TALLINN UNIVERSITY OF TECHNOLOGY

School of Business and Governance

Sohaib Arshad

PERFORMANCE OF AI GENERATED CONTENT IN CONTENT

MARKETING
Master’s thesis
MA International Business Administration

Supervisor: Tarmo Koppel, PhD

Tallinn 2023
I hereby declare that I have compiled the thesis independently
and all works, important standpoints, and data by other authors
have been properly referenced and the same paper
has not been previously presented for grading.

The document length is 8675 words from the introduction to the end of the conclusion.

Sohaib Arshad
09/05/2023
TABLE OF CONTENTS

ABSTRACT ....................................................................................................................................5
INTRODUCTION ...........................................................................................................................6
1. LITERATURE REVIEW ............................................................................................................8
1.1. Content Marketing ................................................................................................................8
1.2 Overview of AI & generative models....................................................................................9
1.3. Role of AI in Marketing .....................................................................................................10
1.4. Performance comparison ....................................................................................................12
1.5. SMCR & AIDA ..................................................................................................................13
1.6. Technology Acceptance Model (TAM) .............................................................................14
2. METHODOLOGY ....................................................................................................................16
2.1. Blog post content experiment .............................................................................................18
2.2. Facebook ads & landing page content effectiveness comparison ......................................20
2.3. Email content effectiveness comparison ............................................................................21
2.4. Online survey for professional writers ...............................................................................22
3. RESULTS ..................................................................................................................................24
3.1. Blog post content experiment results .................................................................................24
3.2. Landing page comparison results .......................................................................................29
3.3. Email results .......................................................................................................................30
3.4. Results for perception of professional writers ....................................................................31
4. DISCUSSION............................................................................................................................35
CONCLUSION .............................................................................................................................38
LIST OF REFERENCES ..............................................................................................................39
APPENDICES ...............................................................................................................................45
Appendix 1. AI generated blog prompt for GPT4 .....................................................................45
Appendix 2. Prompts and content results for emails .................................................................46
Appendix 3. Prompt and content results for AI landing page and its Ad copy .........................50
Appendix 4. Questionnaire for Blog Posts ................................................................................51
Appendix 5. Questionnaire for professional writers ..................................................................53
Appendix 6. Descriptive Statistics for writer’s questionnaire ...................................................56
Appendix 7. Correlation analysis for writer’s questionnaire.....................................................57
Appendix 8. Kruskal-Wallis H for writer’s questionnaire 1 .....................................................58
3
Appendix 9. Kruskal-Wallis H for writer’s questionnaire 2 .....................................................59
Appendix 10. Non-exclusive licence .........................................................................................60

4
ABSTRACT

This study investigates the performance of AI-generated content in comparison to human-written

content. By employing the SMCR&AIDA models, the study evaluates the performance of content
from both sources using marketing performance metrics across emails, blog posts, and landing
pages. Additionally, the study relies on the TAM model and evaluates the perceptions of the
usefulness of AI content among both writers and readers.

Four separate experiments were conducted, including a survey of professional writers to

understand their perception and usage of AI content and tools. The findings indicate that AI-
generated content can be competitive with human-written content in certain contexts and users are
unable to accurately differentiate between the two. Additionally, professional writers find AI
content to be effective, credible, and high-quality while acknowledging its limitations.

From a managerial perspective, the study suggests that content managers and marketers should
consider integrating AI-generated content into their strategies to improve efficiency and cost-
effectiveness. This research contributes to the growing body of literature on AI-generated content
performance and provides a foundation for future investigations in this area.

Keywords: AI-generated content, AI in marketing, performance testing,

5
INTRODUCTION

After the emergence of digital marketing, content marketing has become a key to successful online
campaigns (Blates,2015). However, creating original content for marketing purposes is expensive
and time-consuming because it involves extensive research, data collection, and skillful writing
that provides meaningful information to the audience and keeps them engaged.

On the other hand, the development of advanced natural language processing (NLP) models has
opened up new possibilities for using artificial intelligence (AI) for writing a variety of content as
they enable AI to write like humans (Schweidel, Reisenbichler, Reutterer & Zhang, 2023). For
instance, companies are using AI for making documentaries and creating content for journalism
(Segarra-Saavedra, Cristòfol & Martínez-Sala, 2019).

Moreover, in the context of content marketing, recent research found that tools such as ChatGPT
can assist marketers in generating content more quickly, potentially achieving quality on par with
human content creators (Rivas & Zhao, 2023) while another expert states AI-generated content to
be faster, cost-effective, and with better Search Engine Optimization (SEO) capabilities (Taylor,
2023). Tang (2021) also considered AI algorithms to be excellent at carrying out routine, rule-
based writing tasks more efficiently than human writers.

However, despite the possibilities and the increasing use of AI technology in content marketing,
there is limited research on the performance of AI-generated content in comparison to human-
generated content. The main problem this study addresses is this knowledge gap.

This study aims to investigate the performance of AI-generated content in content marketing
through A/B testing experiments for blog posts, emails, and landing pages in a real-world scenario
using AI-generated and human-written content. Metrics such as click-through rate (CTR),
conversions, open rates, and other relevant metrics (Rancati & Gordini, 2014) for each experiment
are used to determine the performance of related types of content. Additionally, the study

6
administered a survey to gather professional writers' perceptions of the effectiveness and
limitations of AI-written content and tools.

To provide a comprehensive analysis of the results, this study incorporates the AIDA and SMCR
models for understanding the quantitative performance of content and the Technology Acceptance
Model (TAM) for exploring professional writers' and readers' perceptions of AI-generated content
in terms of effectiveness, content quality, and credibility.

The academic contribution of this study is to advance the understanding of AI-generated content's
performance in content marketing across 3 different content types. Managerially, it can help
businesses make more informed decisions about incorporating AI-generated content in their
content marketing strategies. This knowledge can lead to lower content marketing costs, improved
marketing outcomes, such as higher audience engagement and conversion rates, and potentially
more efficient use of resources.

Research Questions:
RQ: What is the performance of AI-generated content in Content Marketing?
Sub-RQ1: How does AI-generated content compare to human-written content in terms of CTR,
conversion rates, engagement rates, and open rates?
Sub-RQ2: How do readers perceive AI content in terms of effectiveness, content quality, and
credibility in comparison to human-written content?
Sub-RQ3: Are readers able to differentiate between AI-generated and human-written content on
the web?
Sub-RQ4: How do professional writers perceive AI-generated content regarding effectiveness,
credibility, and content quality?

7
1. LITERATURE REVIEW

1.1. Content Marketing

In the highly competitive digital landscape of the 21st century, businesses constantly seek effective
ways to stand out and connect with their target audiences. As a powerful tool for achieving this,
content marketing has gained prominence, with a recent survey conducted by Stahl (2022)
revealing that 71% of B2B marketers believe content marketing has grown in importance for their
organization over the past year. Although content marketing has historical roots, it has evolved
considerably over time and now encompasses a wide array of formats and approaches to cater to
diverse industries and objectives (Pulizzi, 2013; Forecey, 2022; Gokhale, 2016; Harris, 2017).

“Content marketing is a strategic marketing approach focused on creating and distributing

valuable, relevant, and consistent content to attract and retain a clearly defined audience, ultimately
driving profitable customer action” (Content Marketing Institute, n.d.). This customer-centric
approach aims to provide value and relevance to audiences. Resultantly, helping businesses
achieve desired results, such as raising brand awareness, generating leads, converting leads into
customers, establishing themselves as industry leaders, engaging with customers, retaining
customers, increasing website traffic, and boosting sales (Gokhale, 2016).

Poradova (2019) asserts that content marketing is the sum of all modern online marketing
strategies, indicating its significance in digital marketing. This evolution has led to the
diversification of content marketing formats, including blog posts, case studies, ebooks, webpages,
community forums, podcasts, videos, infographics, templates, and emails (Forecey, 2022;
Gokhale, 2016; Harris, 2017). However, the choice of content type depends on the industry and
the objectives of the business. For example, a company selling toys for kids may use blog posts
explaining product benefits and videos showing actual products or people using them, while a
brand selling organic skincare products might also employ podcasts, case studies, or community
forums to educate consumers.

8
1.2 Overview of AI & generative models

To comprehend the evolution of generative AI, which enables the production of human-like text,
image generation based on textual inputs, and more, it is essential to examine AI and its associated
sub-fields, including machine learning (ML), deep learning (DL), natural language processing
(NLP), and other generative AI models.

Artificial Intelligence (AI) was first introduced by John McCarthy in 1956 (Bini, 2018) and has
since branched into various sub-fields, including ML and DL (LeCun, Bengio, & Hinton, 2015).
ML focuses on developing algorithms and techniques that allow computers to learn from and make
predictions or decisions based on data without explicit programming (Alpaydin, 2020). As a subset
of ML, DL utilizes multi-layered artificial neural networks to capture complex patterns within
datasets (LeCun et al., 2015).

Generative AI (GenAI) is a subfield of AI that specializes in creating various data types, such as
images, videos, audio, text, and 3D models (GenerativeAI, n.d.). Within GenAI, Natural Language
Processing (NLP) is the area that concentrates on enabling computers to understand, interpret, and
generate human language (Chopra, Prashar, & Sain, 2013). Natural Language Generation (NLG)
is a subtask under NLP that involves creating natural language text that fulfills specific
communication objectives (Dong et al., 2022).

NLP has played a significant role in the development of powerful language models, such as GPT
(Radford et al., 2018), GPT-2 (Radford et al., 2019), GPT-3 (Brown et al., 2020), and the latest
GPT-4 model (OpenAI, 2023). Pre-trained transformer-based models (GPT) trained on extensive
datasets introduced by OpenAI (Radford, Narasimhan, Salimans, & Sutskever, 2018) have been
the real game-changers in the field of AI-generated text.

In the realm of AI image generation, Generative Adversarial Networks (GANs) and Variational
Autoencoders (VAEs) are the most common frameworks used today. GANs involve competition
between two machine learning models that are typically built using neural networks. One generates
fake data similar to real data and the other model tries to differentiate real from fake (Goodfellow
et al., 2020). VAEs use a neural network to understand patterns within input data and create new
images by drawing examples from these identified patterns (Kingma & Welling, 2013). Prominent
AI image generation models include Midjourney V5 (Midjourney, 2023), DALL-E 2 (Marcus,

9
Davis, & Aaronson, 2022), and Stable Diffusion (Rombach, Blattmann, Lorenz, Esser, & Ommer,
2022).

To sum it up, generative AI has evolved rapidly, encompassing various sub-fields that have enabled
advances in AI content generation. Understanding the development and capabilities of these
generative models is crucial for exploring their potential applications in diverse domains, including
content marketing.

1.3. Role of AI in marketing

Over the last decade, AI has emerged as a significant driving force within the marketing sector,
offering immense potential for a variety of tasks ranging from personalizing business
communication to identifying well-performing promotions and uncovering new or missing
information distribution channels (Stone et al., 2020).

AI's anticipated impact on marketing and communications is substantial, with 55% of CMOs
expecting it to be more influential than social media (Weber Shandwick, 2016). With the
widespread adoption of AI, consumer purchase and consumption patterns will be readily available,
making the real business competition about the processing qualities and capabilities of AI systems
(Vishnoi, Bagga, Sharma, & Wani, 2018).

Examples of AI in marketing include Amazon's use of drones for delivery automation through
Prime Air, Macy's use of NLP for in-store personal assistant services, and Replika, a chatbot
offering emotional support by imitating users' communication styles (Huang & Rust, 2021). To
harness AI's potential in marketing, Paul Roetzer, CEO of Marketing AI Institute, introduced the
5Ps of the Marketing AI framework (Roetzer, 2017).

AI's impact on marketing is evident in various functions and approaches, such as campaign
optimization, content personalization, customer segmentation, task automation, content
generation, and sentiment analysis on social media.

Campaign optimization is crucial in digital marketing (Alsoud et al., 2023), allowing marketers to
allocate funds to cost-effective activities that yield desired results. Machine learning-powered tools

10
like Google Analytics provide insights into customer journeys, user behavior, and campaign
performance, enabling marketers to optimize campaigns and maximize return on investment
(Gaur, Singh, & Kumar, 2016).

Personalization is central to modern algorithmic tools, as our purchases, news consumption, music
preferences, and other aspects of daily life increasingly rely on tailored recommendations (Kotras,
2020). For instance, Netflix uses algorithms to provide users with content that aligns with their
preferences (Gomez-Uribe, & Hunt, 2015), while TikTok's recommendation system curates videos
based on various factors, contributing to its success as a social media platform (Boeker, & Urman,
2022).

Advancements in AI have led to the creation of tools that generate new content for marketing
purposes more efficiently than humans, such as writing product descriptions, social media posts,
and SEO-optimized content for higher search rankings (Tang, 2021; Schweidel, Reisenbichler,
Reutterer & Zhang, 2023).

Customer segmentation involves dividing customers into distinct groups based on behavioral
similarities (Kansal, Bahuguna, Singh, & Choudhury, 2018). AI segmentation offers flexibility by
treating each customer as a unique segment or combining dispersed long-tail groups into a single
segment (Huang & Rust, 2021), enabling marketers to tailor messages and product offerings
accordingly.

Sentiment analysis helps marketers understand users' feelings about a company or its products
(Taherdoost & Madanchian, 2023). Companies increasingly use AI-powered sentiment analysis to
comprehend customer feedback, service issues, social media mentions, and employees' feelings
and thoughts (Taherdoost & Madanchian, 2023).

In conclusion, AI has a significant role in shaping modern marketing strategies and offers a myriad
of opportunities for marketers to optimize their campaigns, personalize content, segment
customers, automate tasks, generate content, and analyze sentiments. As AI technology continues
to advance, its impact on marketing will only grow.

11
1.4. Performance comparison

In terms of KPIs, scholars have relied on subjective assessments due to the absence of widely
recognized and consistent objective measures for evaluating the efficacy of content marketing
(Koob, 2021). However, Rancati & Gordini (2014) provided quantitative metrics and categorized
these metrics into four categories; “Consumption, Sharing, Lead Generation, and Sales Metrics”.
The metrics under each category used in the literature are explained in Table 1.

Table 1. Content marketing metrics & their categories

Category metrics
Consumption page views
downloads
visitors
time on page
social chatter
Sharing likes, shares, tweets
email forwards
inbound links
Lead generation form completion
email, newsletter, blog subscriptions
blog comments
cookies
RSS feed
conversion rate
Sales metrics online and offline sales
customer retention
cost saving
content marketing ROI
Source: Adapted from Rancati and Gordini (2014).

In addition to these, experts also use the tone of voice, grammar, writing style, in-depth topic
coverage, and command of language as measures to evaluate the quality of content in marketing
(SEMrush, 2021).

The combination of these metrics can be used to assess the effectiveness of content marketing
campaigns based on their objective. For instance, in case the goal of a campaign is to generate new
leads for an online education company, that company can use metrics such as form completion or
email & newsletter subscriptions to understand whether the campaign was effective or not.

12
1.5. SMCR & AIDA

In the context of this study, two prominent models have been chosen to analyze the performance
of AI-generated content in content marketing in comparison to human-written content. These
models are the Sender-Message-Channel-Receiver (SMCR) model and the AIDA model.

SMCR model also known as Berlo’s model of communication (Muyanga & Phiri, 2020), states
that there are 4 key components of effective communication; source, message, channel, and
receiver. Furthermore, the model highlights the significance of each component in effective
communication and suggests that any disruption or weakness in any of these components can lead
to miscommunication or failure in achieving the desired communication objectives (Berlo, 1960).
In recent years, the model has been used widely in the literature for understanding the dynamics
of marketing and communication studies (Sutanto, 2016; Dhar & Bose, 2016; Pornsuksawat &
Kheokao, 2019; Rubino, Tian, & Pelkki, 2022).

Figure 1. SMCR Model

Source: Berlo (1960)

The AIDA Model stands for A-attention, I-interest, D-desire, and A-action. That outlines a
consumer's cognitive journey that progresses through four primary stages from gaining attention
to taking the desired action (Lewis, 1908). Using this model, marketers can develop effective
marketing campaigns by guiding consumers through four psychological stages, transitioning from
a lack of awareness to tangible purchases (Ghirvu, 2013).

13
 Attention

Interest

Desire

Action

Figure 2. AIDA model

Source: Adoption based on Lewis (1908)

In relation to marketing, the model has been used widely in literature for understanding and
enhancing social media performance (Hassan, Nadzim & Shiratuddin, 2015; Mumtaz, 2019) to
comparing different marketing campaigns (Rehman, Nawaz, Ilyas & Hyder, 2014) and tourism
websites (Prathapan, Sajin Sahadevan & Zakkariya, 2018).

1.6. Technology Acceptance Model (TAM)

The Technology Acceptance Model (TAM), proposed by Davis (1985), is a dominant widely used
theoretical framework for understanding the factors that influence the adoption and use of
technology (Marangunić & Granić, 2015). This model focuses on perceived usefulness and
perceived ease of use as two main determinants of technology acceptance.

Researchers have used this model to explain the adoption of new technologies in both business
and social aspects such as the acceptance of viral marketing (Aghdaie, Sanayei, & Etebari, 2012),
the adoption of e-learning (Ibrahim et al., 2017), and social media usage (Rauniar, Rawski, Yang,
& Johnson, B, 2014). While some researchers found it to be very effective (King & He, 2006),
others found it to be lacking in extensive and thorough research (Chuttur, 2009; Legris, Ingham,
& Collerette, 2003).

14
Figure 3. Technology Acceptance Model
Source: Davis (1985)

The application of TAM in this study offers a structured and theoretically grounded approach to
understanding the acceptance and adoption of AI-generated content among professional writers
and readers. By relating the key factors of interest to the established constructs of Perceived
Usefulness and Perceived Ease of Use, this study provides valuable insights into the potential
barriers and facilitators of AI-generated content adoption in content marketing, contributing to
both academic knowledge and managerial practice.

15
2. METHODOLOGY

The objective of this study is to analyze and understand the performance of AI-generated content
in comparison to human-written content in content marketing. To achieve this objective, the author
has analyzed AI-generated content from different perspectives and using different content types.

Consequently, this study employs a multi-faceted methodology and the research design consists of
4 distinct yet complementary experiments involving diverse participant groups. These experiments
include A/B testing for three different types of content; blog posts, Emails & landing pages, along
with an online questionnaire to professional writers and marketers in order to gather their
perception of the effectiveness of AI-generated content.

A/B testing, also known as controlled experiment and split testing, is a widely accepted method
for measuring the performance of different content variations to determine which version resonates
better with the target audience (Kohavi, Henne, & Sommerfield, 2007).

In content marketing, it has been used widely to compare two or more versions of elements such
as headlines, social media posts, emails, web versions, and more to understand which one performs
better (Yang & Hayashi, 2021; Feit & Berman, 2019; Turunen, 2021; Hagar & Diakopoulos, 2019).
Figure 1. shows the basic structure for A/B testing by Yang & Hayashi (2021).

Figure 4. The basic structure of A/B testing.

Source: Yang & Hayashi (2021)

16
For this research, A/B testing will be used to compare the performance of different types of content
in content marketing such as emails, blog posts, and landing pages based on a variety of marketing
metrics. The structure for that is shown in Figure 5.

Human written
content (control
Group A group) KPI’s
Bano Analysis
Herbal’s & results
Audience AI generated content
Group B (treatment group) KPI’s

Figure 5. A/B testing structure for this study

Source: Author’s adaptation of A/B testing structure provided by Yang & Hayashi (2021)

To allow for direct comparison as per the A/B testing method, each experiment – Blog posts,
Facebook Ad + Landing page, and Email - has two versions. One version is written by AI while
the second version is written by a human. All the content for these 3 experiments has been
developed following the brand’s target audience, messaging, and values.

In order to ensure writers do not use AI to write content for the experiments, clear instructions as
well is the intent for which the content will be used were provided. Additionally, all the content
meant to be written by human writers was assessed using AI content detection tools and it was
ensured that the content meets the requirements for the experiments.

The rationale behind choosing different types of content (blog posts, emails, landing pages) and
conducting 3 different experiments includes the following reasons;
1. Each of these content types serves a specific purpose in the overall content strategy.
Therefore, using different types of content to assess the effectiveness can help to identify
the strengths and weaknesses of AI-generated content in different contexts.
2. To gain a more comprehensive understanding of the effectiveness of AI-generated content
in content marketing
3. To reduce potential biases and limitations associated with using a single approach

17
Furthermore, for the first 3 experiments, Bano Herbal which is a small home-based herbal skin
care products manufacturer from Pakistan was used to ensure that the research is anchored in the
real-world context and that the findings are more relevant and applicable to content marketing
practices.

Table 2. The research design for the performance of AI-generated content in content marketing

Object activity target group metrics

Performance blog posts + Bano herbal’s target engagement metrics
comparison questionnaires audience proxies questionnaire
Performance ads + landing page potential customers CTR
comparison who have interacted Avg. time on page
online with Bano’s engagement rate
previous ad conversion rate
Performance promotional emails Bano herbal’s email open rate
comparison subscribers CTR
conversion rate
Perception about the online questionnaire professional writers questionnaire
effectiveness of ai
generated content
Source: Author’s Analysis

By implementing this comprehensive framework, the goal is to obtain an in-depth understanding

of the performance of AI-generated content in the context of content marketing.

In the following sections, the author has outlined each experiment, providing information on the
procedures used, the participants involved, and the data collection & analysis techniques that were
applied. This comprehensive overview of the methodology will make it easier to replicate the study
for suture researchers. Furthermore, it will promote transparency throughout the research process,
and enable a thorough evaluation of the results.

2.1. Blog post content experiment

To understand the effectiveness of AI-generated blog posts, A/B testing was performed using 2
blog posts, one written by a human writer and one by AI.

As a part of the experiment, a professional writer was hired and provided with a basic overview of
the topic relevant to Bano Herbal’s audience rather than giving an outline of what to include. On
the other hand, ChatGPT with GPT4 (the latest natural language processing model developed by
18
OpenAI that was released on March 14, 2023) was used and provided with a base prompt that had
similar information to what was provided to the writer. The prompt also included information about
the writing style, word limit, and other structural information to ensure the AI written content is in
the form of a blog post and has an appropriate structure.

Once both versions of the blog were ready, they were published on Bano Herbal’s website and
were presented to 40 different people with similar characteristics to the target audience of Bano
Herbal via links inside the questionnaire. That allowed the author to not only measure responses
about the blog’s effectiveness, content quality, and credibility but also made it possible to measure
qualitative metrics such as engagement rate, events per session, event count, etc. The UTM
parameters were used to differentiate between the traffic from both blog posts. Though it helped
to achieve the objective, however, small discrepancies were observed in the collected data which
point out the limitations of UTM parameters for tracking.

The questionnaire was divided into five sections with the first section comprising two demographic
questions (e.g., age and gender). Sections 2, 3, and 4 each contained four questions focused on
effectiveness and its two sub-sets; content quality and credibility, using the Likert scale (Kim et
al., 2020; Haim & Graefe, 2017; Liaw & Huang, 2013). The last section had only one question,
which asked participants to rate their perception of the author, also measured by the Likert scale.
Furthermore, the questionnaire had two versions with the same questions but each led to a different
blog post.

All the respondents were randomly presented with one blog post using QR codes in a way that in
the end, 20 people received the AI blog post and 20 received the blog post that was written by a
human. Once they had read it, they were asked to share their perception of the effectiveness,
content quality, credibility, and likelihood of content being written by humans or AI via the
questionnaire. None of the readers knew beforehand whether the blog post presented to them is
written by a human writer or AI.

On the one hand, this approach allowed the author to collect interaction data, such as page views
and average session duration, for both blog posts. On the other hand, it enabled the author to gather
valuable insights from readers about the effectiveness, credibility, and content quality of both blog
posts through the questionnaire, providing us with data for comparison.

19
The subsequent statistical tests were utilized to examine the data; independent sample t-test, and
chi-square test.

To compare average ratings concerning effectiveness, content quality, and credibility between AI-
generated and human-written content, the Independent Samples t-test was employed. This test was
chosen due to its ability to compare the means of two distinct groups—those exposed to AI-
generated content and those exposed to human-authored content.

The Chi-square test was applied to assess whether participants could differentiate between AI-
generated and human-authored content. The rationale for choosing this test lies in its ability to
analyze the association between two categorical variables: the respondents' perception of the
content's author and the actual author.

2.2. Facebook ads & landing pages performance comparison

While the audience’s perception of AI-generated content and being able to differentiate is one
thing, understanding how the users interact with the content when presented to them in a real-
world scenario is another. Therefore, the idea of A/B testing was applied to landing pages to assess
the effectiveness of AI-generated content against human-written content in terms of consumption
and sales metrics (Rancati & Gordini, 2014).

Furthermore, an underperforming product “Onion Oil” was chosen rather than a well-known
product for two reasons; to mitigate the biases and the potential to create a higher impact. And
since the product already had a product page with human-written content from Bano Herbal’s in-
house writer, content for the second landing page was created using ChatGPT with the GPT4
model. All instructions provided by the AI were followed to the point except for image/video
suggestions (due to the unavailability of such content for the product) and client testimonials.

In addition to that, to bring users to the product pages and compare their effectiveness, A/B testing
ads were used where two ad copies were created for Facebook ads, following the same human vs
AI principle. Both ads had similar visuals and were connected to their respective pages.
Additionally, the A/B testing ad was created with the objective to drive traffic to the landing pages.

20
Both ads were targetting different audience segments to minimize the chance same users seeing
both ads.

Moreover, UTM parameters for both landing pages with different campaign and source names
were created. That was done to ensure that the traffic to those pages from ads and the website
traffic coming to these pages from other sources can be differentiated and that only those visitors
that came through the ads to landing pages are counted. Also, to address the cookies tracking
limitations, final data from google analytics was considered for analysis.

2.3. Email content performance comparison

In order to evaluate the performance of AI-generated email content in comparison to human-

written emails, an experiment was conducted using samples of Bano Herbal's audience. The
participants in this study received 2 sets of emails and 4 promotional emails in total. For each set,
one of the emails was crafted by an experienced human writer, while the other was generated by
AI.

For the first set, the selected promotion was “free delivery” and the chosen sample consisted of
494 people who had not yet made any purchases but had subscribed to receive emails from the
company. For the second email group, a bundle deal was selected offering 3 products at a
discounted price and with free delivery. This email was sent to a sample of 1025 people who had
subscribed to receive emails from the company.

Additionally, for every email, all aspects including the email subject and preview text was also
written by the respective writers of the email, whether AI or human. Furthermore, to ensure
independence, each email was sent to users after a specified period.

The experiment aimed to determine whether there were any significant differences in the
performance of the AI-generated email content compared to the human-written email content. The
total number of emails delivered, opened, clicked, and converted were recorded for both the AI-
generated and human-written emails. The open rate, click rate, and conversion rate were calculated
for each email to compare their performance.

21
2.4. Online survey for professional writers

In addition to carrying out experiments to evaluate the performance of AI-generated content, an

online survey was conducted targeting professional writers. The questionnaire aimed to obtain
insights into various aspects of AI-generated content, delving into the writers' viewpoints on the
effectiveness, credibility, content quality, and limitations of such content, as well as their
experiences with AI-assisted content generation tools in order to evaluate it using the TAM model.

This questionnaire contained 24 questions, divided into seven sections. The first section inquired
about demographics, such as age, gender, location, and education. The second section addressed
experience and industry, while the third section focused on any AI tools the respondents may have
used and how often they utilized them. Lastly, sections 4 through 7 explored the effectiveness,
credibility, content quality (Kim et al., 2020; Haim & Graefe, 2017; Liaw & Huang, 2013) and
limitations of AI-generated content and tools.

To ensure a diverse sample of respondents, the questionnaire was initially shared using Facebook
groups for content writers, online forums, direct messages to writers via LinkedIn, and personal
connections. However, the response rate was quite low and resulted in only 18 responses.

To improve the response rate, the author reached out to admins of online communities & Facebook
groups and sent reminders to personal connections after one week after the initial contact. As a
result, the number of respondents increased significantly to a total of 111. However, 31 respondents
left the survey halfway and only 80 people completed it. The overall method ensured that the study
includes respondents from different industries and with a wide range of experiences from 19
different countries.

Table 3. Demographic profile of survey respondents

Variable N (%)
Gender
Men 54 (67.5%)
Women 26 (32.5%)
Education
High school diploma or equivalent 4 (5.0%)
Bachelor's 47 (58.8%)
Master's 22 (36.3%)
Source: Author’s calculations

22
 Distribution of respondents based on industry
60.00%
48.80%
Percentage of respondents

50.00%

40.00% 35%

30.00% 27.50%
21.30% 20%
20.00%

10.00%

0.00%
Technology Education Entertainment Helathcare Travel &
Hospitality
Industries

Percentage

Figure 6. Distribution of respondents based on industry (considers writers who have written for
multiple industries).
Source: Calculated by the author

From the above table, it is evident that a majority of the professional writers (67.5%) who
responded to our survey are men whereas 32.5% of respondents are women. Moreover, all the
writers either have a High school, Bachelor’s, or Master’s degree. However, the number of writers
with a Bachelor’s degree (58.5%) is significantly higher than the ones with a Master’s degree
(36.3%). Only 5% of the respondents have a High school or equivalent degree.

Lastly, in the context of industry, 48.8% of writers write for the technology industry, followed by
education at 35%, entertainment at 27.50%, health care at 21.30%, and travel & hospitality at
23.6%.

Furthermore, for easier analysis, the collected data was coded and an average was taken for Likert
scale responses of all variables related to effectiveness, credibility, content quality, and limitations
separately to create these new variables for each metric and to conduct statistical tests using SPSS.
These tests included descriptive statistics, Spearman’s correlation, and the Kruskal-Wallis H test.
The purpose of these tests was to determine the relationships between the variables and understand
how they affect each other.

23
3. RESULTS

3.1. Combined conceptual model of SMCR & AIDA

In the context of this study, the source in SMCR represents AI-generated content and human-
written content. The distinction between these sources allows for an investigation into the impact
of altering the origin of content on the overall effectiveness of content marketing campaigns. As
for the message part, it represents the informational/promotional messages of the Bano Herbal
brand. The channel part represents the channels used to deliver those messages such as email,
website landing page, etc. and the receiver part represents the audience of the message.

The integration of the AIDA model into the SMCR model at the receiver stage then measures the
impact of the message from both sources using KPIs at each stage of the AIDA model, evaluating
how the change in the source of content (AI-generated vs. human-written) influences the
performance.

Figure 6. Integration of SMCR & AIDA model for this study

Source: Author’s adaptation of SMCR & AIDA models

24
Performance KPIs related to each AIDA stage are as follow:
• Attention stage - open rates, page views
• Interest stage - click-through rates (CTR), time spent on the page, page views, engagement
rate
• Desire stage - Return visits, Number of clicks on call-to-action (CTA)
• Action stage, conversion rates
This combined approach is expected to offer a comprehensive understanding of the quantitative
performance of AI-generated content in comparison to human-written content in content marketing
and contribute to the academic and managerial implications of the research.

3.2. TAM conceptual model for perception of writers and readers

In this study, Technology Acceptance Model (TAM) has been used as a theoretical framework to
understand the perceptions of professional writers and readers regarding AI-generated content. The
model's key constructs, Perceived Usefulness (PU) and Perceived Ease of Use (PEOU) have been
adapted to capture the specific factors relevant to written content in content marketing.

Figure 7. TAM for the perception of professional writers

Source: Author’s adaptation of the Technology Acceptance Model for the perception of
professional writers

25
In the context of this study, TAM is used to examine the perceptions of professional writers
regarding the effectiveness, credibility, content quality, and limitations of AI-generated content,
providing insights into their acceptance and potential adoption of AI-generated content in their
work. For professional writers, Perceived Usefulness includes their perceptions of the
effectiveness, content quality, credibility, and limitations of AI-generated content. These factors
represent the writers' overall evaluation of the usefulness of AI-generated content in meeting their
objectives.

Figure 8. TAM for the perception of readers

Source: Author’s adaptation of TAM for the perception of readers

For blog posts, TAM is used to examine the perception of readers about effectiveness, content
quality, and credibility for both blog posts as to whether there is a difference and its impact on
satisfaction. In the case of readers/potential customers, Perceived Usefulness captures their
perception of the effectiveness, quality, and credibility of AI-generated content. By analyzing
readers' evaluations of AI-generated content, it is possible to understand how they perceive the
utility of such content in meeting their informational and entertainment needs and whether engage
with the content and find it satisfactory or not.
26
3.3. Blog post content experiment results

The analysis of the data from blog posts data provided the author with significant insights related
to the performance of AI-generated blog posts in comparison to human-written blog posts. Since
the blog post data was measured from both an analytics perspective and via a questionnaire, it was
analyzed separately to draw significant conclusions. Table 4 provides results from the google
analytics platform for both blog posts.

Sub-RQ1:How does AI-generated content compare to human-written content in terms of CTR,

conversion rates, engagement rates, and open rates?

Table 4. Google analytics results for blog posts.

Content source average engaged events per engagement event count

engagement sessions per session rate %
time per user
session
AI-generated 134.42 1.05 8.58 100% 163
Human-written 87.67 1.12 6 90.48% 150
Source: Author’s calculation

Based on these results, AI-generated blog post had 53.3% higher engagement time, 9.52% higher
engagement rate, and more events per session compared to the human-written blog post. Since
both of the blog posts were meant to be informational, their conversion rates were 0%. From
AIDA's perspective, higher engagement time, more events per session, and higher engagement
rate suggest that AI-generated content was more effective at capturing readers’ attention and
maintaining their interest.

Sub-RQ2: How do readers/customers perceive AI content in terms of effectiveness, content

quality, and credibility?

To answer Sub RQ2, independent Samples t-tests were conducted to compare average ratings for
effectiveness, content quality, and credibility between AI-generated and human-written content.
The results of the Independent Samples t-tests for effectiveness, content quality, and credibility are
presented in Table 5.

27
Considering the significance level (0.05) for the t-test, there were no significant differences
between the AI-generated and human-written blog posts in terms of effectiveness (p = 0.638),
content quality (p= 0.844), and credibility (p = 0.089).

Table 5. Independent samples t-test for effectiveness, content quality, and credibility of blog posts

Variable group mean standard degrees of t-value p-value

deviation freedom (df) (significance)
Effectiveness AI 3.8500 0.72275 38 -0.475 0.638
Human 3.9625 0.77492
Content AI 4.5000 0.65895 38 0.198 0.844
Quality
Human 4.4625 0.53358
Credibility AI 3.6750 0.92871 38 1.746 0.089
Human 3.1250 1.05911
Source: Calculated by the author

These results indicate that the perceived difference in ratings from readers for effectiveness,
content quality, and credibility of AI-generated blog post and human-written blog post is not
significant. From the lens of TAM, these survey results suggest that readers found both blog posts
similarly useful in terms of effectiveness, content quality, and credibility as there is no significant
difference in the results.

Sub-RQ3: Are readers able to differentiate between AI-generated and human-written content on
the web?

To answer Sub-RQ3, a Chi-square test was conducted to assess if participants could differentiate
between content created by AI and content written by humans. The test compared their perception
of the blog post's author (whether they believed it was written by a human) to the actual author
(AI-generated or human-written).

Table 6. Chi-square test results for respondents' perception of the author

Test name degrees of chi-square value p-value

freedom (df) (χ²) (significance)
Chi-square test 4 5.399 0.249
Source: Author’s calculations

The Chi-square test found no significant correlation between perceived and actual authorship, χ²(4,
N=40) = 5.399, p = .249. This suggests that participants were not able to accurately distinguish
28
between AI-generated and human-written blog posts, and this is consistent with previous studies
(Köbis & Mossink, 2021; Graefe et al., 2018).

3.4. Landing page comparison results

The analysis of the collected data revealed several insights about the performance of human-
written and AI-generated content in terms of engagement and conversion rates for landing pages.
The key metrics for each content source are summarized in the table below:

Table 7. Human VS AI landing pages key metrics results

Content source engagement rate conversion rate avg. engagement events per
time (s) session
Human-written 84.18% 0.60% 19.21 8.72
AI-generated 42.62% 0.00% 19.20 6.74
Source: Author’s calculations

The human-written content outperformed the AI-generated content with a 97.37% higher
engagement rate. The engagement rate for the human-written content was 84.18%, compared to
the AI-generated content's engagement rate of 42.62%. The conversion rate for the human-written
content was 0.60%, while the AI-generated content had a conversion rate of 0%. The results from
both of these metrics show that a human-written landing page was more effective than an AI-
generated landing page.

While analyzing the average engagement time per session, there wasn't any significant difference
between content written by a human writer and that generated by AI. This indicates that users spent
roughly the same amount of time on each landing page, irrespective of whether the content was
written by a person or an AI.

In addition, the metric for events per session demonstrated that users engaged more with human-
written content, boasting an average of 8.72 events per session. In contrast, AI-generated content
averaged 6.74 events per session. This suggests that users found the human-written content on the
landing page more engaging than the one generated by AI.

In conclusion, the human-written content demonstrated superior performance in terms of

engagement rate, conversion rate, and user interaction compared to the AI-generated content.
29
There was no significant difference in the average engagement time per session between the two
content types. From the lens of SMCR & AIDA, human-written landing page performed better on
all stages of AIDA as compared to AI-generated content. This information addresses Sub-RQ1 of
the study, focusing on the comparison between human-written and AI-generated content in terms
of engagement and conversion rates for landing pages.

3.5. Email results

Both emails for the email experiment (human-written and AI-generated) were sent using the email
app within Shopify for Bano Herbal. The findings from the email experiment are summarized in
the table below:

Table 8. A summary of findings from the email experiment

delivered opened open clicked click- converted conversion

rate through rate
-rate
Human_email_1 481 84 17.70 4 0.80% 0 0%
%
AI_email_1 478 113 23.60 3 0.63% 1 33.30%
%
Human_email_2 986 129 13.1% 4 0.41% 0 0%
AI_email_2 996 203 20.4% 4 0.4% 1 25%
Source: Author’s calculation

The first AI-generated email was delivered to 478 customers and had an open rate of 23.6%.
Against that, the human-written email was delivered to 481 customers and had a slightly lower
open rate of 17.46%. This suggests that the AI-generated email was more effective in capturing
the recipients' attention from AIDA & SMCR perspectives and had a 33.3% higher open rate.

In terms of CTR (click-through-rate), the human-written email had a CTR of 0.80%, while the AI-
generated email had a CTR of 0.63%. This indicates that the human-written email was slightly
more successful in driving customer engagement and capturing customer’s interest and desire once
the email was opened as AI-generated email had a 21.25% lower click-through rate.

As for conversion rate, the AI-generated email had a conversion rate of 33.3%, indicating that one
out of the three customers who clicked on the email took the desired action. In contrast, the human-

30
written email had a conversion rate of 0% which indicated that none of the four customers who
clicked on the email made a purchase.

The second AI-generated email was delivered to 996 subscribers and had an open rate of 20.4%.
On the other hand, human-written email was delivered to 986 people and had an open rate of
13.1%. Overall, AI email had a 55.73% higher open rate suggesting that AI-generated subject lines
and preview text are more effective than human-written counterparts. From the lens of AIDA, AI-
generated email was more effective in capturing the audience’s attention.

As for CTR, both emails performed almost the same with human-generated email displaying a
CTR of 0.41% and AI-generated email having a CTR of 0.4%. However, when it comes to
conversion rates, human-written emails showed a conversion rate of 0% as none of the customers
who opened the emails made a purchase. On the other hand, AI-generated email resulted in 1
conversion, displaying a conversion rate of 25%. From the SMCR & AIDA point of view, AI-
generated email was found to be either more effective or performed similarly to human-written
email on almost all AIDA stages.

However, for both CTR and conversion rates, the number of people who clicked and converted is
too small. That makes it difficult to draw any significant conclusion. And even though the available
data suggests that the human-written email had a higher CTR in email group 1 and the AI-
generated email had higher conversion rates for both email groups, the limited number of clicks
and conversions calls for further research and larger samples to draw more significant conclusions.

3.6. Results for the perception of professional writers

Based on the survey of professional writers conducted to understand the perception of professional
writers for the effectiveness, credibility, content quality, and limitations of AI-generated content
and tools, the author coded the data and analyzed the results using SPSS. Testing methods used for
analysis included descriptive analysis, Spearman’s correlation, and the Kruskal-Wallis H test.

The results from descriptive statistics for the frequency of AI tools usage are presented in Table 9,
showing the frequency of using AI tools among professional writers.

31
Based on these results, 18.2% of writers use AI tools daily while 34.5% use them every week.
Additionally, 34.5% of the writers hardly every use AI tools for content generation as they only
use it on a monthly or rare need basis. The number of respondents who have never used AI tools
is only 12.7%.

Table 9. Usage frequency of AI tools among professional writers

Frequency of use N (%)

Daily 10 (18.2%)
Weekly 19 (34.5%)
Monthly 8 (14.5%)
Rarely 11 (20%)
Never used AI tools 7 (12.7%)
Source: Author’s Calculations

On the other hand, as illustrated in Figure 9, a significant majority of content writers (83.60%)
said they have used ChatGPT. This indicates ChatGPT to be the most popular AI tool among
professional writers who are using AI tools. Other AI tools such as Copy.ai (23.60%), Word Hero
(18.20%), and Jasper.ai (14.50%) are also being used by writers but their percentage is
considerable lower.

AI Tool Usage Percentage Among Professional Writers who

have used AI tools
90.00% 83.60%
80.00%
Percentage of writers using tool

70.00%

60.00%

50.00%

40.00%

30.00% 23.60%
18.20%
20.00% 14.50%

10.00%

0.00%
ChatGPT Copy.ai Word Hero Jasper.ai (Jarvis)
AI Tools

Percentage of users

32
Figure 9. AI tools usage percentage among professional writers who have used AI tools
Source: Author’s calculations

The descriptive statistics results for effectiveness, credibility, and content quality variables (Table
10) show that, on average, respondents hold a fairly positive opinion about the effectiveness of AI-
generated content (mean = 3.7844). However, they gave slightly lower ratings for credibility (mean
= 3.6344) and content quality (mean = 3.5813). As a whole, the results from the first three variables
are positive. Interestingly, writers also rated the limitation of AI-generated content positively
(mean = 3.4656). This indicates that while they find the AI content to be effective, they also
acknowledge that AI-generated content might have some limitations such as misleading content
or potential struggle with conveying complex topics.

Table 10. Descriptive statistics for effectiveness, credibility, content quality, and limitations.

Descriptive statistics
N minimum maximum mean std. deviation

Effectiveness 55 1.00 5.00 3.7844 1.01939

Credibility 55 1.00 5.00 3.6344 1.00862

Content quality 55 1.00 5.00 3.5813 0.90163

Limitations 55 1.00 5.00 3.4656 0.91726

Source: Author’s calculations from results of Appendix 6

To dig deeper, Spearman’s correlation and Kruskal-Wallis H test were also conducted by the
author. The correlation analysis involved the new variables of effectiveness, content quality,
credibility, limitations (averages of Likert scale responses), frequency of usage, and years of
experience.

The results from Spearman’s correlation analysis (appendix. 6) showed no significant correlation
between experience and variables related to AI-generated content. However, between frequency
of usage and AI content variables of effectiveness (correlation coefficient = 0.327, p = 0.003),
credibility (correlation coefficient = 0.299, p = 0.007), and content quality (correlation coefficient
= 0.305, p =0.006), it shows a moderate positive correlation which is statistically significant (p <
0.05).

33
In addition to that, it also shows that effectiveness has a strong positive correlation with both
credibility (correlation coefficient = .651, p < .001) and content quality (correlation coefficient =
.671, p < .001). This means that writers who perceive AI content as effective positively also tend
to perceive content quality and credibility as positive. In other words, it means if the perception of
the effectiveness of AI-generated content increases, it also tends to increase the credibility and
content quality of the AI-generated content for writers. In addition to that, credibility is also
positively correlated with content quality (correlation coefficient = 0.734, p < .001), indicating a
strong positive correlation.

As for the Kruskal-Wallis H test, it involved the new variables of effectiveness, content quality,
credibility, and limitations (averages of Likert scale responses) with gender and education as
grouping variables (appendix 6). The results from the gender test indicate that there is no
significance difference in perception of effectiveness, content quality, and limitations. However,
for credibility (Kruskal-Wallis H = 4.822, sig = 0.028), there is a significant difference in the
perception of both genders. The results from the education group indicate no significant difference
between the perception of writers for effectiveness, content quality, credibility, and limitations
based on their education.

From the perspective of the TAM framework, the results from descriptive statistics (Table 10.)
suggest that professional writers perceive AI-generated content to be fairly useful but with some
limitations. And since a majority of writers are using AI tools to some extent, it indicated their
adoption behavior as per the TAM framework.

34
4. DISCUSSION

The findings from this study provide valuable insights into the performance of AI-generated in
comparison to human-written content. By analyzing the data from professional writers'
questionnaires, blog posts experiment, email experiments, and landing pages, we can draw several
conclusions.

The AI-generated blog post demonstrated higher engagement time, engagement rate, and events
per session compared to the human-written blog post. However, there was no significant difference
in readers' perception of effectiveness, content quality, and credibility between the two types of
content. This suggests that AI-generated content can be as effective as human-written content in
terms of engaging readers and providing useful information. Additionally, the Chi-square test also
showed that readers were not able to accurately differentiate between AI-generated and human-
written content, which implies that AI-generated content can effectively mimic human writing.

For the landing page comparison, human-written content performed better than AI-generated
content in terms of engagement rate, conversion rate, and user interaction. However, there was no
significant difference in the average engagement time per session between the two content types.
These results show that human-written content is more effective than AI-generated content for
landing pages, particularly in capturing user attention and driving users to action.

In both email experiment groups, AI-generated content had higher open rates than human-written
content, suggesting a better ability to capture recipients' attention. However, the CTR and
conversion rates despite AI showing a higher tendency to drive users to take desired action in both
emails, were inconclusive due to the small number of clicks and conversions. This requires further
research with larger samples to draw more significant conclusions.

Lastly, the professional writers' questionnaire results suggest that writers perceive AI-generated
content to be useful and rated it positively in terms of effectiveness, credibility, and content quality.
Additionally, they also acknowledge the limitations such as lack of creativity & originality, and a

35
higher chance of misleading information. Furthermore, he results from the Kruskal-Wallis H test
indicate a significant difference between gender and credibility. Spearman's correlation test
revealed a significant correlation between the frequency of AI tools usage and content quality,
effectiveness, & credibility variables indicating that the more writers use AI to generate content,
the more they find it useful.

This study adds to the growing body of knowledge on AI-generated content performance in
comparison to human-written content and provides evidence on the performance of AI-generated
content across emails, blog posts, and landing pages. This information can be used as a foundation
for future research in this field. Furthermore, the study broadens the application of the SMCR,
AIDA, and TAM models in the context of AI-generated content.

The findings provide valuable insights into the performance of AI-generated content and the
perceptions of writers and readers about its usefulness. This can be further investigated in future
research to better understand the factors related to AI-generated content. Additionally, future
researchers can compare the performance of AI content with human-written content, and a third
comparison group of AI-human-written content to explore the performance and effectiveness of
each.

From a managerial standpoint, this study offers various practical implications. First, content
managers and marketers should consider integrating AI-generated content into their strategies, as
it has been demonstrated to be competitive with human-written content for some content types.
This can help companies produce content more efficiently and cost-effectively as AI generates
content faster and at lower costs than humans. Additionally, since users are unable to accurately
differentiate between AI and human-written content, managers can use AI-generated content
without negatively impacting their brand image or the credibility of the content. Lastly, given the
limitations of AI content, managers can offer suitable training and support for writers who might
be collaborating with AI tools to create content, aiding them in better understanding and utilizing
the technology.

Aside from these academic and managerial implications, the study also has some limitations.
Firstly, the sample size for some experiments, particularly email experiments, was small, with only
four emails sent in total. This resulted in smaller conversion rate numbers and limited the ability

36
to draw significant conclusions. Future research should consider using larger sample sizes to
strengthen the findings.

Secondly, the study focused on a limited number of content types, such as email, blog, and landing
page, which might not cover the entire range of content produced by AI and humans. Future
research can explore other content types and formats to provide a more comprehensive
understanding of AI-generated content performance. Lastly, for Facebook Ads and landing pages,
UTM parameters were used to track users, which can lead to discrepancies between platforms and
might not accurately track data across different devices.

Other than that, the study also encountered some technical limitations such as the results from
landing pages and blog posts obtained from Google Analytics were not individual results but
overall results which limited the scope of analysis.

37
CONCLUSION

The primary objective of this study is to understand the performance of AI-generated content in
comparison to human-written content along with assessing the perception of professional writers
and readers. The analysis conducted for the study provides valuable insights into the performance
of AI-generated and human-written content across 3 different content types from the lens of SMCR
and AIDA, along with the perception of professional writers and readers related to effectiveness,
content quality, and credibility of the AI content from the perspective of TAM.

While there were no significant differences in perceived effectiveness, content quality, and
credibility for blog posts among readers, AI-generated blog performed better than human-written
blog in terms of engagement rate, events per session, and average engagement time per session.

For landing pages, human-written content outperformed AI-generated content and had a 97.37%
higher engagement rate, making it more successful in capturing the audience’s attention and desire.

On the other hand, the results for email experiments showed AI content to be more effective in
capturing attention and bringing users to the desired action but were inconclusive due to small
numbers and warrant further investigation with larger sample sizes. The survey of professional
writers about the perception of effectiveness, content quality, credibility, and limitations found that
writers consider AI-generated content to be effective, high quality, and fairly credible while
acknowledging its limitations.

Practically, this information can help marketers and business managers to make informed decisions
about the incorporation of AI-generated content in overall content strategy and its possible impact.
Future research can further explore the potential of AI-generated content in various content types
and contexts, as well as potential moderating factors that may influence its performance.

38
LIST OF REFERENCES

Aghdaie, S. F. A., Sanayei, A., & Etebari, M. (2012). Evaluation of the consumers' trust effect
on viral marketing acceptance based on the technology acceptance model. International
Journal of Marketing Studies, 4(6), 79.

Alpaydin, E. (2020). Introduction to machine learning. MIT Press.

Alsoud, M., Sharari, H., Helalat, A., Abuhjeeleh, M., Trawnih, A., Mahrakani, N., & Alsoud, M.
(2023). USING ARTIFICIAL INTELLIGENCE MARKETING TO OPTIMIZE
CUSTOMER SATISFACTION IN THE HOSPITALITY INDUSTRY. Journal of
Southwest Jiaotong University, 58(2).

Arshad, S. (2023). Onion Oil - DIY Nature’s Magical Antidote for Hair Loss. Supplementary
data for the study on Performance of AI-generated content in content marketing
[human-written blog]. Bano Herbal website. https://banoherbal.com/blogs/news/onion-
oil-diy-nature-s-magical-antidote-for-hair-loss

Arshad, S. (2023). Regrow Tonic - Onion Oil. Supplementary data for the study on Performance
of AI-generated content in content marketing [default landing page for the product].
Bano Herbal website. https://banoherbal.com/blogs/news/onion-oil-diy-nature-s-
magical-antidote-for-hair-loss

Arshad, S. (2023). Say Goodbye to Hair Loss, Embrace a New Era of Hair Care. Supplementary
data for the study on Performance of AI-generated content in content marketing [AI
landing page]. Bano Herbal website. https://banoherbal.com/pages/say-goodbye-to-hair-
loss-onion-oil

Arshad, S. (2023). Unraveling the Mysteries of Hair Thinning: Causes, Prevention, and the
Game-Changing Role of Onion Oil. Supplementary data for the study on Performance
of AI-generated content in content marketing [AI-generated blog]. Bano Herbal
website. https://banoherbal.com/blogs/news/hair-thinning-causes-prevention-role-of-
onion-oil

Baltes, L. P. (2015). Content marketing-the fundamental tool of digital marketing. Bulletin of the
Transilvania University of Brasov. Economic Sciences. Series V, 8(2), 111.

Berlo, D. (1960). The Process of Communication. Holt, Rinehart and Winston Inc., New York.
Bini, S. A. (2018). Artificial intelligence, machine learning, deep learning, and cognitive
computing: what do these terms mean and how will they impact health care?. The
Journal of arthroplasty, 33(8), 2358-2361.

Bock, P. (1985). The emergence of artificial intelligence: Learning to learn. AI Magazine, 6(3),
180-180.

39
Boeker, M., & Urman, A. (2022, April). An Empirical Investigation of Personalization Factors
on TikTok. In Proceedings of the ACM Web Conference 2022 (pp. 2298-2309).

Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., ... & Amodei, D.
(2020). Language models are few-shot learners. Advances in neural information
processing systems, 33, 1877-1901.

Chopra, A., Prashar, A., & Sain, C. (2013). Natural language processing. International journal of
technology enhancements and emerging engineering research, 1(4), 131-134.

Chuttur, M. (2009). Overview of the technology acceptance model: Origins, developments and
future directions.

Content Marketing Institute. (n.d.). What is content marketing? Retrieved March 31, 2023, from
https://contentmarketinginstitute.com/what-is-content-marketing/

Davis, F. D. (1985). A technology acceptance model for empirically testing new end-user
information systems: Theory and results (Doctoral dissertation, Massachusetts Institute
of Technology).

Dhar, S., & Bose, I. (2016). Framework for using new age technology to increase effectiveness
of project communication for outsourced IT projects executed from offshore. In E-Life:
Web-Enabled Convergence of Commerce, Work, and Social Life: 15th Workshop on e-
Business, WEB 2015, Fort Worth, Texas, USA, December 12, 2015, Revised Selected
Papers 14 (pp. 207-211). Springer International Publishing.

Dong, C., Li, Y., Gong, H., Chen, M., Li, J., Shen, Y., & Yang, M. (2022). A survey of natural
language generation. ACM Computing Surveys, 55(8), 1-38.

Feit, E. M., & Berman, R. (2019). Test & roll: Profit-maximizing A/B tests. Marketing Science,
38(6), 1038-1058.

Forsey, C. (2022). How to Develop a Content Strategy in 7 Steps: A Start-to-Finish Guide.

HubSpot. https://blog.hubspot.com/marketing/content-marketing-plan

Gaur, L., Singh, G., & Kumar, S. (2016, March). Google Analytics: A Tool to make websites
more Robust. In Proceedings of the Second International Conference on Information
and Communication Technology for Competitive Strategies (pp. 1-7).

GenerativeAI (n.d.). Retrieved from https://generativeai.net/

Ghirvu, A. I. (2013). The AIDA model for advergames. The USV Annals of Economics and
Public Administration, 13(1 (17)), 90-98.

Gokhale, N. (2016). Content Marketing-From Genesis to Future. Journal of Indian Management

Research and Practices, 1(01), 13-23.

40
Gomez-Uribe, C. A., & Hunt, N. (2015). The Netflix recommender system: Algorithms, business
value, and innovation. ACM Transactions on Management Information Systems
(TMIS), 6(4), 1-19.

Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., ... & Bengio,
Y. (2020). Generative adversarial networks. Communications of the ACM, 63(11), 139-
144.

Graefe, A., Haim, M., Haarmann, B., & Brosius, H. B. (2018). Readers’ perception of computer-
generated news: Credibility, expertise, and readability. Journalism, 19(5), 595-610.

Hagar, N., & Diakopoulos, N. (2019). Optimizing content with A/B headline testing: Changing
newsroom practices. Media and Communication, 7(1), 117-127.

Haim, M., & Graefe, A. (2017). Automated news: Better than expected?. Digital journalism,
5(8), 1044-1059.

Harris, J. (2017, October 9). Email content: Increase reach, impact, and subscriber satisfaction.
Content Marketing Institute. https://contentmarketinginstitute.com/articles/emails-
reach-subscriber-satisfaction/

Hassan, S., Nadzim, S. Z. A., & Shiratuddin, N. (2015). Strategic use of social media for small
business based on the AIDA model. Procedia-Social and Behavioral Sciences, 172,
262-269.

Huang, M. H., & Rust, R. T. (2021). A strategic framework for artificial intelligence in
marketing. Journal of the Academy of Marketing Science, 49, 30-50.

Ibrahim, R., Leng, N. S., Yusoff, R. C. M., Samy, G. N., Masrom, S., & Rizman, Z. I. (2017). E-
learning acceptance based on technology acceptance model (TAM). Journal of
Fundamental and Applied Sciences, 9(4S), 871-889.

Jarek, K., & Mazurek, G. (2019). Marketing and Artificial Intelligence. Central European
Business Review, 8(2).

Kansal, T., Bahuguna, S., Singh, V., & Choudhury, T. (2018, December). Customer
segmentation using K-means clustering. In 2018 international conference on
computational techniques, electronics and mechanical systems (CTEMS) (pp. 135-139).
IEEE.

Kim, J., Shin, S., Bae, K., Oh, S., Park, E., & del Pobil, A. P. (2020). Can AI be a content
generator? Effects of content generators and information delivery methods on the
psychology of content consumers. Telematics and Informatics, 55, 101452.

King, W. R., & He, J. (2006). A meta-analysis of the technology acceptance model. Information
& management, 43(6), 740-755.

Kingma, D. P., & Welling, M. (2013). Auto-encoding variational bayes. arXiv preprint
arXiv:1312.6114.

41
Köbis, N., & Mossink, L. D. (2021). Artificial intelligence versus Maya Angelou: Experimental
evidence that people cannot differentiate AI-generated from human-written poetry.
Computers in human behavior, 114, 106553.

Kohavi, R., Henne, R. M., & Sommerfield, D. (2007, August). Practical guide to controlled
experiments on the web: listen to your customers not to the hippo. In Proceedings of the
13th ACM SIGKDD international conference on Knowledge discovery and data mining
(pp. 959-967).

Koob, C. (2021). Determinants of content marketing effectiveness: Conceptual framework and

empirical findings from a managerial perspective. PloS one, 16(4), e0249457

Kotras, B. (2020). Mass personalization: Predictive marketing algorithms and the reshaping of
consumer knowledge. Big Data & Society, 7(2), 2053951720951581.

LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015). Retrieved from
https://doi.org/10.1038/nature14539

Legris, P., Ingham, J., & Collerette, P. (2003). Why do people use information technology? A
critical review of the technology acceptance model. Information & management, 40(3),
191-204.

Lewis, E. S. (1908). Financial advertising (Vol. 1628). Рипол Классик.

Liaw, S. S., & Huang, H. M. (2013). Perceived satisfaction, perceived usefulness and interactive
learning environments as predictors to self-regulation in e-learning environments.
Computers & Education, 60(1), 14-24.

Marangunić, N., & Granić, A. (2015). Technology acceptance model: a literature review from
1986 to 2013. Universal access in the information society, 14, 81-95.

Marcus, G., Davis, E., & Aaronson, S. (2022). A very preliminary analysis of dall-e 2. arXiv
preprint arXiv:2204.13807.

Midjourney. (2023). Models. Retrieved from https://docs.midjourney.com/docs/models

Mumtaz, R. (2019). Awareness and perspectives social media as new strategic marketing
approach in minor industries; notion grounded on AIDA model. Journal of Content,
Community Communication, 10.

Muyanga, C. C., & Phiri, J. (2020). Assessment of effective communication in international

schools in developing countries based on the Berlo’s SMCR model. Open Journal of
Business and Management, 9(1), 448-459.

OpenAI. (2023). GPT-4 technical report. arXiv. https://arxiv.org/pdf/2303.08774.pdf

Poradova, M. (2020). Content marketing strategy and its impact on customers under the global
market conditions. In SHS Web of Conferences (Vol. 74, p. 01027). EDP Sciences.

42
Pornsuksawat, S., & Kheokao, J. (2019). Marketing communication for LGBT in Thailand
Tourism Industry: A preliminary Findings.

Prathapan, M., Sajin Sahadevan, D., & Zakkariya, K. A. (2018). Effectiveness of digital
marketing: Tourism websites comparative analytics based on AIDA model.
International Journal of Innovative Research & Studies, 8(4), 262-273.

Pulizzi, J. (2013). Epic Content Marketing: How to Tell a Different Story, Break through the
Clutter, and Win More Customers by Marketing Less. McGraw Hill.

Radford, A., Narasimhan, K., Salimans, T., & Sutskever, I. (2018). Improving language
understanding by generative pre-training.

Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., & Sutskever, I. (2019). Language models
are unsupervised multitask learners. OpenAI blog, 1(8), 9.

Ramesh, A., Pavlov, M., Goh, G., Gray, S., Voss, C., Radford, A., ... & Sutskever, I. (2021,
July). Zero-shot text-to-image generation. In International Conference on Machine
Learning (pp. 8821-8831). PMLR.

Rancati, E., & Gordini, N. (2014). Content marketing metrics: Theoretical aspects and empirical
evidence. European Scientific Journal, 10(34).

Rauniar, R., Rawski, G., Yang, J., & Johnson, B. (2014). Technology acceptance model (TAM)
and social media usage: an empirical study on Facebook. Journal of enterprise
information management, 27(1), 6-30.

Rehman, F. U., Nawaz, T., Ilyas, M., & Hyder, S. (2014). A comparative analysis of mobile and
email marketing using AIDA model. Journal of Basic and Applied Scientific Research,
4(6), 38-49.

Rivas, P., & Zhao, L. (2023). Marketing with ChatGPT: Navigating the Ethical Terrain of GPT-
Based Chatbot Technology. AI, 4(2), 375-384.

Roetzer, P. (2017, September 30). The 5Ps of Marketing Artificial Intelligence. Marketing AI
Institute. Retrieved from https://www.marketingaiinstitute.com/blog/the-5ps-of-
marketing-artificial-intelligence

Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2022). High-resolution image
synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on
Computer Vision and Pattern Recognition (pp. 10684-10695).

Rubino, E. C., Tian, N., & Pelkki, M. H. (2022). Improving communications to increase
nonindustrial private forest landowner (NIPF) participation in forest certification
programs: a case study in Arkansas, USA. Forests, 13(1), 86.

Schweidel, D. A., Reisenbichler, M., Reutterer, T., & Zhang, K. (2023). Leveraging AI for content
generation: A customer equity perspective. In Artificial Intelligence in Marketing (pp.
125-145). Emerald Publishing Limited.

43
Segarra-Saavedra, J.; Cristòfol, F. J. and Martínez-Sala, A. M. (2019). Artificial intelligence (AI)
applied to informative documentation and journalistic sports writing. The case of
BeSoccer. Doxa Comunicación, 29, pp. 275-286.

SEMrush. (2021). State of Content Ops & Outsourcing Report 2021. Retrieved from
https://www.semrush.com/marketplace/state-of-content-ops-and-outsourcing-report/#0

Stahl, S. (2022, October 19). B2B Content Marketing Insights for 2022: More Budget, More
Work, More Empathy. Content Marketing Institute.
https://contentmarketinginstitute.com/articles/b2b-content-marketing-research-trends-
statistics

Stone, M., Aravopoulou, E., Ekinci, Y., Evans, G., Hobbs, M., Labib, A., ... & Machtynger, L.
(2020). Artificial intelligence (AI) in strategic marketing decision-making: a research
agenda. The Bottom Line, 33(2), 183-200.

Sutanto, S. M. (2016). The Use of Facebook Fanpage as Promotional Media for the Novel Book
fans by using the Model of SMCR Berlo.

Taherdoost, H., & Madanchian, M. (2023). Artificial Intelligence and Sentiment Analysis: A
Review in Competitive Research. Computers, 12(2), 37.

Tang, Y. (2021, October). A robot wrote this?: An empirical study of AI's applications in writing
practices. In Proceedings of the 39th ACM International Conference on Design of
Communication (pp. 380-381).

Taylor, T. (2023, March 21). The Pros and Cons of AI-Generated Content. HubSpot Blog.
Retrieved from https://blog.hubspot.com/marketing/ai-generated-content

Turunen, L. (2021). Improving Effectiveness and Customer Experience with Personalized Email
Marketing: Case: Raflaamo Newsletter.

Vishnoi, S. K., Bagga, T., Sharma, A., & Wani, S. N. (2018). Artificial intelligence enabled
marketing solutions: A review. Indian Journal of Economics & Business, 17(4), 167-
177.

Weber Shandwick. (2016, June). Global consumers are seven times more likely to see a positive
than negative impact of artificial intelligence. Retrieved from
https://www.webershandwick.com/news/global-consumers-are-seven-times-more-
likely-to-see-a-positive-than-negativ/

Yang, X., & Hayashi, T. (2021, September). Exploring the Effects of Internet Memes in Social
Media Marketing through A/B Testing. In 2021 IEEE 23rd Conference on Business
Informatics (CBI) (Vol. 2, pp. 97-106). IEEE.

44
APPENDICES

Appendix 1. AI-generated blog prompt for GPT4

Prompt: Write a 1000+ words blog that educates readers on the topic "Unraveling the Mysteries
of Hair Thinning: Causes, Prevention, and the Game-Changing Role of Onion Oil" and provides
practical tips. Use a conversational and engaging tone that reassures readers and draws on research
and references to support your claims. Break up the content with short paragraphs, bullet points,
and subheadings to make it easy to read and identify key takeaways. End the blog with a call to
action that inspires readers to take action and learn more about Onion Oil.

45
Appendix 2. Prompts and content results for emails

Email 1 Prompt:
So I am running this offer of Free Delivery! for customers who haven't made a purchase yet on
Bano Herbal's Store. The free delivery offer is not limited to just one product or category, but all
skin and hair care products available on Bano Herbal. Therefore, I need you to write an engaging
email that converts them. Determine the most appropriate, creative, engaging language, CTA,
Title, content, preview text etc. Make sure it is brief yet effective rather than writing too much text
and keep the body of the email brief. Also, don’t include unnecessary details as I will be adding a
direct link to website.
Prompt Result:

Source: Author’s snapshot from Bano Herbal’s Shopify

46
Appendix 2 continued

Human Email Group 1:

Source: Author’s snapshot from Bano Herbal’s Shopify

Prompt Email Group 2:

I have this deal:
Acne Free Trio, Sale price Rs.2,000.00, Regular price Rs.2,050.00, SALE, Tax included. Shipping
calculated at checkout.
One Complete Solution for Acne-Prone Skin.

For the last four years, Bano Herbal is spreading smiles in youth, suffering from ACNE. Our Acne
Cure Paste, Neem Facewash, and Neem Active Spray is one complete solution for all types of
Acne. This amazing trio has thousands of satisfied customers all over Pakistan. #0Chemical

Best for 1. Acne 2.Pimples 3. Acne Marks 4. Deep cleansing 5. Blemishes 6. Pigmentation 7.
Suntan 8. Open Pores 9. Excessive Oil

Key Ingredients:
Mint, Basil, Neem Leaves, Moringa, Vitamin E Oil, Rosemary Oil, Zinc Oxide, Rose, Turmeric
Powder, Nutmeg, Starch, Sana Maki, Fuller’s Earth, Cinnamon, Fenugreek, Colin Clay,
Paraben-Free. Chemical Free. Cruelty-Free.
47
Appendix 2 continued

For further queries:

Insta: @bano.skincare
WhatsApp: 03079403557

#herballifestyle #0chemical #pakistan

Customers also get free delivery automatically since the order price is 2000.

Following the best email suggestion and elements, i need you to write an email for customers of
my shopify store, Bano Herbal.

Results:
Subject: Exclusive Offer: Save on Our Acne Free Trio + FREE Shipping!
Preheader: Limited Time Sale - Transform Your Acne-Prone Skin with Bano Herbal's Acne Free
Trio
Dear [Customer's Name],
We're excited to announce an exclusive offer just for you! For a limited time, save on our
bestselling Acne Free Trio and get on the path to clearer, healthier skin. Plus, enjoy FREE shipping
with your purchase! Don't miss out on this amazing opportunity.
Acne Free Trio: Special Sale Price Rs.2,000.00 (Regular price Rs.2,050.00)
What's included:
Acne Cure Paste
Neem Facewash
Neem Active Spray
Why choose our Acne Free Trio? Trusted by thousands of satisfied customers across Pakistan
100% natural, chemical-free, and cruelty-free Effective against acne, pimples, blemishes,
pigmentation, and more Formulated with powerful herbal ingredients like mint, neem,
moringa, and turmeric FREE shipping on your order

Don't wait! This offer is available for a limited time only.

[Claim Your Discount & Get FREE Shipping Now!]

For any questions or assistance, feel free to reach out to us on Instagram @bano.skincare or
WhatsApp at 03079403557.
Thank you for being a valued Bano Herbal customer. We can't wait to help you achieve beautiful,
acne-free skin.
Happy shopping!

Best regards,
[Your Name] [Your Title/Position] Bano Herbal [Company Contact Information]

[Optional: Social Media Links or Newsletter Subscription Information]

Preview Text: Save BIG on Acne Free Trio + FREE Shipping! Limited time offer to achieve
clearer, healthier skin with our natural, chemical-free solution.
48
Appendix 2 continued

Human Email Group 2:

Source: Author’s snapshot from Bano Herbal’s shopify

49
Appendix 3. Prompt and content results for AI landing page and its Ad copy

Landing Page Prompt:

I want you to act as a product landing page content planner and writer, crafting well-structured,
persuasive, and engaging content that effectively showcases a product's features, benefits, and
unique selling points. Your responses should focus on addressing the needs and desires of the
target audience, encouraging them to take action, such as making a purchase. Additionally,
consider using clear and concise language, persuasive calls-to-action, and visually appealing
elements like images, videos, or graphics to create an impactful and high-converting landing page.
Your first task is to plan and write the content for a product landing page for "Onion Oil," a new
Paraben-Free, Chemical-Free, and Cruelty-Free oil for hair loss etc. Highlight the key features,
benefits, and reasons why customers should choose Onion Oil over other Hair loss products on the
market. Additionally, remember to use Pakistani names in the testimonials and provide answers to
FAQ section instead of simply providing questions. Also, don’t provide 30 days guarantee or
discount as it is not for a promotion.

Ad copy prompt:

Considering the landing page content that you provided above, what should be the content for
Facebook ads and what kind of image or video should be used along with call to action.
You can also provide me details for the suitable target audience and regions in Pakistan where this
ad should be running for better results.

Ad copy result:
Response:
[Headline] Unlock the Secret to Healthy, Radiant Hair with Onion Oil!

[Body] Are you struggling with hair loss, breakage, and thinning? It's time to explore the power
of our all-natural Onion & Coconut Oil blend. Transform your hair and regain your confidence!
Visit our website to learn more about this incredible solution.

[Call-to-Action] Learn More

Image/Video Suggestions:
An engaging image or video showcasing the benefits of using Onion Oil, such as promoting hair
growth, reducing breakage, and restoring shine.
A captivating image of the Onion Oil bottle, highlighting its natural ingredients (onions and
coconut oil).
A short video featuring expert advice or testimonials from satisfied customers, sharing their
experiences with Onion Oil.
Remember to optimize the ad copy, images, and targeting to encourage more clicks and visits to
the website. The "Learn More" CTA is an excellent choice for driving traffic to your site, allowing
potential customers to explore Onion Oil's benefits and features in greater detail before making a
purchase decision.

50
Appendix 4. Questionnaire for Blog Posts

Thank you for participating in this survey. You will be asked to read a blog post and provide your
feedback on its quality and effectiveness. Here is a link to the blog post (inserted link for the
version of blog post that was being served to them).
Your responses will help us better understand the perception and evaluation of content quality and
credibility. Your participation is voluntary, and your answers will be kept confidential.

Please read the blog post carefully and then answer the questions that follow.

Demographics (optional)

Please specify your age.

(OPEN TEXT)

What is your gender?

a) Man
b) woman
c) Non-binary
d) Prefer not to say

Effectiveness

Please rate the following statements regarding the effectiveness of the blog post on a scale of 1 to
5 (1 being strongly disagree, 5 being strongly agree).
The blog post …
1. effectively conveyed its intended message.
2. provided valuable and informative content.
3. engaged and captured my attention.
4. was persuasive in promoting the solution to the problem.

Content Quality

Please rate the following statements regarding the quality of the blog post on a scale of 1 to 5 (1
being strongly disagree, 5 being strongly agree).

The blog post…

1. was well-written and free of errors.
2. maintained a consistent tone and style.
3. had a clear structure and flow.
4. was easy to read and understand.

Credibility

Please rate the following statements regarding the credibility of the blog post on a scale of 1 to 5
(1 being strongly disagree, 5 being strongly agree).

51
Appendix 4 continued

The blog post…

1. provided reliable information.
2. was written in a professional and credible manner.
3. demonstrated expertise on the subject matter.
4. built trust with the reader.

Section 4: Perception of the Author

The text you read was randomly selected from the pool of articles written by both human
professionals and AI. Assess the following statement on a scale of 1 to 5, with 1 being "Strongly
Disagree" and 5 being "Strongly Agree”.

The text was written by a human.

Closing:

Thank you for participating in this survey. Your feedback is greatly appreciated and will help us
better understand the subject matter. If you have any additional comments or suggestions, please
feel free to share them below or by sending an email at: [email protected].

52
Appendix 5. Questionnaire for professional writers

Thank you for participating in this survey!

The purpose of the survey is to gather the opinions and experiences of professional writers
regarding the use of AI-powered writing tools in the industry. The survey aims to understand the
perception and attitudes of writers towards AI-generated content.

Your responses will be anonymous and will be used for research purposes only.

Demographic Questions:

1. Age: Please specify your age (eg., 25)?

(Open Text)

2. What is your gender?

a) Man
b) Woman
c) Non-binary
d) Prefer not to say

3. Country: Please specify your country of residence (eg. Estonia)

(Open Text)
4. What is your highest level of education?

a) High school diploma or equivalent

b) Associate degree
c) Bachelor's degree
d) Master's degree
e) Doctoral degree

Professional experience:

5. How many years of experience do you have in professional writing?

(Open Text)

6. Which industry do you primarily write for? (Select all that apply)

a) Technology
b) Healthcare
c) Finance
d) Retail
e) Education
f) Travel & Hospitality
g) Entertainment
h) Other (please specify)

Experience with AI Generated Content:

53
Appendix 5 continued

7. Which AI content generation tools have you used? (Select all that apply)
a) OpenAI's Chat GPT
b) AI Writer
c) Copy.ai
d) Jarvis (Conversion.ai)
e) Other (please specify)

8. How frequently do you use AI-powered writing tools?

a) Daily
b) Weekly
c) Monthly
d) Rarely
e) Never used AI-powered writing tools

Perception of AI Generated Content:

Effectiveness:
Please rate the following statements regarding the effectiveness of AI-generated content on a scale
of 1 to 5 (1 being strongly disagree, 5 being strongly agree).

AI-generated content…
1. is effective in conveying the intended message.
2. is informative and provides value to the reader.
3. can significantly save time in the content creation process.
4. can improve personalization and targeting in content marketing.

Credibility:
Please rate the following statements regarding the credibility of AI-generated content on a scale
of 1 to 5 (1 being strongly disagree, 5 being strongly agree).

AI-generated content
1. is well-researched and provides reliable information.
2. is written in a professional and credible manner.
3. is generally perceived as credible by readers.
4. is effective in building trust with the target audience.

Content Quality:
Please rate the following statements regarding the quality of AI-generated content on a scale of
1 to 5 (1 being strongly disagree, 5 being strongly agree).

AI-generated content…

1. is of high quality and meets professional standards.

2. can match the quality of human-written content.
3. maintains consistent tone and style.
4. requires minimal editing and proofreading.

54
Appendix 5 continued

Limitations:

Please rate the following statements regarding the limitations of AI-generated content on a scale
of 1 to 5 (1 being strongly disagree, 5 being strongly agree).

AI-generated content…

1. has a higher risk of misleading or incorrect information.

2. has limitations in creativity and originality.
3. may struggle to convey complex or nuanced topics.
4. poses ethical concerns in content marketing.

55
Appendix 6. Descriptive Statistics for writer’s questionnaire

Source: SPSS descriptive statistics results based on author’s calculations

56
Appendix 7. Correlation analysis for writer’s questionnaire

Source: Author’s analysis

57
Appendix 8. Kruskal-Wallis H for writer’s questionnaire 1

Source: SPSS results for Kruskal-Wallis H test with gender as grouping variable

58
Appendix 9. Kruskal-Wallis H for writer’s questionnaire 2

SPSS results for Kruskal-Wallis H test with gender as grouping variable

59
Appendix 10. Non-exclusive license

A non-exclusive license for the reproduction and publication of a graduation thesis1

I Sohaib Arshad

1. Grant Tallinn University of Technology a free license (non-exclusive license) for my thesis
Performance of AI-generated content in content marketing, supervised by Tarmo Koppel,

1.1 to be reproduced for the purposes of preservation and electronic publication of the
graduation thesis, incl. to be entered in the digital collection of the library of Tallinn University of
Technology until the expiry of the term of copyright;

1.2 to be published via the web of Tallinn University of Technology, incl. to be entered in the
digital collection of the library of Tallinn University of Technology until the expiry of the term of
copyright.

2. I am aware that the author also retains the rights specified in clause 1 of the non-exclusive
license.

3. I confirm that granting the non-exclusive license does not infringe other persons' intellectual
property rights, the rights arising from the Personal Data Protection Act or rights arising from other
legislation.

09.05.2023

1
The non-exclusive licence is not valid during the validity of access restriction indicated in the student's application
for restriction on access to the graduation thesis that has been signed by the school's dean, except in case of the
university's right to reproduce the thesis for preservation purposes only. If a graduation thesis is based on the joint
creative activity of two or more persons and the co-author(s) has/have not granted, by the set deadline, the student
defending his/her graduation thesis consent to reproduce and publish the graduation thesis in compliance with clauses
1.1 and 1.2 of the non-exclusive licence, the non-exclusive license shall not be valid for the period

60