Generative AI and ChatGPT in Financial Markets and Corporate Policy: A

Comprehensive Review

William Mbanyele

Center for Economic Research, Shandong University, China

[email protected]

Abstract: This article conducts a comprehensive review of the rapidly increasing research on the

implications of generative AI, particularly Large Language Models (LLMs) such as ChatGPT, for

financial markets. We first review studies that utilize ChatGPT to interpret sentiment within

financial reports, conference calls, and analyst reports, assessing how these insights can influence

evaluations of corporate performance. Additionally, our review extends to exploring the broader

impacts of generative AI on corporate policies, shedding light on the transformative potential and

challenges posed by these technologies. Through this synthesis, we aim to provide a holistic

overview of the current research in finance, highlighting significant findings, identifying existing

knowledge gaps, and proposing future directions for scholarly investigation. This review is

intended as a valuable resource for academics, industry practitioners, and policymakers, offering

critical insights into the integration of generative AI technologies in financial analysis and

corporate strategy formulation. It underscores the rapid evolution of this interdisciplinary field and

advocates for continued exploration to understand the potential of generative AI in shaping the

future of financial markets and corporate strategies.

Key Words: Generative AI, Large Language Models (LLMs), ChatGPT, Financial markets,

Corporate performance, Sentiment analysis, Natural language processing (NLP), Artificial

intelligence (AI) in finance

Electronic copy available at: https://ssrn.com/abstract=4745990

1. Introduction

The comprehensive review provides an insightful examination of the burgeoning field of

generative artificial intelligence, with a particular focus on Large Language Models (LLMs) like

ChatGPT, and their implications for financial markets and corporate strategies. This review

synthesizes current research exploring the utilization of ChatGPT for interpreting sentiments

within financial reports, earnings calls, and analyst reports. It assesses how insights derived from

these analyses can influence evaluations of corporate performance and potentially alter corporate

policies.

At the heart of the review is the exploration of generative AI's transformative potential and

the challenges it introduces. The review serves as a holistic overview of the present research

landscape, identifying significant findings and gaps in knowledge, while suggesting future

research directions. It is positioned as a valuable resource for academics, industry practitioners,

and policymakers, offering critical insights into the integration and impact of generative AI

technologies in the domain of financial analysis and corporate strategy formulation.

The review emphasizes the need for future research to dig deeper into the ethical, privacy, and

regulatory challenges posed by AI in finance, explore AI's influence on employment within the

financial sector, particularly regarding skill development and labor substitution, and assess the

long-term economic effects of integrating generative AI into financial markets and corporate

strategies. Through this, the review outlines a future research agenda aimed at fostering a nuanced

understanding of generative AI's role in the financial and corporate world.

Electronic copy available at: https://ssrn.com/abstract=4745990

2. Use of LLM in finance research

2.1 Generative AI as a tool

Jha, Qian, Weber, & Yang (2023) examine the application of ChatGPT in forecasting corporate

investment decisions based on earnings call transcripts. They introduce a novel firm-level

ChatGPT investment score, which assesses managers' expectations on capital expenditure changes,

validated against CFO survey responses. This score not only predicts future capital expenditures

effectively but also forecasts investments in intangibles and R&D, showing ChatGPT's utility in

analyzing corporate policies beyond traditional models. The study also explores the impact of high

investment scores on future stock returns, revealing significant negative returns for firms with high

investment scores, which suggests that the market underestimates the information derived from

earnings calls processed by ChatGPT.

Li, Tu, and Zhou (2023) investigate the accuracy of ChatGPT, specifically the GPT-4

model, in making future earnings predictions alongside justifications compared to consensus

forecasts by human analysts. The findings reveal that ChatGPT's predictions generally exhibit

larger errors and tend to be more optimistic than those of human analysts. However, the accuracy

of ChatGPT's forecasts improves in environments with better information quality and when

earnings press releases contain more substantive content. This suggests that while generative AI

like ChatGPT shows potential in aiding financial market participants with information processing

and content creation, there is a significant risk associated with over-reliance on these models in

the absence of high-quality information. The research underlines the dual promise and peril of

employing generative AI in financial forecasting and analysis, emphasizing the technology's

Electronic copy available at: https://ssrn.com/abstract=4745990

current limitations and the need for careful application in contexts where accurate and reliable

information is crucial

Lopez-Lira and Tang (2023) explore the capacity of ChatGPT to forecast stock price

movements by analyzing the sentiment of financial news headlines. They discovered that

ChatGPT's sentiment scores have a statistically significant predictive power on daily stock market

returns, underscoring the potential utility of ChatGPT as a valuable tool for sentiment analysis in

financial markets. This research contributes to the understanding of how generative AI can be

leveraged for predictive analytics in financial markets.

Li, Feng, Yang and Huang (2024) investigates the capacity of ChatGPT, a large language

model, to act as a financial advisor in forecasting listed firm performance, particularly focusing on

constituent stocks of the China Securities Index 300. Comparing ChatGPT's forecasts for key

financial performance metrics with those of human analysts and actual realized values, our

research reveals ChatGPT's ability to rectify optimistic biases often observed in human forecasts.

By showcasing ChatGPT's potential as a financial advisor and its role in mitigating human biases

in financial decision-making, this study contributes to the existing literature in exploring the

transformative impact of AI in finance.

Kim, Muhn, and Nikolaev (2023) explore the potential of generative AI tools like ChatGPT

in summarizing complex corporate disclosures and their impact on investor information

processing. The authors investigate whether summaries generated by ChatGPT can effectively

distill relevant information from corporate disclosures, focusing on Management Discussion and

Analysis (MD&A) sections and earnings conference calls as primary data sources. Their findings

suggest that ChatGPT-generated summaries are significantly shorter than original disclosures, yet

they amplify the information content, making the sentiment of summaries more pronounced. This

Electronic copy available at: https://ssrn.com/abstract=4745990

indicates that summaries could potentially be more effective at explaining stock market reactions

to disclosed information. Furthermore, the study introduces a measure of information "bloat" to

quantify the degree of redundant textual information in corporate disclosures, finding that bloated

disclosures are associated with negative capital market consequences, such as lower price

efficiency and higher information asymmetry. The study contributes to the literature on the

economic usefulness of generative AI in processing financial information, providing insights into

how these technologies might help investors navigate the vast amounts of textual data in corporate

disclosures more efficiently.

Kim et al., (2023) explore the application of generative AI, specifically large language

models like GPT-3.5, in analyzing and extracting insights on corporate risks from earnings call

transcripts. The study innovatively applies GPT-based techniques to develop firm-level risk

exposure measures to political, climate, and AI-related risks. By comparing these AI-generated

measures against traditional bigram-based and dictionary-based approaches, the researchers

demonstrate the superior capacity of generative AI in capturing nuanced and predictive insights

on risk exposures. They highlight the significant variance in risk exposure attributable to firm-

specific factors, underscoring the complexity and individuality of corporate risk profiles.

Additionally, the paper examines the economic implications of these risk measures, linking them

to stock price volatility and corporate decision-making processes related to investments, lobbying

activities, and innovation efforts. The research not only validates the predictive power and

economic relevance of GPT-derived risk assessments but also illuminates the evolving landscape

of corporate risks, particularly the rising significance of AI-related risks in recent years.

2.2 Corporate policies

Electronic copy available at: https://ssrn.com/abstract=4745990

Jia, Li, Ma, and Xu (2023) present an insightful examination of how companies have reacted to

the advent of generative artificial intelligence (GAI), specifically after the release of ChatGPT in

November 2022. The research examines GAI within corporate conference calls, highlighting an

increased focus on GAI topics across various industries, particularly among firms with higher

innovation intensity, cybersecurity concerns, product differentiation, labor exposure to AI, and

customer operations.

The authors argue that managerial discussions suggest a belief in GAI's beneficial impact

on firms known for their innovation and those facing cybersecurity threats, whereas companies

with significant product differentiation view GAI as more detrimental. Mixed perspectives were

noted on GAI’s effects on labor and customer operations, with a general tilt towards a positive

outlook. The study also distinguishes between discussions containing actionable initiatives versus

general commentary, finding a greater inclination towards the former in firms engaged in

innovation, facing cybersecurity threats, and with extensive customer operations.

This research contributes to understanding the corporate perception and response to GAI,

indicating a broad spectrum of implications for innovation, cybersecurity, product strategy, labor,

and customer engagement. It underscores the transformative potential of GAI across different

business dimensions, while also acknowledging the diverse reactions and strategies adopted by

firms in anticipation of or in response to these technological advancements.

Bai, Boyson, Cao, Liu, and Wan (2023) study the informational disparity between

corporate executives' responses during earnings calls and responses generated by Large Language

Models (LLMs) like ChatGPT, Google Bard, and an open-source LLM. The authors introduce a

novel metric, Human-AI Differences (HAID), to quantify the unique information content provided

by executives that surpasses the knowledge base of these advanced AI models. They demonstrate

Electronic copy available at: https://ssrn.com/abstract=4745990

that a higher HAID value is strongly predictive of stock liquidity, abnormal returns, and analyst

forecast accuracy following earnings calls, emphasizing the crucial role of human insights over

AI-generated content in providing new, market-moving information. This study underscores the

potential of using LLMs as tools for investors to sift through layers of disclosed information to

discover hidden insights, highlighting the significant economic implications of generative AI in

the financial domain

2.3Employment

Brynjolfsson, Li, & Raymond (2023). explore the effects of a new generative AI-based

conversational assistant on job performance, using data from 5,179 customer support agents.

Results show that the introduction of the AI tool leads to a 14% increase in productivity,

particularly benefiting novice and low-skilled workers with a 34% improvement, while

experienced workers show minimal changes. The AI model helps spread best practices, aids in

skill development for newcomers, enhances customer sentiment, boosts employee retention, and

fosters potential learning among workers. Overall, the findings indicate that access to generative

AI has the potential to significantly enhance productivity across different skill levels.

Dogukan, Naumovska, and Aggarwal (2023) examine the impact of artificial intelligence

(AI) on labor substitution, particularly in the context of language translation services and

knowledge work. Using the introduction of Google's neural network-based translation (GNNT)

and the emergence of ChatGPT as case studies, the authors analyze the effects of these

technologies on human translators' job volume and the broader implications for skilled knowledge

workers.

The study provides empirical evidence that the adoption of GNNT led to a reduction in the

demand for human translation services in an online labor market, particularly for tasks with a

Electronic copy available at: https://ssrn.com/abstract=4745990

higher analytical component compared to those requiring cultural and emotional insights. This

distinction highlights a heterogeneous substitution effect, where AI's capability to replace human

tasks varies based on the nature of the task. Furthermore, the paper explores the implications of

ChatGPT in question and answer forums, suggesting that AI technologies, while capable of

performing analytical tasks, may still struggle with tasks requiring deep cultural understanding or

emotional intelligence.

By documenting the substitution effect of AI on specific human tasks, the research

contributes to the ongoing discourse on AI's role in the economy, its potential to disrupt traditional

employment structures, and the importance of adapting human skills to complement emerging

technologies. The findings underscore the nuanced nature of AI's impact on the workforce,

emphasizing that while some tasks may be automated, others will continue to require the

irreplaceable insights and abilities of human workers

Brynjolfsson, Li, and Raymond (2023) examine the impact of introducing a generative AI-

based conversational assistant on the productivity of customer support agents. Analyzing data from

5,179 agents, the study finds that access to the AI tool led to a 14% increase in the number of

issues resolved per hour, with the most significant benefits observed among novice and lower-

skilled workers. This suggests the AI tool may disseminate tacit knowledge from more experienced

or skilled workers to newer ones, helping them advance more rapidly on the learning curve.

Moreover, the AI assistance was found to improve customer sentiment, reduce managerial

intervention requests, and enhance employee retention. The paper contributes to understanding the

economic implications of generative AI in the workplace, highlighting its potential to redistribute

skills and affect labor dynamics

Electronic copy available at: https://ssrn.com/abstract=4745990

 Eisfeldt et al., (2023) examine the impact of Generative AI on the valuation of firms,

focusing on the labor market effects. They introduce a new metric for measuring a firm's workforce

exposure to Generative AI and presents findings that firms with high Generative AI exposure saw

significant market value increases shortly after the release of ChatGPT. This effect is attributed

mainly to the labor substitution channel, where Generative AI replaces or augments human labor,

particularly in tasks that are core to specific occupations. The study also explores the broader

adoption of Generative AI across industries, highlighting its role as a general-purpose technology

that enhances productivity and operational efficiency.

2.4 Stock performance

Liu, Zhuang, Liu and Han (2024) study the impact of ChatGPT's introduction on Chinese listed

companies' cumulative abnormal return (CAR). The findings indicate that events surrounding

ChatGPT releases, such as GPT-3.5 and GPT-4, significantly benefit companies focused on

ChatGPT. Initially, these firms experience undervaluation in the stock market, attributed to

information asymmetry and competitive pressures from international markets. However, increased

media coverage over time enhances social attention and recognition of ChatGPT's value,

amplifying its impact.

Bertomeu, Liu, Lin, and Ni (2023) investigate the capital market effects of Italy's prohibition

of ChatGPT due to privacy law violations on March 31, 2023. This unique situation provided a

natural experiment to observe generative AI's influence on the economy. The study found that

Italian companies highly engaged with generative AI technology faced a notable

underperformance—approximately 9%—relative to less exposed firms during the ban period. The

impact was more pronounced on smaller and newly established companies, suggesting a tie to the

concept of creative destruction. Additionally, the ban's enforcement led to a deterioration in the

Electronic copy available at: https://ssrn.com/abstract=4745990

information environment, evidenced by a decrease in forecasts from analysts based in Italy

compared to their foreign counterparts and an increase in bid-ask spreads, particularly for

companies with limited institutional investors or analyst coverage. These observations underscore

generative AI's dual role in enhancing firm productivity and information processing capabilities.

The research offers significant insights into generative AI's economic value and its broad

implications for market dynamic

3. Future research agenda

This review serves as a foundation for identifying gaps in the current understanding and suggests

several avenues for future exploration:

3.1 AI's Role in Corporate Strategy and Policy: Future studies could focus on how generative

AI influences corporate strategic decisions, including investments in innovation, R&D, and human

capital using long term data. Exploring the balance between AI-driven efficiency gains and the

potential for creative destruction would provide valuable insights into corporate policy

formulation.

3.2 Regulatory and Ethical Implications: There's a need for research on the regulatory

challenges and ethical considerations arising from AI adoption in financial markets. This includes

issues of privacy, data security, and the potential for AI to exacerbate market inequalities.

3.3 GAI's Application in Financial Analysis: While AI has shown promise in analyzing financial

documents and market sentiment, further research could explore methods to improve the accuracy

and reliability of AI-generated analyses, particularly in complex, uncertain market environments.

3.4 GAI's Impact on Employment and Skill Development: Researchers have started examining

the nuanced effects of AI on labor markets, including the potential for AI to substitute or augment

Electronic copy available at: https://ssrn.com/abstract=4745990

human labor in various financial and corporate roles. Future research could focus on how GAI

reshapes the demand for specific skills and the implications for workforce development.

3.5 Long-term Economic Effects of GAI: Longitudinal studies could investigate the broader

economic consequences of integrating generative AI into financial markets and corporate

strategies, including effects on productivity, competitiveness, and innovation.

By addressing these areas, future research can build on the foundational insights provided by this

review, exploring the complex interplay between generative AI technologies, financial market

dynamics, and corporate policy decisions.

4. Conclusion

This review synthesizes early research to showcase how ChatGPT can distill sentiment from

financial documents, aid in interpreting corporate communications, and potentially reshape

corporate strategy and policy. This review serves as a call to action for researchers, highlighting

the importance of continued exploration into the multidimensional effects of generative AI on the

financial industry and corporate practices. By addressing the outlined research gaps, scholars can

contribute to a more nuanced comprehension of AI's potential to revolutionize financial markets

and corporate governance, ensuring that the benefits of these technologies are maximized while

mitigating associated risks and challenges.

Electronic copy available at: https://ssrn.com/abstract=4745990

References

Eisfeldt, A. L., Schubert, G., & Zhang, M. B. (2023). Generative AI and firm values (No. w31222).

National Bureau of Economic Research.

Bai, J.J., Boyson, N.M., Cao, Y., Liu, M., Wan, C. (2023). Executives vs. chatbots: Unmasking

insights through human-AI differences in earnings conference Q&A. Working paper

Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at Work. NBER Working Paper

No. 31161. National Bureau of Economic Research.

Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work (No. w31161). National

Bureau of Economic Research.

Jia, N., Li, N., Ma, G., & Xu, D. (2023). Corporate Responses to Generative AI: Early Evidence

from Conference Calls, Working paper

Kim, A. G., Muhn, M., & Nikolaev, V. V. (2023). Bloated Disclosures: Can ChatGPT Help

Investors Process Information? Retrieved from SSRN: https://ssrn.com/abstract=4425527.

Kim, A. G., Muhn, M., & Nikolaev, V. V. (2023). From Transcripts to Insights: Uncovering

Corporate Risks Using Generative AI. arXiv preprint arXiv:2310.17721.

Li, E. X., Tu, Z., & Zhou, D. (2023). The Promise and Peril of Generative AI: Evidence from

ChatGPT as Sell-Side Analysts. Available at SSRN 4480947

Lopez-Lira, A., & Tang, Y. (2023). Can ChatGPT Forecast Stock Price Movements? Return

Predictability and Large Language Models. arXiv. Retrieved from

https://arxiv.org/abs/2304.07619

Electronic copy available at: https://ssrn.com/abstract=4745990

Wu, Qinqin and Zhuang, Qinqin and liu, yitong and Han, Longyan, Technology Shock of Chatgpt,

Social Attention and Firm Value: Evidence from China. Available at

SSRN: https://ssrn.com/abstract=4739412 or http://dx.doi.org/10.2139/ssrn.4739412

Xiaoyang Li, Haoming Feng, Hailong Yang & Jiyuan Huang (2024) Can ChatGPT reduce human

financial analysts’ optimistic biases?, Economic and Political Studies, 12:1, 20-33, DOI:

Electronic copy available at: https://ssrn.com/abstract=4745990