Paper: arXiv 2512.19705

Authors: Jian Xue, Qian Zhang, Wu Zhu

Abstract

We study how generative artificial intelligence (AI) transforms the work of financial analysts. Using the 2023 launch of FactSet’s AI platform as a natural experiment, we find that adoption produces markedly richer and more comprehensive reports – featuring 40% more distinct information sources, 34% broader topical coverage, and 25% greater use of advanced analytical methods – while also improving timeliness. However, forecast errors rise by 59% as AI-assisted reports convey a more balanced mix of positive and negative information that is harder to synthesize, particularly for analysts facing heavier cognitive demands. Placebo tests using other data vendors confirm that these effects are unique to FactSet’s AI integration. Overall, our findings reveal both the productivity gains and cognitive limits of generative AI in financial information production.

Complexity vs Empirical Score

  • Math Complexity: 2.0/10
  • Empirical Rigor: 6.0/10 (capped from 9.5: no out-of-sample test; see penalty flags)
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper employs sophisticated econometric methods (DiD, PSM, entropy balancing) on a large proprietary dataset of analyst reports, but uses minimal complex mathematics or formulas. The analysis is highly data-driven and backtest-ready, focusing on empirical causal inference.

Research Flowchart

  flowchart TD
  A["Research Goal:<br/>How does Generative AI<br/>transform financial analysts' work?"] --> B["Natural Experiment:<br/>2023 FactSet AI Platform<br/>Launch"]
  B --> C["Data/Inputs:<br/>Analyst Reports<br/>Forecast Accuracy"]
  C --> D["Computational Process:<br/>Difference-in-Differences Analysis<br/>Placebo Tests (Other Vendors)"]
  D --> E{"Key Findings/Outcomes"}
  E --> F["Productivity Gains:<br/>+40% Info Sources<br/>+34% Topical Coverage"]
  E --> G["Cognitive Limits:<br/>+59% Forecast Errors<br/>Harder Synthesis"]

Cited References (125)

  1. The impact of generative AI on information processing: Evidence from the ban of ChatGPT in Italy (2025) — link
  2. AI as ``Co-founder’’: GenAI for Entrepreneurship (2025) — link
  3. Textual factors: A scalable, interpretable, and data-driven approach to analyzing unstructured information (2025)
  4. Generative AI use by capital market information intermediaries: Evidence from seeking alpha (2026)
  5. Does generative AI facilitate investor trading? Evidence from ChatGPT outages (2025)
  6. How stock market participants use generative artificial intelligence: Evidence from user-platform interaction data (2025)
  7. Generative AI and investor processing of financial information (2026)
  8. AI democratization and trading inequality (2026)
and 117 more
  1. Word on the street: Field evidence of artificial intelligence use in sell-side equity research (2025)
  2. Generative AI in financial reporting (2026)
  3. Interactivity and illusions of ability: How using generative AI affects investor judgments (2026)
  4. Empirical asset pricing via machine learning (2020)
  5. How generative AI could disrupt creative work (2023)
  6. Artificial intelligence, education, and entrepreneurship (2024)
  7. Divergent LLM Adoption and Heterogeneous Convergence Paths in Research Writing (2024)
  8. GPT-4o system card (2024) — link
  9. Generative AI at work (2025)
  10. AI , skill, and productivity: The case of taxi drivers (2025)
  11. The rapid adoption of generative AI (2024)
  12. The unequal adoption of ChatGPT exacerbates existing inequalities among workers (2025)
  13. GPTs are GPTs : Labor market impact potential of LLMs (2024)
  14. AI and the Extended Workday: Productivity, Contracting Efficiency, and Distribution of Rents (2025)
  15. Generative AI and firm values (2023)
  16. Generative AI as Seniority-Biased Technological Change: Evidence from US Resume and Job Posting Data (2025)
  17. Large language models, small labor market effects (2025)
  18. The labor market effects of generative AI : A difference-in-differences analysis of AI exposure (2025)
  19. Artificial intelligence adoption and workplace training (2025)
  20. How Retrainable Are AI-Exposed Workers? (2025)
  21. Evidence of a social evaluation penalty for using AI (2025)
  22. Aggregate productivity growth: Lessons from microeconomic evidence (2001)
  23. Creative accounting or creative destruction? Firm-level productivity growth in Chinese manufacturing (2012)
  24. A Model of Growth Through Creative Destruction (1992)
  25. How Adaptable Are American Workers to AI-Induced Job Displacement? (2025)
  26. Productivity performance, distance to frontier and AI innovation: Firm-level evidence from Europe (2024)
  27. Free range startups? Market scope, academic founders, and the role of general knowledge in AI (2025)
  28. Data-driven investors (2025)
  29. Local Heterogeneity in Artificial Intelligence Jobs over Time and Space (2025)
  30. Generative artificial intelligence in entrepreneurship education enhances entrepreneurial intention through self-efficacy and university support (2025)
  31. AI Regulation and Entrepreneurship (2024)
  32. AI and Operational Losses: Evidence from US Bank Holding Companies (2024)
  33. Artificial Intelligence and Mergers and Acquisitions (2025)
  34. Artificial Intelligence in Recruitment: Examine the Effects on Hiring and Firm Outcomes (2025)
  35. Artificial intelligence and high-skilled work: Evidence from analysts (2020)
  36. Identity disclosure and anthropomorphism in voice chatbot design: A field experiment (2024)
  37. Artificial intelligence, lean startup method, and product innovations (2025)
  38. Man versus machine: Complex estimates and auditor reliance on artificial intelligence (2022)
  39. Can artificial intelligence reduce the effect of independence conflicts on audit firm liability? (2024)
  40. Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses (2021)
  41. Home-tutoring services assisted with technology: Investigating the role of artificial intelligence using a randomized field experiment (2022)
  42. Does CSR reduce idiosyncratic risk? Roles of operational efficiency and AI innovation (2021)
  43. Is artificial intelligence improving the audit process? (2022)
  44. Does automation improve financial reporting? Evidence from internal controls (2025)
  45. Predicting Material Misstatements Using Machine Learning (2025)
  46. The Janus face of artificial intelligence feedback: Deployment versus disclosure effects on employee performance (2021)
  47. Super mario meets AI : Experimental effects of automation and skills on team performance and coordination (2025)
  48. Capitalism, socialism and democracy (1943)
  49. A model of growth through creative destruction (1990)
  50. Selection and the Evolution of Industry (1982)
  51. Entry, exit, and firm dynamics in long run equilibrium (1992)
  52. Generative AI and Finance (2024)
  53. Innovation strategy after IPO : How AI analytics spurs innovation after IPO (2025)
  54. Organizational Technology Ladders: Remote Work and Generative AI Adoption (2025)
  55. The role of entrepreneurship in US job creation and economic dynamism (2014)
  56. Who creates jobs? Small versus large versus young (2013)
  57. Digital economics (2019)
  58. UMAP : Uniform manifold approximation and projection for dimension reduction (2018) — link
  59. HDBSCAN : Hierarchical density based clustering. (2017)
  60. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation (2024) — link
  61. Shifting work patterns with generative ai (2025)
  62. Generative AI in knowledge work: Design implications for data navigation and decision-making (2025)
  63. ChatGPT and DeepSeek: Can they predict the stock market and macroeconomy? (2025) — link
  64. Generative AI and Asset Management (2026) — link
  65. Analyzing the analysts: Career concerns and biased earnings forecasts (2003)
  66. Divergent llm adoption and heterogeneous convergence paths in research writing (2025) — link
  67. Discussion of ``Analysts’ industry expertise’’’ (2012)
  68. Gpt-4 technical report (2023) — link
  69. Seeing like an ai: How llms apply (and misapply) wikipedia neutrality norms (2024) — link
  70. Arbiters of Ambivalence: Challenges of Using LLMs in No-Consensus Tasks (2025) — link
  71. Machine learning methods that economists should know about (2019)
  72. Recursive partitioning for heterogeneous causal effects (2016)
  73. From man vs. machine to man+ machine: The art and AI of stock analyses (2024)
  74. Human versus machine: A comparison of robo-analyst and traditional research analyst investment recommendations (2022)
  75. Automated earnings forecasts: Beat analysts or combine and conquer? (2018)
  76. Man versus machine learning: The term structure of earnings expectations and conditional biases (2023)
  77. Generative AI in Action: LLM Applications for Financial Analysis (2025)
  78. The promise and peril of generative AI: Evidence from GPT-4 as sell-side analysts — → 2412.01069
  79. The Memorization Problem: Can We Trust LLMs’ Economic Forecasts? — → 2504.14765
  80. Analyst report readability (2015)
  81. Information content of equity analyst reports (2005)
  82. Analyst information discovery and interpretation roles: A topic modeling approach (2018)
  83. Evidence on the information content of text in analyst reports (2014)
  84. Reading between the lines: An empirical examination of qualitative attributes of financial analysts’ reports (2012)
  85. Following the leader:: a study of individual analysts’ earnings forecasts (2001)
  86. Financial analyst characteristics and herding behavior in forecasting (2005)
  87. Team earnings forecasting (2009)
  88. Capital market effects of media synthesis and dissemination: Evidence from robo-journalism (2018)
  89. The tool for the automatic analysis of text cohesion (TAACO): Automatic assessment of local, global, and text cohesion (2016)
  90. Entropy balancing for causal effects: A multivariate reweighting method to produce balanced samples in observational studies (2012)
  91. Analyst forecast accuracy: Do ability, resources, and portfolio complexity matter? (1999)
  92. Analysts’ motives for rounding EPS forecasts (2012)
  93. Expertise in forecasting performance of security analysts (1999)
  94. Analyst forecasting ability and the stock price reaction to forecast revisions (2000)
  95. Decision fatigue and heuristic analyst forecasts (2019)
  96. Face value: Trait impressions, performance characteristics, and market outcomes for financial analysts (2022)
  97. Concurrent earnings announcements and analysts’ information production (2020)
  98. Analyst career concerns, effort allocation, and firms’ information environment (2019)
  99. Using and interpreting fixed effects models (2024)
  100. Analyst responsiveness and the post-earnings-announcement drift (2008)
  101. Locked-in at home: Female analysts’ attention at work during the COVID-19 pandemic (2020)
  102. When security analysts talk, who listens? (2007)
  103. Information interpretation or information discovery: Which role of analysts do investors value more? (2012)
  104. An analyst by any other surname: Surname favorability and market reaction to analyst forecasts (2019)
  105. Investor reaction to celebrity analysts: The case of earnings forecast revisions (2007)
  106. Self-selection and the forecasting abilities of female equity analysts (2010)
  107. Political contributions and analyst behavior (2016)
  108. The effect of social skills on analyst performance (2023)
  109. Disclosure processing costs, investors’ information choice, and equity market outcomes: A review (2020)
  110. Limited attention, information disclosure, and financial reporting (2003) — link
  111. Implications of rational inattention (2003)
  112. When attention is away, analysts misplay: Distraction and analyst forecast performance (2024) — link
  113. Air pollution, affect, and forecasting bias: Evidence from Chinese financial analysts (2021)
  114. When do associate analysts matter? (2022) — link
  115. Managing decision fatigue: Evidence from analysts’ earnings forecasts (2024) — link
  116. Analyst teams (2021)
  117. A neuro-metabolic account of why daylong cognitive work alters the control of economic decisions (2022) — link

2 of 125 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.