When AI Is the Story: LLM-Based Evidence on AI Narratives, Disclosure Credibility, and IPO Performance

Artificial intelligence (AI) has become an increasingly important aspect of the U.S. IPO market. However, it is noteworthy that investors may encounter difficulties in distinguishing issuers with concrete AI capabilities from those who primarily utilize AI as a strategic narrative. The present study investigates the impact of AI-related disclosures, both in terms of content and credibility, on IPO pricing and subsequent performance. The sample comprises U.S. IPOs completed between 2020 and 2025 and identified through Renaissance Capital’s IPO Pro database. Registration statements and final prospectuses are collected from the SEC's EDGAR archive. The approach involves the application of a large language model to analyze AI-related statements in the prospectuses, and the classification of each statement according to its intensity, specificity, verifiability, and business relevance. These dimensions are then aggregated into an AI Narrative Credibility Score. Subsequently, cross-sectional models of first-day underpricing, six- and twelve-month returns, and post-IPO volatility are estimated. The specifications control for offer size, firm age, profitability, industry, market conditions, underwriter reputation, and listing-year fixed effects. Preliminary evidence suggests a link between more intensive AI narratives and stronger first-day returns. However, this association is weaker when disclosures are broad, generic, or weakly linked to identifiable products, business models, or operational capabilities. It has been demonstrated that firms which make more specific and verifiable disclosures regarding artificial intelligence demonstrate more stable aftermarket performance. Conversely, issuers which are narrative-heavy exhibit greater volatility and weaker medium-term returns. The findings indicate that investors may initially reward AI-related positioning but subsequently differentiate substantive technological disclosure from symbolic claims. The study contributes to research on AI in finance by applying LLM-based text analysis to IPO prospectuses. It also offers governance implications for issuers, underwriters, and regulators, emphasizing the importance of transparent, specific, and verifiable AI communication in capital-market transactions.