Using a large language model (LLM), we construct a narrative sentiment index to investigate the predictability of risk premia in the Chinese Treasury bond market. Both in-sample and out-of-sample analyses demonstrate that narrative sentiment significantly and negatively forecasts future bond excess returns, yielding significant economic value for mean-variance investors. By incorporating this measure into an affine term structure model (ATSM), we identify a dual-channel effect: narrative sentiment relates positively to expected future short rates but negatively to excess returns, with the impact intensifying at longer maturities. Risk premia estimated via this sentiment-augmented ATSM exhibit more pronounced countercyclical dynamics. Finally, we establish that narrative sentiment carries forward-looking signals regarding both subjective expectations and future fundamentals, providing a coherent economic foundation for its predictive power in the bond market.

Against a backdrop of economic transformation and industrial upgrading, financing constraints have become a significant barrier to the growth of microeconomic entities. Investor sentiment, a recognized source of irrationality in capital markets, is understood to influence the financing environment through various channels. Concurrently, advances in generative artificial intelligence (AI) are generating novel methods for the precise measurement of this sentiment. Leveraging data from Chinese A-share listed corporations from 2011 to 2023, this study constructs a double machine learning framework to examine the impact of investor sentiment on financing constraints and its underlying transmission mechanisms. Our findings indicate that elevated investor sentiment significantly reduces these constraints. This effect operates primarily through an enhanced capacity for debt financing, a reduction in market scrutiny, and a stimulation of consumer demand that boosts corporate revenue. Conversely, periods of subdued sentiment tend to trigger risk-averse behavior, intensify liquidity pressures, and further restrict access to capital. A heterogeneity analysis reveals that the effects of sentiment vary across industries, ownership structures, and regional levels of information infrastructure. The impact is particularly pronounced within the real estate sector and in more digitally advanced regions. Interestingly, the effect does not fundamentally differ across ownership types, suggesting a degree of market neutrality. In sum, this study offers fresh empirical evidence on the mechanisms through which sentiment drives financing behavior. It also expands the application of generative AI, demonstrating its value as a tool for financial research.

This paper examines how potential disruptions from generative artificial intelligence (AI) to firms' core businesses affect their cross-industry acquisition decisions. Using data on Chinese A-share listed firms, we employ a large language model to assess the potential positive and negative effects of generative AI on core businesses and construct a measure of net AI risk exposure. We measure cross-industry M&A using the semantic distance between acquirers' and targets' business descriptions. A difference-in-differences analysis exploiting the commercial diffusion of generative AI shows that more exposed firms acquire more distant targets, with an effect equivalent to 8.2% of the sample mean and robust to alternative specifications. The effect is stronger among firms with narrower innovation boundaries and more concentrated business portfolios, consistent with innovation capability reconfiguration and strategic risk diversification. It is also stronger among firms with smaller market shares, in more competitive industries, and in more marketized regions. Further analyses show that high-exposure firms acquire targets with higher AI patent shares and receive more favorable announcement-day market reactions to distant acquisitions. Our findings suggest that potential AI-related disruptions shape acquisition target selection and firms' business boundaries.

This paper examines how major artificial intelligence (AI) breakthroughs affect the market pricing of firms’ pre-existing AI capabilities. Using Chinese listed firms’ 2023–2024 AI patents classified with a large language model, we estimate dynamic stock return and volatility responses to the DeepSeek-R1 release. Stronger pre-event AI capabilities are associated with higher post-release returns and greater short-term volatility, with volatility effects persisting longer. Positive return responses concentrate in algorithms and models, not computing infrastructure, while industry applications show later volatility increases without comparable return gains. These findings highlight technology-specific differences between return and volatility responses to AI breakthroughs.

Can fully agentic artificial intelligence predict equity returns in real time? We build a novel dataset of daily AI-generated attractiveness scores spanning the entire Russell 1000 universe, continuously updated since April 2025. Each trading day, the model freely searches the live internet and produces quantitative scores for every constituent stock before the next market open, eliminating both look-ahead bias and attribution bias. A daily value-weighted portfolio of the 20 highest-ranked stocks earns a Fama-French six-factor alpha of 16.5 basis points per day (41.5% per annum), with an average bid-ask spread of 1.92 basis points. Predictability is sharply asymmetric: the model reliably identifies future outperformers but not future underperformers. The evidence points to stock picking rather than market timing: performance is strongest on down-market days, and the alpha does not come from betting on aggregate market direction. The AI’s reasoning traces help explain this predictability. Alpha is concentrated in traces with richer live-web evidence and stronger analytical structure. In the theme attribution, most active alpha reflects firm selection under an investment lens, not broad theme choice.

We investigate the capabilities of Large Language Models (LLMs) in generating novel alpha factors for option returns. Utilizing a structured prompt-engineering approach, LLMs like GPT-5 can directly create factors for two distinct options markets: the mature U.S. market and the emerging Chinese market. Empirical analysis further reveals that the LLM-generated factors exhibit remarkable and robust performance, delivering statistically signifcant returns in both all-sample and extensive out-of-sample tests. Beyond their statistical signifcance, such factors are economically meaningful. They display low self-correlation, indicating genuine innovation, and are grounded in sound economic rationale derived from market microstructure and behavioral fnance principles, showcasing a key advantage over traditional machine learning models.

Technology firms are often financed on narratives about products, contracts, customers, and technological progress well before these developments appear in accounting statements. We ask when such narratives become economically informative. Our central idea is that narratives should matter more once they can be linked to later verifiable outcomes rather than treated as stand-alone text.Using listed Chinese technology firms, we develop a validated corporate narrative framework for bank-affiliated investment, a setting in which investors must screen with soft information ex ante and then monitor hard realization and downside risk ex post. We use GPT-5.1 to extract business claims from management discussion, investor-relations records, exchange Q&A, and earnings-roadshow materials, and to label later claim–evidence pairs as support, partial support, conflict, duplicate, or irrelevant. We then connect these labels to official announcements, procurement awards, permits, project updates, and negative-event disclosures to construct a validated firm-month signal. The broad merged panel contains 592 firms and 30,169 firm-month observations; the main return tests use 576 firms and 18,230 firm-month observations over 2022–2024. A simple production rule that combines a low-narrative-premium component with hard-narrative and hard-event anchors, together with a separate downside-risk gate, delivers an implementable annualized long-short return of 8.93% in bank-invested firms after trading costs. The signal is much weaker in non-bank firms, predicts future gross-margin improvement more strongly than future ROE, and improves downside screening.

We investigate whether overnight news sentiment adds predictive value for overnight returns. We focus on the CSI300 Index, whose ETFs are widely held by Chinese retail investors. Sentiment indi-cators are constructed from minute-level overnight news using a fine-tuned RoBERTa model. These indicators are combined with market-based variables to predict overnight returns via regression and machine learning. Results show that while the sentiment alone has predictive value, its incremental contribution disappears once the A50 overnight return is included.

We document “foreign bias” in AI financial predictions, reversing the classic home bias. U.S.-based ChatGPT is systematically more optimistic than China-based DeepSeek about Chinese firms—in price predictions and directional forecasts—yet significantly less accurate. Evidence supports an information-availability mechanism: bias is strongest when U.S. media coverage of Chinese firms is limited and attenuates for cross-listed firms. Crucially, injecting Chinese news eliminates the prediction gap. Both models produce similar forecasts for U.S. firms, consistent with broader worldwide coverage. LLMs trained in different information environments can create divergent signals, with implications for investors and policymakers as AI increasingly intermediates global markets.

We propose the AI Narrative Gap as a novel firm characteristic—the systematic divergence between a firm’s AI strategic narrative intensity and its subsequent AI capital expenditure commitment—and document its capital market consequences. Using Chinese A-share listed firms from 2015 to 2022, we show that firms with a wider AI Narrative Gap attract significantly more optimistic and less accurate analyst earnings forecasts. These distorted expectations, in turn, predict lower subsequent stock returns, lower industry-adjusted abnormal returns, and weaker future accounting performance. A double-sort portfolio placing firms simultaneously in the highest tercile of the AI Narrative Gap and highest tercile of analyst optimism earns a mean return 22.8 percentage points below that of the lowest tercile on both dimensions (t = −5.10). The return reduction in the AI Narrative Gap coefficient is attenuated but not eliminated after controlling for optimism, consistent with a partial expectation-distortion channel. Collectively, these results establish the AI Narrative Gap as a cross-sectionally informative firm characteristic that captures the credibility of a firm’s AI strategic identity, with systematic implications for analyst expectations and asset prices.