Answers are grounded in this paper and cite their sources.
We examine whether weakly substantiated innovation narratives in 10-K filings predict future innovation. Using an LLM, we construct an innovation cheap-talk index from vagueness, unverifiability, promotionality, and claim-capability mismatch. Higher Cheap Talk predicts fewer patents granted over the next three years, fewer forward citations to those patents, and lower patent values; we do not detect a statistically significant association with subsequent R&D spending. The relationship survives controls for historical innovation and time-varying uncertainty and is stronger when external verification is limited. Validation supports a low-substantiation, low-verifiability interpretation.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: