The First Annual Reports of the LLM Era

This study examines purposively selected narrative passages from 150 SEC Form 10-K filings for a fixed, sector-stratified sample of 50 companies across FY2019, FY2022 and FY2024. A single author scored six directional dimensions of language change on an ordinal scale from zero to two. Five dimensions cover changes in claims, hedging, framing, named specificity and sentence expansion. The sixth concerns newly introduced or reorganised sections and is reported separately.

Across the five editorial dimensions, the score allocation gives annual aggregate rates of 35.3 points during 2019–2022 and 44.0 points during 2022–2024, an increase of 24.5 per cent. The median company difference is zero, a company-resampling bootstrap interval extends from approximately −0.10 to 0.45, and no timing convention gives either interval a majority of companies. The result depends upon purposive passage selection, interpretive scoring and interval allocation. It describes the scored sample and rubric. It does not establish a population trend, identify generative AI use or support causal attribution.