Disclosure written for machine readers
Information transparency and corporate disclosure are rising in the accounting journals, and language models now read filings for investors. Yet few disclosure papers mention language models or XBRL. Is disclosure written for machines changing what transparency means?
What prompted this
On the accounting dashboard, "information transparency" (5 papers in July–September 2026 against about 1 expected) and "corporate disclosure" (7 against 3) are among the fastest-growing keywords. Since October 2025, 1,136 papers carry a disclosure or transparency keyword; only 43 of them also carry a language-model, generative-AI or XBRL keyword, and "XBRL" itself appears on 2.
Questions
- Are filings that machines parse easily (structured tags, clean tables) priced faster after release?
- Earlier text-analysis work found that firms avoid the words that machine readers count as negative. Do language-model readers see through that?
- Does machine readability help small investors, or mostly the fastest traders?
Data that could help
- SEC EDGAR filings with acceptance timestamps, inline XBRL data and intraday prices.
- The SEC's phase-in of inline XBRL (2019–2021) as a staggered change in machine readability.
Next step
Measure how fast prices react to 10-K releases before and after each group of filers moved to inline XBRL.