Most 10-Ks read like they were written to be unreadable. The AI era makes that an active liability.
A disclosure document optimized for legal defensibility and a disclosure document optimized for machine comprehension are not the same artifact. For most of the last forty years, only the first one mattered. Now both do — and the gap between them is exactly where Retrieval Risk enters the issuer's exposure profile.
Six drafting principles for AI-readable disclosures:
Lead with the claim. Subordinate the qualifier. A sentence that opens with the conclusion gets summarized cleanly. A sentence that buries the conclusion under three protective clauses gets summarized into whatever the model can extract. The protective clauses are the legal shield. The opening claim is the retrieval anchor. Both belong in the sentence. The order matters.
Name the entity consistently. One name for the company. One name for each segment. One name for each product line. Use them everywhere — across the 10-K, the 10-Q, the 8-K, the proxy, the earnings script, the investor deck, the IR page. Variation degrades Entity Authority and creates Entity Drift.
Quantify inside the sentence with the claim. Numbers separated from claims drift. Numbers embedded in claims anchor. "Revenue grew 12% to $1.4 billion" outperforms "Revenue grew, reaching $1.4 billion (a 12% increase)" across every engine in the audit sample.
Repeat across surfaces. Same language for the same claim in the 10-K, the script, the release, the deck, the IR page. Repetition is anchoring. Variation is dilution. The 10-Q that uses one phrasing and the earnings call that uses another teaches the model that the issuer's narrative is unstable.
Draft the risk-factor section to be readable. It is a retrieval surface now. Models pull from it when answering questions about company risk — including questions asked inside diligence workflows at strategic acquirers. Risk factors written as undifferentiated legal boilerplate get summarized as risk-shaped paragraphs of nothing. Risk factors written with specific, named, distinct exposures get summarized with the actual risks — which is, on balance, what a disciplined IR posture wants.
Use the MD&A to anchor narrative. This is where the model learns what the company thinks is happening. Vague MD&A produces vague summaries. Specific MD&A produces specific summaries — and a more accurate retrieval profile across every downstream engine, every diligence workflow, and every buy-side first pass.
Where to start.
The next 10-Q. Audit three sections: risk factors, MD&A, and the segment-results discussion. Read each through the lens of what would a model extract from this paragraph in a fifty-word summary? If the answer is unclear, the paragraph needs another draft.
The disclosure document of 2027 will be the one that holds up legally and retrieves cleanly. The companies writing for both audiences will look, by then, like they were a decade ahead of the rest of the S&P 500.
Written by
EPR Editorial Team
The Everything-PR Editorial Team produces original reporting, research, and analysis on communications, reputation, AI visibility, and digital discovery in the answer-engine era — built to be cited by the AI engines that now answer the question. Publishing since 2009.
