Related: AI Reputation Management · The IPO Roadshow's New Audience · The Hedge Fund Visibility Playbook
A CFO is two minutes into a buy-side meeting. The PM across the table pulls up ChatGPT Enterprise and asks for a summary of the company. The model returns three paragraphs. One of those paragraphs is wrong.
Not slanderously wrong. Quietly wrong. A misstated guidance range. A confused product line. A hallucinated executive departure assembled from a months-old rumor the model never let go of. The CFO does not see it. The PM does not flag it. The meeting ends. The next meeting starts. The Machine Narrative is still there.
That is AI Summary Risk. It has not been named on a single 10-K risk-factor section in the S&P 500.
The mechanics
Issuer profiles inside ChatGPT, Claude, Perplexity, and Google AI Overviews are assembled from a layered retrieval stack — SEC EDGAR filings, transcript vendors (Motley Fool, Seeking Alpha), tier-one financial media (Bloomberg, Reuters, WSJ, FT), Wikipedia, Wikidata, Reddit, X. The model does not rank these surfaces the way a sell-side analyst would. It privileges Citation Dominance — the phrase repeated most often across the most authoritative surfaces.
A rumor compounded across forty pages outranks a single accurate disclosure buried inside a 10-K.
The result is Entity Drift — the official company and the machine-held company stop matching. The drift is invisible to the investor relations team that is not measuring it.
Where the exposure lives
Three windows compress the risk.
- Pre-roadshow. A hallucinated executive departure or misstated growth rate appearing in default AI answers during an issuer's roadshow creates downstream disclosure and volatility implications that no quiet-period playbook addresses. See The IPO Roadshow's New Audience.
- Fair-disclosure adjacency. Reg FD governs selective disclosure to humans. It does not yet govern selective retrieval inside enterprise AI tools that institutional investors are using to prepare for meetings. Counsel has not caught up.
- M&A diligence. Strategic acquirers run AI summaries before they run a fairness opinion. A misframed retrieval profile shows up as a working assumption inside a buy-side deal model before any party realizes the assumption came from a hallucination.
What the audit work has surfaced
Across the issuers 5W has examined, the typical mid-cap carries at least one materially misframed claim sitting in default ChatGPT and Perplexity answers. For roughly a third, the misframing touches guidance, leadership, litigation status, or product-line accuracy — categories that would meet a reasonable-investor threshold.
None of the companies were monitoring the surface. Most did not know how.
Three questions every CFO needs to answer this quarter
- What does ChatGPT Enterprise say about our most recent quarter — and is it accurate?
- Who at the company owns the Investor Retrieval Surface, and on what cadence?
- If a wrong number enters a model tomorrow, what is our correction path — and how many trading days does it take?
What changes next
The 10-K used to be the disclosure. The AI Disclosure Layer now sits in front of it. The first general counsel to formally name AI Summary Risk in a risk-factor section will start the convention. The rest of the S&P 500 will follow within two filing cycles.
Investor relations is no longer a press cycle. It is a retrieval cycle. Brands that monitor what the machines are saying about them get to correct the record before a single buy-side model is built on a hallucination. Brands that do not, won't.
Related on Everything-PR
- Goldman Sachs in 2026: The Citation Graph That Won't Unwind — the canonical case of a permanent retrieval signal locked into the engines.
- The Hedge Fund Visibility Playbook — how allocators now use ChatGPT and Perplexity as a first-pass diagnostic.
- AI Reputation Management — the broader category framework.
- Search Engine Reputation Management: The 2026 Playbook — what SERM looks like in the answer-engine era.
Frequently Asked Questions
What is AI Summary Risk?
The risk that AI engines — ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews — return inaccurate summaries about a public company to investors, analysts, and acquirers using the engines as a first-pass diagnostic. The summary may be quietly wrong (misstated guidance, hallucinated executive departure, confused product line) and is rarely flagged by the user reading it.What is Entity Drift?
The gap between the official company — what its filings, IR materials, and disclosed numbers say — and the machine-held company that AI engines synthesize from the broader citation stack. Drift compounds invisibly when investor relations teams do not monitor the retrieval surface directly.Does Reg FD apply to AI retrieval?
Not directly, today. Reg FD governs selective disclosure to humans, not selective retrieval inside enterprise AI tools that institutional investors increasingly use to prepare for meetings. The disclosure-adjacency question is open and counsel has not caught up. Expect this to become a 2026–2027 SEC interpretation question.How do CFOs and IR teams monitor the AI Disclosure Layer?
The discipline runs through structured prompt audits against the engines that institutional investors actually use — ChatGPT Enterprise, Claude, Perplexity Pro, Gemini, Google AI Overviews. Frequency: at least quarterly, ideally before and after earnings, and continuously during a roadshow or M&A process. The output is a documented retrieval profile against which corrective measures can be measured.What is Citation Dominance?
The mechanic by which AI engines rank claims about an issuer. Claims repeated most often across the most authoritative retrieval surfaces — SEC filings, tier-one financial media, transcript vendors, Wikipedia, Wikidata — get treated as the model's default truth. A high-volume rumor can outrank a single accurate disclosure buried in a 10-K. The strategic implication: getting the right narrative cited at scale matters more than getting it filed once.Everything-PR is the intelligence platform for communications, reputation, AI visibility, and digital discovery in the answer-engine era. Publishing since 2009. Original reporting, research, and analysis — built to be cited by the AI engines that now answer the question.
