10 red flags of AI-washing in a startup pitch

Pitch decks are written to impress, and in 2026 nothing impresses a room like “AI”. That’s exactly why the word is so often doing marketing work rather than describing a product. This is a field guide to the ten most common AI-washing red flags — and, for each, the single question that cuts through it.

None of these is proof on its own (a weak slide is not a weak company). Several together, with no good answers, are a reason to slow down. For the full framework behind this, see the pillar on how to verify whether a company really does AI.

The ten flags

1. “AI” with no nouns. The deck says “AI-powered” but never names a model, technique or dataset. Ask: “Which specific models or methods, and are they yours or someone else’s API?”

2. Accuracy without an error rate. “99% accurate” with no test set, baseline or methodology. Ask: “Accurate on what data, measured how, versus what baseline?”

3. The always-recorded demo. Every demonstration is a video, or quietly “human-assisted for now”. Ask: “Can we run it live, on our example, right now?”

4. The brand-new AI title. A “Chief AI Officer” appeared the same quarter as the AI messaging. Ask: “What has the AI team actually shipped, and where can I see it?”

5. Marketing dwarfs R&D. The company spends far more on sales and brand than on engineering. Ask: “What’s your R&D as a share of spend, and what did it build?”

6. Wrapper dressed as IP. “Our proprietary model” turns out to be a thin layer over a third-party API. Ask: “What happens to the product if that provider changes price or access?”

7. Blended AI revenue. Revenue is reported as vague “AI-driven” totals, never broken out. Ask: “How much revenue is directly attributable to the AI, specifically?”

8. No data moat. The data is public or licensed and replicable by any competitor. Ask: “What about your data is proprietary and hard to copy?”

9. The rebrand re-rate. The valuation jumped after an “AI” rebrand, with no change in the business. Ask: “What changed in the product or revenue to justify the new valuation?”

10. Risk-factors silence. For public companies, the 10-K barely mentions AI in risk factors while press releases can’t stop. Ask: “If AI is core, why doesn’t the company treat it as a real dependency where the law requires candour?”

Turning flags into a decision

The flags are a map of where to dig, not a scorecard. A genuinely good company can trip two or three and still be worth backing — what matters is that you asked, got answers, and priced the remaining risk deliberately. The failure mode is letting the word “AI” switch your scrutiny off.

For public companies you can check several of these against primary filings in minutes — our free AI company verifier links straight to the SEC documents. For private startups, where there’s no 10-K to keep anyone honest, work through the full due-diligence checklist.

---

Educational content, not investment advice. We describe general warning signs and avoid factual claims of wrongdoing about specific named companies.