Original Study · 18 Aug 2026

1 in 9 Businesses Recommended by OpenAI Had an Incorrect Claim

We checked 1,257 testable claims across 182 recommended businesses. Twenty businesses had at least one incorrect claim, and another 479 claims could not be verified from the available evidence.

Study sample

1,257 claims across 182 businesses

Measured outcome

Factual claim accuracy

AI platform

OpenAI

What this means for marketers

Monitor the exact facts answer engines repeat about the brand. Fix inconsistent identity, service, credential, and location information across owned and authoritative third party sources.

• Audit the facts attached to every monitored recommendation, not only whether the brand appeared.
• Correct business names, addresses, phone numbers, opening hours, service areas, and branch details first.
• Align services, credentials, awards, and commercial claims across the website and authoritative third party profiles.
• Escalate high-severity errors and repeat the same prompts after source corrections to see whether the answer changes.

What we tested

How often do OpenAI local business recommendations contain incorrect or unverifiable factual claims? The study covered 1,257 claims across 182 businesses.

What the result does not prove

Unsupported means the bounded evidence did not verify a claim; it does not mean the claim was false. The study covers recommended businesses, not every business in the 3,000-row panel.

• Unsupported claims were not proven false; the bounded evidence set simply did not confirm them.
• The audit covered 182 identity-validated businesses that OpenAI recommended, not all 3,000 sampled businesses.
• Citation presence was measured at the recommendation level and did not prove that every sentence was supported by a cited page.
• The uncited group contained only 12 businesses, so its accuracy comparison is descriptive rather than decisive.