Can an AI search engine cite my page for something my page does not say?

How often do AI citations point to a page that does not back the claim?

Often in older answer engines, much less often in Google’s current AI Overviews.

Narayanan Venkit and colleagues (opens in a new tab) ran 303 real questions through You.com, Perplexity and BingChat in 2024, scoring all 909 answers. For each cited statement, an AI judge checked whether the cited page actually supported it.

The authors note the problem went beyond statements with no support at all. Even when one of the listed sources did support a statement, engines often cited a different source that did not. That is the case most relevant to you: a real claim, attributed to the wrong page.

Two cautions apply. These were 2024 engines, and ChatGPT and Google’s AI Overviews were not tested. The AI judge also agreed with human checkers only moderately, so the figures are estimates.

Is it better in Google’s AI Overviews?

Much better, though not zero.

Xu, Iqbal and Montgomery (opens in a new tab) captured Google’s AI Overviews (the AI summary at the top of Google’s results) for 55,393 trending searches over 40 days in spring 2026. They split the Overviews into 98,020 single claims and checked each one against the pages it cited.

• 2.66% of claims were contradicted by a page the Overview cited.
• 6.98% were not mentioned by any cited page at all.
• 41.9% of Overviews had every claim backed by their sources.
• 2.74% of Overviews had fewer than half their claims backed.

Some unbacked claims may come from social and video pages the researchers could not read. Even assuming all of those were fine, the unbacked rate would fall only from 11.0% to roughly 5.3%. For a business, that means a small but real share of claims arrive with a link that does not say what the summary says.

What does misattribution look like in practice?

A correct-sounding claim, a real link, and a page that says something different or nothing on the point.

In the 2024 user study, all 21 expert participants spotted misattribution at some point. One said a statement “doesn’t seem to be in the source,” even though the statement itself was true. Others described engines picking only one side of a source that discussed both.

Our own reputation study found a commercial version of this. We asked four AI engines whether 79 brands were legitimate and checked 240 cited claims against the pages they cited:

• Where a cited page could be read, it supported the claim fully or in part 72.4% of the time.
• Of the checkable claims saying a problem was common, 71.4% cited evidence showing only individual reports, or nothing about frequency.
• 119 of the 240 claims could not be checked at all, because review sites block automated reading.

That second point is the subtle risk. Your page, or a review page about you, can be cited for a broad claim such as “customers often complain” when it only shows a few cases.

Do readers notice when a citation is wrong?

Mostly not, and a citation can make a wrong answer more convincing.

Li and Aral (opens in a new tab) ran a randomized experiment with 4,927 adults, chosen to represent the US population. Adding reference links to AI search answers raised trust. The increase was the same whether the links were valid or invented and broken.

People also rarely click through. A Pew Research study, reported by Huang and colleagues (opens in a new tab), found users clicked sources cited inside Google’s AI summary in only 1% of visits. So a misattributed claim with your name on the link will usually be read, not checked.

Why would an AI attach your URL to a claim you never made?

Because an answer blends several pages, and engines often link a sentence to the wrong one.

An earlier study cited by Huang and colleagues found only 51.5% of generated statements were fully supported by their citations. The pattern recurs across engines and years.

What should you do about it?

Monitor what AI answers attribute to you, make your own claims unmistakable, and keep evidence of what your pages said.

1. Search AI engines for your brand and key claims. Click the citations to your pages and check that each linked page says what the answer says.
2. State key facts once, plainly and specifically. Clear, quotable sentences give an engine less room to blend your page with someone else’s.
3. Keep dated copies of important pages. If an answer attributes a statement to you, you can show what the page said at the time.
4. Watch review sites, not just your own pages. Broad claims about your reputation are often tied to review pages showing individual cases.
5. Report serious misattributions to the platform. Where a wrong claim carries legal or safety risk, raise it through the engine’s feedback channel and get legal advice.

If you want help tracking how AI engines describe and cite you, see our generative engine optimization service.

What does the research not tell us yet?

The research measures how often citations miss, but not how often that harms the business being cited.

• The engine-by-engine citation accuracy figures come from 2024 engines. Comparable current figures for ChatGPT and Perplexity are scarce in the papers we reviewed.
• Most audits use AI judges to decide whether a page supports a claim, which adds error.
• The AI Overview audit used trending searches, not the commercial searches buyers make about vendors.
• Our reputation study could not check half its sampled claims because review sites block automated reading.
• No study measures the reputational or legal consequences for a company wrongly cited.

Frequently asked questions

Can Google’s AI Overviews misquote my website?

Occasionally. In a 2026 audit of 98,020 claims, 2.66% were contradicted by a page the Overview cited, and 6.98% were not mentioned by any cited page.

Which AI search engine cites sources most accurately?

No current head-to-head answer exists. In a 2024 audit, Perplexity’s citations supported their statements 49.0% of the time, against 68.3% for You.com.

Do people check the sources AI answers cite?

Rarely. Pew found users clicked sources cited in Google’s AI summary in only 1% of visits, and links raised trust even when invalid.

What should I do if an AI answer attributes a false claim to my company?

Record the answer and your page as it stood, then report it to the platform. For serious cases, get legal advice.

Sources

• Narayanan Venkit, Laban, Zhou, Mao and Wu (2024), Search Engines in an AI Era: The False Promise of Factual and Verifiable Source-Cited Responses (opens in a new tab), arXiv:2410.22349.
• Xu, Iqbal and Montgomery (2026), Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact (opens in a new tab), arXiv:2605.14021.
• Li and Aral (2025), Human Trust in AI Search: A Large-Scale Experiment (opens in a new tab), arXiv:2504.06435.
• Huang, Goyal, Saha and Chandrasekharan (2026), Answer Bubbles: Information Exposure in AI-Mediated Search (opens in a new tab), arXiv:2603.16138.
• Underneath (2026), “Is this brand legit?” How AI assistants build a reputation