What is a brand hallucination?
A brand hallucination is a false claim about a specific business that an AI tool states as fact. Ask an AI assistant about a software company and it may quote a price the company never charged, describe a feature that does not exist, name the wrong founder, or say the business has closed when it has not. Each of those is a brand hallucination.
The term borrows from "AI hallucination," the general name for fluent AI output that is not true. Brand hallucination narrows it to the cases a business cares about: wrong facts about its own name, products, people and reputation, shown to potential customers who may never check.
Google is plain about the underlying risk. Its help page for AI Overviews says the technology "may provide inaccurate or offensive information" and that "AI Overviews can and will make mistakes" [1].
What are examples of brand hallucinations?
Brand hallucinations fall into five recurring types. The documented cases below involve Google's AI Overviews, because those are the ones widely reported, but the same types appear in other AI assistants.
| Type | What it looks like |
|---|---|
| Invented facts | A price, feature, policy, location or opening hours that are not real |
| Entity confusion | Two businesses or people with similar names merged into one |
| False relationships | A claim that a company owns, partners with or is related to another |
| Outdated facts presented as current | A discontinued product or former executive described as current |
| False accusations | A claim that a business runs scams or a person committed offenses |
Three documented cases show how serious this can get, all summarized in Wikipedia's AI Overviews article [2]:
- False accusations against businesses. Two Munich-based publishers sued Google after AI Overviews said they were "performing scams, subscription traps, and other immoral business practices." A Munich regional court ruled that AI Overviews generate "independent, new, and substantive statements" and that Google could be held liable for them [2].
- False claims about a public figure. An AI Overview wrongly stated that Canadian musician Ashley MacIsaac had criminal convictions. The false claim led to the cancellation of a December 2025 performance, and he took legal action against Google [2].
- Entity confusion. An AI Overview stated that footballers Joaquín Correa and Ángel Correa are brothers, although they are unrelated [2].
The first two are reputational. The third is the mechanism behind many smaller brand errors: two names that look alike treated as the same entity.
Why do AI tools hallucinate about brands?
AI tools hallucinate about brands because they produce the most plausible-sounding answer from what they have, and for most businesses what they have is thin, inconsistent or out of date. Four causes account for most cases:
- Little training data. A large brand appears in thousands of documents. A small business may appear in a handful. With few examples, a model fills gaps with patterns from similar businesses.
- Ambiguous names. Businesses that share a name with another company, a place or a person are easily merged, as the Correa case shows [2].
- Bad retrieved sources. AI search features ground answers in web pages they retrieve. If the best-ranking page about a brand is wrong, a joke, or a competitor's comparison, the answer can repeat it. Google's AI Overviews were briefly restricted in May 2024 after suggesting people eat rocks and put glue on pizza [2].
- Content the crawler cannot see. Most major AI crawlers do not run JavaScript [7]. A pricing page or product list that only appears after JavaScript runs is invisible to them, so the model has no correct source to use.
What is the difference between a brand hallucination and negative sentiment?
A brand hallucination is factually false, while negative sentiment is an unfavorable but possibly accurate opinion. "Their support is slow, according to many reviews" is negative sentiment, and may be fair. "They charge a $500 cancellation fee" when no such fee exists is a brand hallucination.
The difference decides the response. Negative sentiment is fixed by improving the product and earning better coverage. A brand hallucination is fixed by getting correct facts in front of the systems that got it wrong, and by reporting the error. Treating a hallucination as a sentiment problem wastes effort on reviews that were never the cause.
How do you check what AI says about your brand?
To check what AI says about your brand, ask each major AI tool a fixed set of factual questions about your business and compare every answer with the truth. A practical audit takes an afternoon:
- Write 15 to 25 questions customers actually ask: "How much does [brand] cost?", "Who owns [brand]?", "Is [brand] legit?", "Does [brand] offer refunds?", "[brand] vs [competitor]", "Where is [brand] based?"
- Ask them in each tool your customers use, such as Google AI Mode, ChatGPT, Gemini, Perplexity, Claude and Copilot. Use a logged-out or fresh session.
- Record every factual claim in a sheet, with the answer's cited sources.
- Mark each claim correct, outdated or false.
- Follow each false claim to its source. Where an answer cites a page, open it. Many brand hallucinations trace to one wrong third-party page.
- Repeat monthly. AI answers change, and a fix in one tool does not carry to the others.
Commercial AI visibility tools can automate this. Profound, for example, offers a check that surfaces "inaccurate claims and their sources" [5]. Search Insider's view is that the manual audit should come first, because it shows which questions matter before paying to track them.
How do you fix a brand hallucination?
You fix a brand hallucination by correcting the sources the AI relies on and reporting the error to the platform. No AI company offers a way to edit an answer directly. What works:
- Publish the facts on your own site, in plain HTML. One page that states pricing, ownership, locations, policies and product names in plain sentences gives retrieval systems a correct source. This is the role of an entity home.
- Correct the third-party pages being cited. If an answer cites an outdated directory listing or review, update or dispute that listing. The AI will keep citing it until it changes.
- Remove ambiguity. Use your full business name consistently and link your official profiles together with structured data, so systems can tell you apart from similarly named entities.
- Allow AI search crawlers. A site that blocks crawlers such as OAI-SearchBot cannot be used as a source in that product's answers [6], which leaves only third-party pages to describe you.
- Report the error. In Google's AI Overviews, use the thumbs down icon and "Report a problem" [1]. Other AI tools have similar feedback buttons.
- Get legal advice for defamatory claims. The Munich ruling suggests the platform, not only the source, can be held responsible for false statements [2].
Opting out of Google's AI features with the Search generative AI control removes a site's own pages from AI Overviews and AI Mode [8]. It does not stop Google's AI from describing the brand using other sites, so it does not fix a hallucination.
Can you prevent brand hallucinations?
You cannot fully prevent brand hallucinations, because no AI company guarantees accuracy and Google says its AI answers will make mistakes [1]. You can reduce the risk by making correct information easy to find, consistent everywhere, and readable without JavaScript.
Google's guidance for appearing in its AI features is standard SEO, with "no additional requirements" [4]. The same applies here: the pages that rank for questions about your brand are the pages AI search features are likely to draw on.
How common are brand hallucinations?
No published study measures how often AI tools get facts wrong about businesses. General hallucination benchmarks test models on standard tasks, not on facts about specific companies, so their rates do not transfer to your brand. Any statistic claiming that a set percentage of AI answers about brands is wrong should be treated as unverified unless it comes with its method and sample.
What is documented is that the problem is real enough for Google to warn users to "always check important info in more than one place" [1], and for Google to note at AI Mode's launch that answers might "unintentionally appear to take on a persona or reflect a particular opinion" [3].
When should you worry about brand hallucinations?
Worry about brand hallucinations when an AI tool is wrong about something a customer uses to decide: price, availability, refund policy, safety, legitimacy or ownership. Those errors cost sales directly. Errors about minor history matter less. Run the audit before a launch, a price change, a rebrand or a merger, because those are the moments when the facts change and AI answers lag behind.