Over 1,700 brands reveal how often AI chatbots produce false facts about them
In one of the largest brand-led surveys of AI misrepresentation for businesses, we asked companies to grade the facts that AI engines produce about them.
- Of more than 32,000 facts that brands reviewed from AI engines, they flagged 9 in every 100 as false.
- AI engines are most likely to get information wrong about a company's pricing, funding & financials, and the software and tools they integrate with.
- 3 in 4 brands that partook in grading their facts found at least one fact to be false.
- Companies in the AI industry saw the most false facts about them, at a rate of 18.5%.
- Companies based in New Zealand, Germany and Canada saw the highest rate of false facts about them.
- Small businesses (≤50 employees) suffer false facts from AI chatbots at a higher rate (9.4%) than both medium (7.4%) and large firms (6.3%).
Tabla de contenidos
How often do brands find incorrect AI-generated facts about them?
When brands reviewed the facts AI engines were stating about them, they flagged 9.2% as incorrect. This was equivalent to 3,005 false facts out of 32,556 reviewed.
of the facts brands checked were flagged incorrect.
That figure may sound small, but AI answers name only a handful of brands and a small number of facts about each. Unlike a Google results page with ten blue links, an AI answer is a limited shelf of options. When a potential customer is misled by wrong or outdated information about a brand, that risks losing a customer's trust entirely.
What type of fact does AI get wrong most often about brands?
The claims AI gets wrong most often are the ones that tend to change more often, such as prices (13.5%) or funding and financial details (10.9%). Even the more stable details, such as a company's services or its history and founding, still see error rates of more than 1 in 20.
When a fact changes often, AI is more likely to be out of date on it. But the ranking is more than a curiosity, because these are the facts that can influence whether someone buys from a business.
Incorrect pricing, for example, could see a customer arrive expecting a number that a company cannot honour. Quoted higher than the real price, and they may never make contact at all. That type of outcome doesn't show up in a website's analytics with 'AI misinformation' as the cause of the issue.
Wrong funding or financial details can reach investors, prospects and journalists doing background checks, and a company described as smaller or shakier than it is loses a potential prospect. Misstated information about tech and integrations can be a direct disqualifier in B2B. If a buyer is told that a company doesn't connect to the CRM they already run, that can rule them out before a salesperson has the chance to be involved.
A common thread here is lack of visibility for brands. A misled customer is unlikely to file a complaint about an AI answer; they'll simply go elsewhere.
What percentage of brands have found false facts
about them on AI chatbots?
Among brands that reviewed at least 20 facts about them, 75% found at least one fact to be false.
brands that checked 20 or more facts found at least one that was wrong.
Brands that only spot-check two or three facts often find nothing wrong. Without spending time prompting for a wider range of questions that customers would ask about them, a small sample of prompts risks misleading brands into thinking they are accurately represented by AI chatbots.
Which sectors are worst-affected by false AI facts?
Break our brand facts data down by sector and one group stands out. AI companies see false facts at nearly double the average rate (18.5%). They tend to be younger than those in the average sector, and there is less settled, authoritative information online about them for AI systems to draw on. When this happens, AI might guess an answer, and often that guess will be wrong.
| # | Sector | % false | |
|---|---|---|---|
| 1 | AI / Machine Learning | 18.5% | |
| 2 | Professional Services / Consulting | 10.8% | |
| 3 | SaaS / Software | 9.5% | |
| 4 | Travel / Tourism | 9.2% | |
| 5 | Health / Wellness / Medical | 8.4% | |
| 6 | Food & Beverage / Hospitality | 8.2% | |
| 7 | Marketing / Advertising | 8.1% | |
| 8 | Financial Services / Fintech | 8% | |
| 9 | E-commerce / Retail | 7.6% | |
| 10 | Media / Entertainment | 7.6% |
The countries where companies are worst-affected by false AI facts
New Zealand businesses have flagged the highest rate (15.7%) of false facts that AI chatbots provide about them. Germany (15%) and Canada (14.3%) follow close behind, with all three between half and two-thirds worse than the average country (9.2%). The two largest markets in the data sit at the safer end: the United States at 8.6% and the United Kingdom at 8.2%.
| # | Market | % false |
|---|---|---|
| 1 | 🇳🇿 New Zealand | 15.7% |
| 2 | 🇩🇪 Germany | 15% |
| 3 | 🇨🇦 Canada | 14.3% |
| 4 | 🇪🇸 Spain | 14.3% |
| 5 | 🇳🇱 Netherlands | 13.6% |
| 6 | 🇦🇺 Australia | 11.7% |
| 7 | 🇦🇪 UAE | 10.3% |
| 8 | 🇿🇦 South Africa | 10% |
| 9 | 🇸🇬 Singapore | 9.8% |
| 10 | 🇸🇪 Sweden | 9.3% |
Why would American and British businesses be better represented? This data can't directly answer that question, and it's not simply an English-language bias, which large language models are also known for. New Zealand and Canada are English-speaking too, yet they sit at the opposite end of the table from the US and UK.
It may be related to how much of the internet's information trail is US and UK-centric. If American or British companies are simply written about more, with more press coverage, more directory listings, more review sites, and more competitor comparisons, that is the corroboration AI relies on to state a fact more accurately.
And language might still play a part. English has historically dominated the text that large language models were trained on. GPT-3's corpus, for example, was over 92% English. That weighting shapes where these models read to gain information.
The smaller the business, the more AI gets wrong
Small businesses with 50 or fewer employees see a higher rate of false AI facts about them than medium or larger companies do.
Small businesses see false facts about 50% more often than large firms.
of small businesses found at least one false fact about themselves, the highest of any size band.
Where AI trips up on small firms
Why might this be? A small business often has a thinner digital footprint, so AI has less to work from and might fill the information gaps with guesses. Part of the fix is to make the moving facts (price, partners, offering) easy to find and consistent everywhere that they already have a web presence.
US brands come off better than most, except on price
The United States ranks outside of the ten countries with the highest rate of incorrect brand facts, but over one in twelve (8.6%) is still notably high. In the world's most competitive market, where a buyer comparing options in an AI answer sees only a handful of brands and a handful of claims about each, one wrong fact in twelve is enough to hand a considered purchase to whichever competitor happened to be described accurately.
of US brands' facts were flagged false.
of US brands that checked 20 or more facts found at least one that was false.
Where AI trips up on US brands
However, US brands see false pricing claims at 16.1%, well above the 13.5% global figure and nearly double these brands' own overall rate. A market with more competitors, more frequent promotions and more price changes gives AI more chances to be out of date.
Which AI engine gets brands wrong most often?
Among the AI engines that were graded most heavily by brands, Perplexity and Google AI Overviews displayed the highest percentage of false facts.
| # | AI engine | % false |
|---|---|---|
| 1 | Perplexity | 11.5% |
| 2 | Google AI Overviews | 11.4% |
| 3 | ChatGPT | 8.1% |
All three of these engines answer from a mix of what they learned in training and what they retrieve from the live web, and the balance shifts from one question to the next. What none of them has is a single authoritative record of your business. They choose between whatever is publicly available about you. If brand information is displayed inconsistently across different sources, then the risk of AI choosing the wrong information remains.
How brands can combat AI hallucinations
You are not responsible for changing how Large Language Models (LLMs) work, but you can start to change the information that feeds them. These five moves, drawn from Searchable's own AI-search playbook, are how brands correct the record. As section 2 showed, prioritise your volatile facts first (price, partners, offering), because those are what AI gets wrong most.
- 1Get your Organization schema and sameAs links right.
This is the fastest lever available. Structured data on your site tells AI, in machine-readable terms, exactly who you are, and sameAs links to your verified profiles (LinkedIn, social accounts) confirm it. Many brands skip this, so doing it well is a quick and genuine advantage.
- 2Tell the same story everywhere.
AI builds its picture of you by cross-checking sources, so make sure your name, location, founding date and description match across your site, social profiles, directories and press. Contradictions between sources are what produce wrong answers.
- 3Claim and populate a Wikidata entry.
The bar is lower than Wikipedia, it is machine-readable, and it feeds both Wikipedia and the data AI models train on. It is a high-leverage and under-used step.
- 4Earn credible third-party coverage.
Independent articles from trusted publications are the one lever that shifts what AI knows at the deepest level, beyond what it finds in a live search. It is slow, and it is what pays off over the long term.
- 5Review and correct the facts in-platform.
Check the facts AI is actually stating about you, flag the ones that are wrong, and keep your brand knowledge current so every downstream answer starts from the right place. Re-check periodically. This is what Searchable's Brand Facts and Knowledge Base are built for.
Methodology
- Searchable surfaces the facts AI engines state about a brand and lets the brand review each one and mark it correct or incorrect. This study draws on 32,556 reviewed facts across 1,704 brands that have participated up until 29th July 2026.
- ChatGPT accounts for 20,359 reviewed facts (8.1% false), Perplexity 6,676 (11.5%) and Google AI Overviews 4,881 (11.4%). Copilot, Gemini, Google AI Mode, Claude, Grok and DeepSeek make up the remainder.
- Each fact is manually graded by an employee of the organisation it describes. No automated counting.
- What “false” includes. False facts here fall into three groups: outdated but once-true details, mistaken identity where AI confuses one brand with another, and genuine fabrications. We use misinformation, false and inaccurate as the umbrella terms for all three, and reserve hallucination for the fabricated group only.
- Every figure for the share of brands that found at least one false fact, whether overall, by company size, or for the US, counts only brands that reviewed 20 facts or more. Fact-level error rates hold at roughly 9% at every cut-off.
Citation. You are free to report on these findings with credit to Searchable and a link to this page.
What is AI saying about your brand?
Three in four brands that looked properly found AI stating something false about them. See the facts ChatGPT, Perplexity and Google AI Overviews hold about your business, and correct the ones that are wrong.
- See every fact AI states about your brand
- Find false, outdated and mixed-up claims
- Track corrections across every major AI engine
Want to chat with an expert? Book a demo
Trusted by over 2,000 marketing professionals and agencies.
