How often AI gets business information wrong: 2026 statistics
of companies had at least one basic fact about them hallucinated or missing entirely from an AI chatbot's answer, across 13,000+ test queries on ChatGPT, Perplexity and Gemini.
Source: Searchable AI accuracy study, July 2026Key takeaways
- 93% of businesses have at least one wrong or missing fact in AI answers (Searchable, 2026)
- 50% vs 32% - small businesses get fabricated facts far more often than large companies (Searchable, 2026)
- 27% - only about a quarter of phone numbers given by AI match the business's Google Business Profile (Seer Interactive, 2025)
- 100% vs 68% - Gemini's business profile accuracy against ChatGPT's and Perplexity's (SOCi, 2026)
- 1.2% of business locations get recommended by ChatGPT at all, vs 35.9% appearing in Google's local 3-pack (SOCi, 2026)
- 45% of consumers now ask AI for local business recommendations - up from 6% one year earlier (BrightLocal, 2026)
What's in this report
- 93% of businesses have at least one fact wrong
- The phone number problem: given 91% of the time, right 27%
- Gemini is right 100% of the time. ChatGPT: 68%
- 45% of your customers now ask AI first
- Where AI gets its facts about you
- 10-prompt self-audit: test what AI says about your business
- Methodology
- FAQ
- Sources
1. 93% of businesses have at least one fact wrong in AI answers
The largest accuracy study to date queried ChatGPT, Perplexity and Gemini more than 13,000 times about real companies and checked every answer against verified records. 93% of companies had at least one basic fact hallucinated or missing - and small businesses were hit hardest on every metric.
| Metric | Value | Source |
|---|---|---|
| Companies with at least one wrong or missing fact | 93% | Searchable, 2026 |
| SMEs receiving at least one fabricated fact | 50% | Searchable, 2026 |
| Large companies (500+ staff) receiving a fabricated fact | 32% | Searchable, 2026 |
| Brand name confused or misattributed - SMEs | 4% | Searchable, 2026 |
| Brand name confused - large companies | 0.7% | Searchable, 2026 |
| Confident fabrications - SMEs vs large brands | 5% vs 2% | Searchable, 2026 |
| Key information missing - SMEs vs large brands | 6.3% vs 4.8% | Searchable, 2026 |
| UK high street retailers with misstated facts | 64% | Searchable retail analysis, 2026 |
AI chatbots confuse small business brand names with other companies five times more often than large company names - 4% of SME answers versus 0.7%.
Searchable, 2026
The pattern is consistent: the smaller the digital footprint, the more the models guess. Service descriptions were accurate for 93% of large brands but only 87% of SMEs, and the fact types AI got wrong most often - company size, contact details and founding year - are exactly the details a potential customer acts on. The study checked answers against Companies House filings and official company profiles, so "wrong" here means verifiably wrong, not merely outdated. One honest caveat: the sample is London-based companies, and no comparable US-wide study exists yet - which says something about how young this research field is.
2. The phone number problem: given 91% of the time, right 27%
When Seer Interactive asked 7 AI models for brand phone numbers across 178 branded queries, the models almost always produced a number - but only 27% of those numbers matched the business's Google Business Profile.
An AI assistant would rather give your customer a number than no number. The gap between confidence and accuracy is the real risk: a wrong number is not a lost impression, it is a customer calling someone else - possibly a competitor, possibly an outdated line. Model choice matters here too. Measured against the brand's own customer service page, Gemini was accurate 89% of the time, ChatGPT only 68% - the least accurate of the seven models tested.
of the sources AI models cite when answering phone number questions are third-party sites - not the business's own pages. Your accuracy depends on other people's databases.
Seer Interactive, 2025
3. Gemini is right 100% of the time. ChatGPT: 68%
SOCi's 2026 Local Visibility Index - nearly 350,000 business locations across 2,751 brands - measured how accurately each AI platform reproduces basic business profile information. Gemini scored 100%. ChatGPT and Perplexity both scored 68%.
Gemini's perfect score has a simple explanation: it is grounded in Google Maps and Google Business Profile data, the same records businesses already maintain. ChatGPT and Perplexity assemble profiles from third-party sources, and roughly one answer in three contains an error. The same study measured how often platforms recommend businesses at all:
| Metric | Value | Source |
|---|---|---|
| Locations recommended by ChatGPT | 1.2% | SOCi, 2026 |
| Locations recommended by Gemini | 11% | SOCi, 2026 |
| Locations recommended by Perplexity | 7.4% | SOCi, 2026 |
| Locations appearing in Google's local 3-pack | 35.9% | SOCi, 2026 |
| Average rating of businesses ChatGPT recommends | 4.3 stars | SOCi, 2026 |
SOCi's framing: AI visibility is 3 to 30 times harder to earn than a local pack position. And note the sample - brands with 50 or more locations. If national chains with dedicated marketing teams reach only 1.2% recommendation rates on ChatGPT, independent businesses are almost certainly doing worse. This selectivity is precisely why AI search visibility behaves differently from classic rankings: the model is not listing ten results, it is naming two or three businesses it "trusts".
4. 45% of your customers now ask AI first
Wrong AI answers would not matter if nobody read them. In BrightLocal's 2026 survey of 1,002 US consumers, 45% had used AI tools to find local business recommendations in the past year - up from 6% the year before.
| Metric | Value | Source |
|---|---|---|
| Consumers using AI for local business recommendations | 45% | BrightLocal, 2026 |
| Same figure one year earlier | 6% | BrightLocal, 2025 |
| AI users who trust AI recommendations | 63% | BrightLocal, 2026 |
| Adults 30-44 asking AI for recommendations | 64% | BrightLocal, 2026 |
| Consumers 60+ doing the same | 24% | BrightLocal, 2026 |
| Using ChatGPT specifically for recommendations | 31% | BrightLocal, 2026 |
| Using Google AI Mode for recommendations | 23% | BrightLocal, 2026 |
| Relying solely on AI review summaries | 23% | BrightLocal, 2026 |
Two numbers from that table belong together: 63% of AI users trust the recommendations, and 23% rely on AI review summaries without reading the underlying reviews. Combine that trust with a 68% profile accuracy rate and the exposure is obvious - a meaningful share of buyers now receives unverified, sometimes wrong information about your business and acts on it directly. In my client work this shows up as a new complaint pattern: "a customer called angry because ChatGPT told them we offer X". The customer never saw the website. There was no click to track.
5. Where AI gets its facts about you
AI assistants do not invent most business errors from nothing - they repeat them from somewhere. Fixing AI answers means fixing the sources the models actually read.
ChatGPT has an official local data partnership with Foursquare, signed in December 2024. Industry measurements - not confirmed by OpenAI, so treat the exact figure with care - attribute roughly 60-70% of ChatGPT's local business results to Foursquare data, cross-checked against Yelp, directories and crawled web pages. Gemini reads Google Business Profile. Perplexity leans on its own crawl and review platforms. Three practical consequences follow.
First, an unclaimed Foursquare listing is now a business risk - for many businesses it quietly became a primary source of truth they have never once updated. Second, consistency beats volume: in Seer's data, 59% of phone number citations were third-party pages, so one stale directory entry can outvote your own website. Third, there is no complaint desk - you cannot ask a model to fix an answer, you can only correct the records it learns from. This source cleanup work - profile claiming, entity consistency, structured data - is the unglamorous half of GEO programs, and by these numbers, the half with the clearest ROI.
6. Test what AI says about your business: the 10-prompt self-audit
You cannot manage what you have not measured. Run these ten prompts in ChatGPT, Gemini and Perplexity - fresh chats, no logged-in context - and tick every answer that is fully correct.
Scoring: 9-10 means you are ahead of roughly nine in ten businesses. 6-8 means the classic fix list applies - claim and update Foursquare, Google Business Profile and Yelp, then align your website's contact and service pages. Below 6, treat it as an active leak: 45% of consumers are hearing these answers, and errors compound as models cite each other's sources. Re-run the audit monthly; answers shift as models refresh their data.
7. Methodology
This report aggregates every primary study measuring AI accuracy on business information published between December 2025 and July 2026: Searchable's 13,000-query accuracy study (July 2026), Seer Interactive's phone number study (178 branded queries, 7 models, December 2025), SOCi's Local Visibility Index (~350,000 locations, 2,751 brands, January 2026) and BrightLocal's Local Consumer Review Survey (n=1,002, 2026). Statistics were verified against the original publications or, where the primary report is gated, against at least two independent press accounts; each number above carries its source inline. A literature check via OpenAlex and Semantic Scholar (July 2026) found no peer-reviewed academic study measuring AI accuracy on individual business facts - industry research is currently the only evidence base, and its limitations are noted in the text.
Three widely-circulated figures were excluded as untraceable or misquoted, including the claim that "45% of customers use ChatGPT to find local services" - the underlying BrightLocal data says 45% use any AI tool; the ChatGPT-specific figure is 31%. Known limitations: the Searchable sample is UK-only, and SOCi measures brands with 50+ locations, so single-location businesses likely face worse numbers than reported here. This page will be updated as new primary data is published.
8. Frequently asked questions
How often does AI give wrong information about a business?
In the largest study to date, 93% of companies had at least one basic fact hallucinated or missing in AI answers (Searchable, 13,000+ queries, 2026). Half of small businesses received at least one outright fabricated fact - against 32% of large companies.
Which AI chatbot is most accurate about business information?
Gemini, by a wide margin. SOCi measured its business profile accuracy at 100% because it reads Google Maps data directly, versus 68% for ChatGPT and Perplexity. For phone numbers, Seer Interactive found Gemini right 89% of the time and ChatGPT only 68%.
Why does ChatGPT get my business hours or phone number wrong?
It assembles facts from third-party sources - industry measurements attribute roughly 60-70% of its local data to Foursquare, its official partner since December 2024 - and 59% of its phone number citations point to third-party sites. If those records are stale, the answer is wrong, confidently.
How many customers actually use AI to find businesses?
45% of consumers used AI for local business recommendations in the past year, up from 6% a year earlier (BrightLocal, 2026). In the 30-44 age group it is 64% - and 63% of AI users trust what the AI tells them.
How can I check what AI says about my business?
Use the 10-prompt self-audit above: ask ChatGPT, Gemini and Perplexity about your hours, phone, address, services and pricing in fresh chats, and score the answers against reality. Repeat monthly - model data shifts.
Can I force ChatGPT to correct wrong information about my business?
No. There is no correction mechanism - only source cleanup. Update what the models read: your website, Foursquare, Google Business Profile, Yelp and the major directories, and keep every record consistent. AI answers converge on what consistent sources repeat.
9. Sources
- Searchable. "AI chatbot accuracy study" - via SME Magazine. smeweb.com. Accessed July 20, 2026.
- Seer Interactive. "AI Models Provide Incorrect Phone Numbers 36% of the Time." seerinteractive.com. Accessed July 20, 2026.
- SOCi. "2026 Local Visibility Index" - via Search Engine Land. searchengineland.com. Accessed July 20, 2026.
- BrightLocal. "Local Consumer Review Survey 2026: AI Trust." brightlocal.com. Accessed July 20, 2026.
- Local Falcon. "ChatGPT Local Search Data Sources." localfalcon.com. Accessed July 20, 2026.
- resultsense. "AI chatbots misstate facts on 64% of UK retailers." resultsense.com. Accessed July 20, 2026.
Last updated: July 20, 2026. Compiled and maintained by Stanislav Peev. Found newer primary data on AI business-fact accuracy? Email me and I'll review it for the next update. Citing this page: link to this URL - the numbers above change as studies are updated.
Worried about what AI tells your customers?
I run AI-visibility baselines and GEO programs for businesses that want to be recommended - and quoted correctly - by ChatGPT, Gemini and Perplexity. Describe your situation in a few sentences and I'll reply personally with an honest first read. No calls, no meetings.
- Email[email protected]
- LinkedInin/stanislav-peev-seo
- Based inBratislava, Slovakia - working worldwide (EN · RU)