Ask an AI where your shop is, and there is a real chance it will hand the customer the wrong postcode. In recent vendor testing of UK high street retailers, roughly one in ten AI answers carried a wrong postcode, and 64 percent of those businesses had at least one false fact circulating in AI responses.
Wrong opening hours. A closed sign on a business that is open. Services you have never offered. These are not edge cases anymore, they are close to the median experience.
A recent piece on Search Engine Journal walked through the test data, and the detail that should stop you is simple. Nobody is telling the business it is happening.
The Blind Spot Is Structural
Traditional SEO gave you a dashboard. Rankings moved, you saw it. Traffic dropped, Search Console flagged it. There was a feedback loop you could act on.
AI answers have no such loop. There is no report that pings you when Gemini tells a buyer you shut down, or when ChatGPT invents a service you do not sell. The error just sits there, quietly rerouting people, and the first sign is a customer who never arrives.
It gets harder because every model answers differently. ChatGPT, Google AI Overviews, Gemini, and Perplexity each pull from a different mix of sources, so the same question about the same business returns four different answers. You cannot check one and assume the rest are fine.
The size effect is the part that should worry smaller operators most. In the testing, about half of smaller businesses received false facts, against 32 percent of larger ones. The reason is structural, not bad luck. Bigger companies leave a thicker trail of consistent third-party data, so the model has more to triangulate against.
A small business with a thin, contradictory footprint gives the model room to guess. So it guesses, and it often guesses wrong.
Picture the real cost. A customer asks Gemini whether the local supplier is open on Saturday, gets a confident no, and quietly buys from someone else. No complaint, no bounce, no line in any report. The sale simply never existed, and the business never learns why.
Multiply that by every buyer who checks an AI before they check you, and you are looking at a slow, invisible leak in your funnel that predates any campaign you are running.
This connects to something I wrote about in being mentioned by AI is not being believed. Presence was never the finish line. Being described accurately is the actual bar, and most brands have not cleared it.
Treat It as a Data Problem, Not a PR Problem
The instinct is to panic and think about reputation. Wrong frame. This is a data hygiene problem, and data problems have procedures.
Start with an audit. Take the ten questions your best customers actually ask before they choose you. Where are you, what do you sell, are you open, what does it cost. Run each one across ChatGPT, Gemini, Perplexity, and Google AI Overviews.
Run each question two or three times. These systems are not deterministic, so one clean answer proves nothing about the next.
Then check what the model cited. When an AI names a source, that source is the leak. If it is pulling your hours from an outdated directory or a stale Google Business Profile, the fix lives upstream, at the source, not in the answer box.
Make your own data agree with itself. The single biggest cause of wrong AI answers is a business that contradicts itself across its own footprint. One address on the website, another on the Business Profile, a third on an old aggregator page. The model cannot tell which is true, so it picks, and it often picks the wrong one.
Consistency is not a nicety here. It is the raw input the machine reads before it describes you to a buyer. I made the broader version of this case in make your site legible to machines, and clean, non-contradictory facts are where that work starts.
I built GEOflux (geoflux.ai) precisely because this checking cannot stay manual as the number of models and questions climbs. Measuring where your brand stands inside AI answers, and why, is a standing job now, not a one-time cleanup you tick off and forget.
The Cost of Waiting Is Invisible
Here is the uncomfortable math. Every wrong answer is a buyer making a decision on bad information, and you are paying for it without ever seeing the invoice.
A closed-business claim does not generate an angry email. It generates silence. And silence does not show up in any dashboard you currently watch.
The brands that win the next two years are not the ones with the most content. They are the ones whose data is clean enough that the machine cannot get them wrong in the first place.
Check before your customers do. That one hour of testing is the cheapest market research you will run all quarter.