AI-drafted answers about your business locations are often wrong, and the safest response is to check them before a customer does. A Search Engine Journal report published on 20 July 2026 highlighted how frequently AI tools return incorrect details about physical stores, and it pointed to a Google safeguard that lets business owners vet AI-written answers before the public sees them.
What the test found
According to the report, a vendor-run test found that AI tools returned at least one false fact about 64% of the UK high street retailers checked. The most common error was a wrong postcode. That figure comes from a test run by a vendor, not an official Google statistic, so treat it as a warning sign rather than a settled measurement. The pattern it describes still matters: an AI answer can sound confident and specific while quietly getting your address, opening hours, or services wrong.
A wrong postcode is not a small thing for a business that depends on foot traffic. It sends drivers to the wrong street, breaks map directions, and can push a store off the local results a nearby customer sees.
How Gemini drafts these answers
Google's Gemini now auto-drafts answers inside Search and Maps by pulling from a business's Google Business Profile, its website, its photos, and its reviews. Instead of only showing the fields an owner typed in, the system generates a written response from whatever it can gather across those sources.
That is convenient when the source data is clean. When it is not, the AI can blend an outdated hour from a review, an old address from a cached page, and a service you no longer offer into one confident answer. The person reading it has no way to tell which parts were verified and which were inferred.
The review queue safeguard
Google's 2026 safeguard is a review queue. AI-drafted answers stay pending in that queue until the business owner approves, edits, or rejects them. Nothing an owner leaves unreviewed is guaranteed to match reality, but the queue gives you a checkpoint before an answer reaches customers.
The catch is that the queue only helps if someone actually checks it. An answer left sitting in the queue is a decision by default, so the review step needs an owner.
How to audit your profile
Start with the source data the AI reads from, then check the answers it produces.
- Open your Google Business Profile and confirm the address, postcode, phone number, and opening hours line up with reality, including holiday hours.
- Check that the services and categories listed are ones you still offer.
- Read your own website and photos the way the AI would, and remove or update stale details it might pick up.
- Scan recent reviews for wrong facts that an AI answer could repeat, such as an old location mentioned in a customer's comment.
- Open the AI review queue and approve, edit, or reject each drafted answer rather than leaving it pending.
For a deeper walkthrough of profile signals and local ranking factors, our local SEO resources cover the setup in more detail.
What this means for Thai marketers
For a business that runs several branches, the risk multiplies with every location. A hotel group, a clinic chain, or a restaurant with three outlets has a separate profile, a separate set of reviews, and a separate AI-drafted answer for each one. One outdated branch address can send a guest to a location that closed.
The practical response is the same at every branch: keep the profile data accurate, then check the AI review queue for each location before customers act on an answer. Assigning a specific person to own that check for each branch keeps it from slipping. Teams handling SEO in Thailand can fold this into the local profile work they already do.
FAQ
Is the 64% figure an official Google number?
No. It comes from a vendor-run test cited by Search Engine Journal, not from Google. Read it as a signal that errors are common, not as a precise rate.
Does the review queue fix errors automatically?
No. It holds AI-drafted answers as pending until you approve, edit, or reject them. If no one reviews the queue, the check does not happen.
The task here is straightforward but easy to postpone: audit the profile behind every location, then clear the AI review queue before an answer goes live. If you want help setting up that check across multiple branches, Relevant Audience can walk your team through it.
