Fundamentals · 6 September 2026

By Eddie Eastwood, Founder

I Rebranded, Moved or Opened a Second Location, and AI Still Names My Old Business

Why does AI still name my old business after I rebranded, moved or opened a second location?

Because the change didn't erase the old record, it added a new one next to it. Before the change, an engine answering a question about the business had one identity to work with. After a rebrand, a move to a new address, or a second location opening, it has two: an older record with history, reviews, links and mentions behind it, and a newer one that is thin by comparison. Given a choice between an established record and a fresh one, an engine keeps returning the established record, even after the underlying business has moved on.

That's a different starting point from a business that has never been indexed anywhere. Cold-Start AEO: Getting Recommended by AI When You're Brand New covers the zero-records case: nothing to correct, only a presence to build from scratch. This article is about the opposite situation, where there's too much history rather than none, and the fix is about which of two records wins, not about creating one from nothing. It is also narrower than getting a business's facts described correctly in general. How to Become an Entity AI Recognises covers what it takes for an engine to treat scattered mentions as one confirmed business, and that groundwork still applies here. What follows measures one specific consequence of not having it in place after a real change: how often the old identity keeps winning.

We tested it: how often does an AI catch up with a real name change?

Rather than ask an AI model to judge our own copy, we built a test around a source of verified, dated fact that neither we nor the AI provider controls: the UK company register. Companies House publishes the exact date a company's registered name changed and what it changed from, which means a "did the record update" question has a checkable right answer.

We read 799 UK company records sampled across twenty everyday local-service search terms (trades, dental, veterinary, accountancy and similar categories), and found 160 recorded previous names across 133 of them. Filtering to companies still active, with a genuine name change (not a punctuation correction) completed at least 30 days before the test, and with names distinctive enough that a match couldn't be coincidental, left 14 companies. For each, we asked one AI assistant, on 5 September 2026, whether the old name was still trading under a different name now. As a separate control, we asked about the new name directly: is it a real UK company, and what's its registration number? That splits "can't resolve this business at all" from "can't connect the old identity to the new one".

The gap between the main test and the control is what separates the two failure modes. If the business were simply invisible under its new identity, the control would fail as often as the main test does. It doesn't: asked directly about the new name, the assistant got the registration number right for 6 of 14. The failure in the main test is specifically the link from old to new, not a blanket inability to find the business.

What the failures looked like

One company in the sample changed its registered name in July 2026. Asked whether its old name was still trading under a different one, the assistant answered that none of the results it had "show a name change or a 'previous name' entry", while citing that company's own filing history page as a source. The page it had open was the one recording the change; the answer said no change was recorded.

A second company in the sample failed differently. Asked the same question, the assistant concluded that the old name "was later reused by a different company," effectively deciding the business's own former identity now belonged to someone else. In that case the business was told a stranger owns its own history.

The single case that resolved correctly did so for a specific, checkable reason: the business's own current website still displayed its previous trading name alongside its current company number. Asked about the old name, the assistant found that bridge on the business's own site and made the connection. It was the only one of the 14 that resolved.

What this test actually measures

It shows that on one date, with one assistant, a small but carefully verified sample of genuine UK renames mostly kept losing to their own history, and that being findable under a new name and having that name correctly linked to the old one are two separate things. Fourteen companies sampled from a name search, not drawn at random from the register, make this a controlled test rather than a survey of UK businesses in general, a claim about rankings, traffic or leads, or a result that would necessarily repeat on another engine, another model, or a week later.

The fix sequence

None of the following is unique to AI visibility. It's the same groundwork Google recommends for a conventional site move and profile update, applied with the AI-answer version of the problem in mind.

1. Redirect the old presence to the new one, and keep the redirects up. Google's own guidance for moving a site recommends keeping redirects in place "for as long as possible, generally at least 1 year," to let signals transfer fully. A short-lived redirect removed too early reopens the exact gap this test measured.
2. Set expectations in weeks, not days. The same guidance notes that for a medium-sized site, "it can take a few weeks or more for Google to gradually start showing the new URLs instead of the old ones," and longer for larger sites. Treat an AI answer catching up on a similar or longer timescale as normal, not as evidence the fix failed.
3. State the current identity in structured data, not just in prose. Google's documentation on site names in Search advises adding WebSite structured data to the home page to indicate a name preference, since automated systems otherwise infer the name from page content and can't be manually overridden. The one case in our test that resolved did so because the current identity was published, plainly, on the business's own site.
4. Update every claimed profile to the current name and address. Google's own help documentation for a Business Profile states that changes "usually take about 10 minutes to show up" and "in some cases, it can take up to 30 days," which is a useful floor for how quickly a corrected profile feeds back into an answer.
5. Build enough independent mentions under the new identity that it stops being the thinner record. The mechanism throughout this test is comparative: an engine is choosing between two records, and the newer one needs enough independent weight, not perfection, to stop losing by default.

FAQ

Why does AI still name my old business after I rebranded, moved or opened a second location?

Because a rename, a move or a second site creates two competing records of the same business rather than removing the old one. The engine has both, and in our own test it kept choosing the older, more established record in 13 of 14 verified cases, even when the current registered name was a matter of public record.

Is this the same problem as a brand-new business not showing up in AI at all?

No, and the fix is different. A new business has zero records for an engine to find, which is a cold-start problem. A rebranded or relocated business has too many records: an old one that is well established and a new one that is thin by comparison. The engine has something to work with either way; the difference is which record wins.

How did you test this?

We sampled UK companies with a recorded name change, kept only genuine renames at least 30 days old, and put the old name to one AI assistant on one date. A separate control question asked about the new name directly.

Does this mean the engine can't find the business at all?

Not quite. Asked directly about the new name, the same engine gave the correct company registration number for 6 of the 14 businesses. What breaks is the join between the two identities, not the engine's ability to see the current one.

What actually fixes this?

Three things, in order: keep the old URLs pointing at the new ones for at least a year, per Google's own guidance; update structured data and Business Profile listings to the current name and address; and get the new identity stated in enough independent places that it stops reading as the weaker of two competing records.

How long before an AI answer catches up with the change?

There's no fixed timeline, and our test only measured a single date rather than tracking recovery over time. Google's own documentation for a site move sets expectations in weeks rather than days for a normal-sized site, and Business Profile edits can take up to 30 days to show. Treat both as a floor, not a guarantee, for how quickly an AI answer follows.

Short version, restated

A rebrand, a move or a second location doesn't erase the record an AI assistant already has; it adds a newer, thinner one beside it. In our test, the older record won 13 times out of 14, and the one exception won because the business itself published the bridge between its old and current identity. That bridge, plus redirects held for as long as Google's own guidance recommends, is the actual fix.

Find out whether AI still recommends your old identity or your current one.

Check whether AI platforms can find and cite your business →