You Rebranded. AI Still Remembers the Old You.

Here is the direct answer: when you rebrand or move domains, search engines catch up in weeks and AI assistants catch up in months — sometimes longer. A redirect tells a crawler where a page went. It does not tell a language model that the company it learned about under one name is the same company now operating under another. Those are two different kinds of memory, and only one of them responds to a 301.

I watch this play out every time a client rebrands. The site migration goes cleanly. Rankings wobble and recover. And then somebody types the new name into ChatGPT and gets a shrug, or types the old name and gets a confident, detailed, completely outdated answer. The web moved. The model did not.

Why a rebrand breaks something a redirect cannot fix

A search engine's job is to find the current best page for a query. When you redirect the old URL to the new one, you have answered the only question it was asking. Authority flows across the redirect because authority, in that system, is attached to URLs and links.

An AI assistant is doing something different. It is assembling an entity — a stable idea of who you are, what you do, who you serve, and whether you are credible — out of everything it has absorbed about you, plus whatever it retrieves at the moment of the question. That entity is held together by name, by description, by the people attached to it, and by corroboration across independent sources. A rebrand attacks every one of those at once.

So you get a predictable split. The old name still has years of accumulated corroboration behind it: articles, directories, reviews, mentions, podcast appearances, somebody's roundup from 2024. The new name has your website and a press release. Guess which one the model trusts.

The three trails you have to rebuild, not just redirect

1. The redirect trail

This is the one everybody does, and it still matters — AI systems that retrieve live pages follow redirects like anything else. Keep them permanent, keep them one hop, and keep them for far longer than the year most teams plan for. I would not retire a rebrand redirect inside three years. The cost of keeping it is a config line. The cost of dropping it is a dead end in the one path that was still working.

2. The corroboration trail

This is the one almost nobody does properly, and it is the one that actually moves AI answers. Every independent source that describes you under the old name is a vote for the old entity. Until a meaningful number of those sources say the new name, the model has no reason to merge them.

The work is unglamorous: your directory listings, your professional association profiles, the chamber of commerce page, the vendor sites that list you as a partner, the conference bios, the podcast show notes, the local news piece from three years ago. You will not get all of them. You do not need all of them. You need enough independent sources saying the same sentence — that the company formerly known as X is now Y — that the connection stops being an assertion only you are making.

3. The entity-record trail

Structured records are the closest thing to telling a machine something directly. Your organization schema, your knowledge-panel data, Wikidata if you qualify for it, your profiles on the platforms that feed structured data outward. These are where an explicit former-name relationship can live rather than being inferred.

Schema has a property for exactly this situation. Use it. State both names in your organization markup, keep the old name present rather than scrubbing it, and let the record carry the continuity your prose cannot.

The mistake that costs the most: erasing the old name

The instinct after a rebrand is to purge. New logo everywhere, old name deleted from the site, old social handles abandoned, old pages removed because they are off-brand now. I understand the impulse. It is also the single most expensive thing you can do to your AI visibility.

The old name is not clutter. It is the bridge. It is the only token that connects a decade of accumulated credibility to the thing you are now. If you delete every instance of it, you have not completed the rebrand — you have orphaned your own history and started from zero, while the old entity keeps floating around in model memory with nothing pointing forward.

Keep a permanent page that says plainly what happened: the old name, the new name, the date, the reason, and the fact that it is the same company and the same people. Link to it. Let others link to it. That page is doing work no amount of new content can do.

A playbook that respects how slowly this actually moves

1. Before launch, write the continuity sentence once and use it verbatim everywhere — press, site, directories, bios. Variation is the enemy here; you want the same string repeated across independent sources, because repetition is what reads as corroboration.
2. At launch, ship the redirects, update organization schema with both names, and publish the permanent "we changed our name" page.
3. Week one to four, work the directory and profile list by hand. Every platform that lets you edit a name field is a cheap vote. Prioritize the ones that are themselves well-cited — the industry associations and established directories carry more weight than volume scraping.
4. Month one onward, pursue earned mentions that use the new name in a context where the old one is acknowledged. A podcast appearance where you say "we used to be X" is worth more than a press release, because it is somebody else's platform saying it.
5. Keep testing. Ask the assistants about both names monthly. You are watching for the moment the old name starts returning the new answer. That is the signal the entities have merged.

How long it really takes

My read, not a guarantee: expect live-retrieval surfaces to reflect the new name within weeks once the redirects and records are clean, and expect the model's own baseline knowledge to lag by considerably longer — a function of training cycles you do not control. Some assistants will get it right because they searched. The same assistant, asked the same question without searching, may still give you the old answer. That inconsistency is normal and it is not a sign your work failed.

What you control is whether, when a model does look, the evidence is unambiguous. That is the whole game. You are not persuading a system. You are removing every reason for it to stay confused.

If you are weighing a rebrand right now

None of this is an argument against rebranding. It is an argument for pricing it honestly. The visibility cost is real, it is front-loaded, and it is mostly paid in manual outreach rather than in design fees. Teams that budget for the corroboration work come out the other side intact. Teams that budget only for the logo spend the following year wondering why the AI still calls them by their old name.

About the author

Scott Tischler is the Founder & Chairman of AIrecommend.ai and a practitioner-authority on AI search and Answer Engine Optimization. With 20+ years in marketing technology — including American Express, MetLife, and UBS — and executive and professional study at Wharton, Harvard, and Oxford, he helps businesses become the ones AI recommends.