How Does AI Search Handle a Company Rebrand?

Why does ChatGPT still call my company by its old name after a rebrand?

AI assistants describe your company from two sources: what they learned during training and what they retrieve from the web right now. A rebrand changes neither on its own. Until enough pages connect the new name to the old one, the model keeps using the name it has seen most often.

This question tends to show up a few weeks after a rename. The new logo is live, the domain redirects work, the press note went out, and then someone asks ChatGPT or Perplexity about the company and gets the old name back. Sometimes it is worse. The assistant describes the old product, the old pricing model, or a founder who left two years ago.

This is not a bug you can file a ticket for. It is how these systems build a picture of an entity. The good news is that the fix is mostly in your hands, and most of it is the same work that makes a rebrand land with people too.

What does an AI engine actually know about your company?

An AI engine knows your company as an entity: a name tied to a set of facts, like what you sell, who you serve, and where you are mentioned. That picture is assembled from many pages. Your own website is one input, and often not the loudest one.

Think about where your old name lives. It sits in directory listings on G2 and Capterra, in a Crunchbase profile, in LinkedIn company pages, in old podcast show notes, in guest posts, in GitHub readmes, and in every blog that ever linked to you. Each of those is a small vote for the old name. When a model like Gemini or Claude forms an answer, it is weighing those votes, whether it is drawing on training or on pages it just retrieved.

Your homepage saying the new name loudly is one vote. A hundred stale profiles saying the old name quietly is a hundred votes. That imbalance is the whole problem in one sentence.

Why do training snapshots and live retrieval behave differently?

Training is a frozen snapshot, so anything learned before your rebrand stays in the model until it is retrained. Live retrieval pulls current pages at answer time, so it can pick up the new name quickly, but only if the pages it finds say the new name clearly and connect it to the old one.

This split explains the confusing behavior founders notice. Ask the same question in two tools and you get two answers. One tool leans on what it remembers. Another runs a search first and quotes what it finds. Even inside one assistant, the answer can change depending on whether the question triggers a web lookup.

You cannot speed up a retraining cycle. You can control what live retrieval finds. So I put almost all of the effort there, because it is the part that responds to work this month rather than whenever the next model ships.

How do you tell AI search that the old name and the new name are the same company?

You state the connection plainly in text and in structured data, on pages engines already trust. A sentence like "Acme, formerly known as Brightline" does more for entity resolution than any logo change. Repeat it on your About page, your homepage footer, your press page, and your Organization schema.

The phrase "formerly known as" matters because it is how people write about renamed companies, so it is how models learn to link two names. I put it in the first paragraph of the About page, not buried in a history section. I add it to the meta description of the About page for a few months. I make sure the press page has a dated announcement that says, in one clean sentence, what changed and why.

On the structured data side, Organization schema gives you the name property for the new name, an alternateName property for the old one, and sameAs links pointing at your official profiles. That markup will not fix anything on its own, but it removes ambiguity for any system that reads it. I covered how assistants read that page in more depth in how AI assistants read an About page.

Which third-party pages should you update first?

Update the profiles that AI answers already cite when someone asks about you. Usually that means review sites, directory listings, LinkedIn, Crunchbase, and any comparison page that ranks for your category. Ask an assistant about your company, note which sources it leans on, and work down that list first.

This is the boring part, and it is where most rebrands stall. Nobody owns the list of profiles. Marketing assumes sales updated LinkedIn. Sales assumes marketing handled G2. The founder assumes it all happened because the announcement went out.

My method is a simple sheet with one row per profile: the URL, who has the login, whether the name is updated, whether the description is updated, and whether the old name appears anywhere on the page. I prioritize by visibility, not by ease. A profile that shows up in an AI answer gets fixed this week. A forgotten directory from years ago can wait.

For pages you do not control, like old guest posts or partner pages, a short, polite email asking for a name update works more often than you would expect. Offer the exact replacement sentence so the editor only has to paste it.

Should you keep the old domain and old brand pages alive?

Keep the old domain and redirect every old URL to its closest new equivalent with a permanent redirect. Do not let old pages die into errors. Redirects carry the history of those URLs to the new name, and they tell crawlers that the move is deliberate rather than accidental.

I see two mistakes here. The first is redirecting every old URL to the new homepage, which throws away the topical meaning each page had. The second is letting the old domain lapse after a year to save a renewal fee. Someone else can then buy it, and whatever they publish there becomes part of how the web describes your old name.

I also keep one page on the new site that explains the rename in plain language. It answers the questions people and machines both ask: what the company used to be called, when the name changed, and whether the product is the same. That page becomes the source an assistant can quote when someone asks what happened to the old brand.

How long does it take for AI answers to catch up?

There is no fixed timeline, and anyone who promises one is guessing. Answers that rely on live retrieval tend to update once the pages they cite are fixed. Answers that rely on training memory update only when the model is retrained. Plan for a long tail, and measure rather than assume.

What I do instead of guessing is keep a short list of prompts and run them on a schedule. "What is the new name of Brightline?" "What does Acme do?" "Who are the alternatives to Acme?" I run those in ChatGPT, Perplexity, Gemini, and Claude and log the answers in the same sheet as the profile updates. That way I can see which fixes moved which answers.

If an answer stays wrong for months, the cause is almost always a specific source that still says the old thing. The log helps you find it. I wrote about a related problem in why AI answers about your company are out of date, and the same detective work applies here.

What should you avoid doing during a rebrand?

Avoid changing the name, the positioning, the URL structure, and the product description all at once. Every change you stack makes it harder for people and engines to see continuity. Avoid deleting old content, avoid vague announcement copy, and avoid hiding the old name as if it were something embarrassing.

A lot of founders treat the old name as baggage. I understand the instinct. But the old name is also the name your existing customers, investors, and search history know. Hiding it breaks the bridge that lets everyone, including AI engines, carry their trust across.

I also tell clients to resist rewriting every page in a new voice on the same day. If the rename comes with a real pivot, fine, say so. If it is mostly a new name for the same product, keep the product pages stable so the only thing that changes is the label. Stable pages make the connection obvious.

A brand new company has the opposite problem, which I covered in how AI search treats a new company with no third-party mentions. A rebrand at least starts with history. Your job is to point that history at the new name.

What should you do next?

Start with one sentence. Write "formerly known as" on your About page, your homepage footer, and your press page today. Then list every third-party profile, rank them by whether AI answers cite them, and fix the top ten. Finally, set up a small prompt log so you can see answers change.

A rebrand is a go-to-market event, not a design project. The logo is the easy part. The hard part is the slow, specific work of telling every system that already knows you what your name is now. I build websites and the automations behind them for lead generation, and the profile tracking sheet and prompt log I described above are the kind of thing I set up for clients so the work does not depend on someone remembering.

If you are planning a rename, or you already did one and the AI answers have not caught up, reach out. I am happy to look at what the engines are saying about you and tell you where I would start.