How to Track a Rebrand in AI Answers

To track a rebrand in AI answers, start by updating your brand profile before launch so everything you measure afterward is testing the new name. Keep the old name as an alias so the trend line stays continuous. Write the new positioning down as three to seven key messages, annotate launch day and every milestone, then watch the two names trade places week over week across ChatGPT, Gemini, Claude, and Perplexity. When the old name lingers, open the citations behind those answers, rank the sources by influence, and correct them in that order. Expect the new name to become dominant one to three months after launch.

A rebrand is fast on your side and slow on the web's. Your site changes overnight. The hundreds of other pages that mention you keep saying the old thing until somebody goes and updates them, and those pages are what AI assistants read when they answer questions about you. The usual holdouts are predictable: a Wikipedia article, two or three industry directories, a comparison roundup from two years ago, and a wire service archive that nobody can edit. That last category is why the switch takes a quarter rather than a week, long after the internal celebration is over.

That gap is the problem this workflow solves. It's also the part nobody staffs for. Teams spend a year on the name, six weeks on the launch, and zero days on the question of whether the systems most buyers now ask about companies have any idea the change happened.

This guide works with any AI visibility tracking setup; the callouts show how each step runs in Spyglasses.

Step	What you're doing	What tells you it's working
1. Update the profile	Rewriting what the tracking knows about you	New prompts ask about the new company
2. Alias the old name	Counting both names as you	One trend line, not two broken ones
3. Write key messages	Turning positioning into trackable phrases	Adoption becomes a percentage
4. Annotate milestones	Dating launch, coverage, and cutover	Movement lines up with real events
5. Watch the swap	Reading both names week over week	The lines cross, and you can say when
6. Fix the sources	Correcting the pages carrying the old story	Old-name citations fall each month

Update the brand profile before launch day

Every tracking setup carries a description of your company somewhere. It's the category you're in, who you sell to, what makes you different, and who you compete with. That description drives which questions get asked on your behalf and which mentions count as you.

If you rebrand and leave that description alone, you spend the next quarter measuring a company that no longer exists. The prompts test the old category. The mention matching looks for the old name. The reports come back clean and mean nothing.

So rewrite it first, ideally the week before launch. Change the name, obviously. Then check the parts people forget. The category, if the rebrand moved you into a new one. The audience, if you're now selling upmarket. The differentiators, because those are usually the reason for the rebrand in the first place. And the competitive set, since a repositioning often changes who you're compared against, and comparison questions are where AI does the most damage or the most good.

One caution about timing. Update the profile before launch, but don't expect the data to change on the same day. What you're doing here is making sure that everything measured from launch forward is asking about the right company.

Keep the old name as an alias

This is the step that saves your reporting, and it's the easiest one to skip in the excitement of launch week.

The instinct is to delete the old name. You're done with it. But the old name is how AI will refer to you for months, and if your tracking only counts the new one, your visibility appears to collapse the day you launch. You'll be explaining a cliff to your CEO that isn't real.

Register the old name as an alias instead. Mentions of either name count as you, so the total holds steady while the mix underneath it shifts. That mix is the actual story. You want to be able to show a chart where one line falls, another climbs, and they cross on a date you can name.

Aliases catch more than the exact old name. Add the common variants people used, any product names that got folded into the new brand, and the legal entity name if it shows up in filings and directory listings. Directory data is stubborn and often quotes the legal name long after the marketing name changed.

While you're here, check one thing that matters more than the name count. Ask an assistant directly whether the old company and the new one are the same. If the answer says the old company was acquired, shut down, or is a competitor of yours, you have a bigger problem than slow adoption, and the fix is making the connection explicit in the places models trust most.

Turn the new positioning into key messages

Most rebrands are not really about the name. The name is the visible part; the point is usually that the company wants to mean something different. That intent is testable, but only if you write it down first.

Key messages are the short phrases you want repeated about you. Under ten words each, specific enough that a competitor couldn't use the same line, and true enough that you can defend them. Three to seven is the right number. A list of twelve is a document nobody reads.

For a rebrand, write two sets. The positive ones are the new framing you're trying to build. The negative ones are the outdated descriptions you're trying to retire, the old category label, the legacy product association, the audience you no longer serve. Tracking both gives you a real measure of adoption, because a rebrand isn't finished when the new story appears. It's finished when the old story stops.

Then set expectations with whoever is watching the numbers. AI describes a company by synthesizing what many sources repeat. A phrase that lives only on your homepage carries almost no weight against a hundred articles using the old language. Message adoption follows coverage, so it moves slowly at first and then faster once a few influential sources pick up the new framing. Getting that framing into a launch announcement is a whole discipline of its own, covered in optimizing a press release for AI citation.

Mark launch day and every milestone

Rebrand data is noisy for the first month. Something moves, three things happened that week, and nobody can say which one did it. Annotations fix that cheaply.

Mark the announcement date. Mark the day the site actually cut over, which is often not the same day. Mark the press release, any major coverage, the analyst briefing, the conference keynote, and the day you finished updating your third-party profiles. Each one takes ten seconds to record and saves an argument later.

The payoff shows up about six weeks in, when the chart has enough shape to read. You'll see a step change after one specific piece of coverage and nothing at all after another, and that difference tells you where to spend the rest of your outreach budget. Without the dates on the chart, all you have is a slope and a theory.

Annotations also protect you from taking credit for the wrong thing. Rebrand launches usually coincide with a website migration, and migrations move visibility all on their own, sometimes downward. A dated note is what lets you separate a positioning win from a technical side effect.

Watch the name swap and the message pull-through

Now you read the data, and there are two questions to keep separate.

The first is the name. Track the old and new names side by side and watch the ratio. Early on the old name dominates, because the web still says it. Over the following weeks the new one climbs. The date those lines cross is the most quotable number you'll get out of this whole exercise, and it's the one your executives will remember.

The second question is harder and more useful. When AI describes the company, is it using the new framing or the old one? A rebrand can move the name across in six weeks and leave the meaning untouched, so you end up with a new logo on the old story. Watching key message pull-through is what catches that, and catching it early is the difference between fixing it during launch and discovering it at the annual review.

Check brand consistency too. Rebrands reliably dent it, because for a while the sources disagree with each other about who you are. Some pages say the old name, some the new, some describe you as both. A consistency dip in month one is normal. A consistency dip that hasn't recovered by month four means the market is still confused about you, and that's a comms problem rather than a data problem.

Watch weekly, not daily. AI answers vary between runs even when nothing changed, so a single day's number is noise. Week over week, on the same set of prompts, is the shortest window that tells you anything real. The general practice of reading these trends is covered in tracking AI visibility over time.

Find the sources still telling the old story

At some point the trend stalls and you need to know why. The answer is always specific pages, and they're findable.

Open the answers that still use the old name and look at what they cite. You'll usually find the same handful of sources over and over. A Wikipedia article that hasn't been updated. Two industry directories. A comparison roundup from two years ago that ranks well and gets read constantly. An old press release still sitting on a wire service. Your own docs subdomain, which somehow never got the memo.

Don't work that list alphabetically. Rank it by how much influence each source actually carries with AI, which is a mix of the site's authority and whether AI crawlers can even read it. A high-authority site that blocks AI crawlers is not your problem right now. A middling site that gets cited constantly is.

Source type	Typical fix	How fast it lands
Your own stray pages	Update or redirect them yourself	Days, once recrawled
Wikipedia and open databases	Submit a sourced correction	Days to weeks
Directories and profiles	Claim the listing, edit the record	Weeks
Old articles and roundups	Ask the publisher for an update note	Weeks to never
Wire service archives	Usually permanent, counter with new coverage	Months

Two of those rows deserve a note. Your own stray pages are the most common finding and the most embarrassing one, because they're entirely within your control. Old subdomains, help centers, changelog archives, and PDF datasheets are where the old name hides.

And the last row is why new coverage matters more than corrections. You can't edit a wire archive, but you can publish enough new material using the new name that the old material stops being the strongest signal. That's the same source-influence logic behind connecting earned media to AI visibility, applied to a name instead of a campaign.

Work this list monthly and the effect compounds. Each corrected source removes one input feeding the old answer, and the answers shift as the inputs do.

Give it a quarter, and check monthly

The rebrand work that moves AI answers is unglamorous. It's a profile update, an alias, five phrases, a few dated notes, and then a monthly hour spent correcting sources in order of influence.

Set the expectation up front that the switch takes a quarter, because the alternative is a nervous executive asking why ChatGPT still uses the old name three weeks after a launch that cost real money. With the trend line and the annotations in front of you, that conversation gets easy. You can show the lines crossing, point at the coverage that moved them, and name the six sources still holding out.

The teams that handle this well treat a rebrand like a slow-motion version of a crisis communications response. Same shape, different urgency. Find what the machines are saying, trace it to sources, fix the sources that matter most, and measure whether the story changed.