How to Track AI Visibility Through a Merger

Tracking AI visibility through a merger means watching two brand names at once, across ChatGPT, Gemini, Claude, Perplexity, and Google AI Overviews, from before the announcement until the acquired name stops showing up on its own. It takes seven steps.

Set the acquirer up as your tracked brand and the acquired company as a competitor, so you get two trend lines instead of one blurry total. Write three or four deal statements down as key messages, plus the wrong claims you expect. Build 15 to 25 questions covering customers, prospects, reporters, and employees, tag them, and run them nightly. Mark the announcement and the close on the timeline.

Then watch share of voice move from the old name to the new one, week by week; a typical transition runs from roughly a 30/45 split to 70/10 over five or six months. Pull the citations behind those answers, where three to eight sources usually carry the whole story. Expect three to nine months before AI answers describe the combined company correctly, because answers only change after the pages behind them change and get recrawled. When the acquired name drops under 10% share of voice and holds there for a month, fold it in as an alias so your trend line stays continuous.

Here's how it usually surfaces. Two weeks after close, a sales rep forwards a screenshot. Somebody asked an assistant about the company you just bought, and the answer describes it as an independent competitor of yours, with pricing from a page that hasn't been touched since last spring.

Why AI answers lag a deal by months

A merger is a legal event with a date on it. AI answers don't work that way.

Assistants build answers about your companies from pages they retrieved; press coverage, product pages, directories, review sites, comparison posts, old funding write-ups. On close day, every one of those pages still says what it said the day before. The deal announcement adds one new page to a pile of hundreds. Retrieval doesn't care that it's the newest one.

So the answer stays stale until three things happen in order. Pages change, whether that's yours getting updated or new coverage getting published. Crawlers come back and pick up the changes, which takes anywhere from a few days on a busy news site to several months on a directory listing. Then the changed pages start beating the old ones for the questions people ask. That sequence takes three to nine months end to end, and it runs at a different speed on ChatGPT, Gemini, Claude, Perplexity, and Google AI Overviews, because each one leans on a different mix of sources.

Comms feels this first, because the deal team will ask how the market is absorbing the news and the usual answers are soft. Press clip counts measure how much got written, not what people are being told. AI answers are what people are actually being told, at the moment they ask. This workflow makes that measurable, and it works with any AI visibility setup; the callouts show how each step runs in Spyglasses.

Step	What you're doing	What tells you it's working
1. Track both brands	Two names, two trend lines	A baseline that predates the news
2. Write the messages	Naming what answers should say	Three or four testable statements
3. Build the prompt set	Asking what each group will ask	Questions that pull real answers
4. Mark the milestones	Annotating announcement and close	Every move has a cause attached
5. Watch the name split	Reading the shift, not the total	One line falling, one climbing
6. Trace the sources	Finding who tells the deal story	A ranked outreach list
7. Retire the old name	Flipping competitor to alias	A continuous line after the merge

Track both companies as separate brands before the announcement

Set this up before anyone announces anything, if you're allowed to. A baseline collected after the news is not a baseline; it's the after picture with nothing to compare it to.

The setup is simple. The acquirer is your tracked brand. The acquired company goes in as a competitor. That feels strange to write down when you're about to own them, but it's the right structure, because a competitor gets its own share-of-voice line. You need both lines separately for the whole transition. Combine them too early and the interesting part disappears into a total that barely moves.

Two or three weeks of data before the announcement is enough. You want to know what a normal week looks like for each name so you can tell later whether a number moved because of the deal or because it always wobbles that much.

Keep the other real competitors in the list too. A deal changes your competitive set, and one of the more useful things this tracking shows is whether the market starts describing the combined company against bigger rivals than either name faced alone.

If neither company has visibility tracking running yet, start with the general setup and layer the deal work on top; baselining a brand's AI visibility covers that from scratch.

Write the deal story as key messages

Every deal comes with an approved narrative. It's in the release, the FAQ, and the talking points. The question nobody usually asks is whether any of it reaches the people asking assistants about you.

Pick three or four statements and write them plainly, the way a customer would say them rather than the way the release says them. "The two companies are now one." "The product is still supported and still sold." "Customers keep their existing contracts." "The combined company covers both markets." Those are testable. "Synergies across a complementary portfolio" is not, and no customer will ever phrase a question that way.

Then write the wrong versions. This is the part teams skip and regret. During a transition, assistants repeat old framing with real confidence, and the specific wrong statements are predictable enough that you can name them in advance.

• The acquired product was discontinued. Almost every deal generates this one, usually from a support page or a forum thread.
• The two companies are competitors. Comparison pages keep this alive long after close, because those pages are built to win exactly the questions you care about.
• The old pricing or old plan names are still current. Pricing pages get cached, quoted, and copied into roundups.
• The deal fell through, or is still pending. Announcement coverage outnumbers close coverage by a wide margin, so the pending version often has more material behind it.

Measure both lists on the same chart. The transition is done when the true statements are rising and the wrong ones are falling, and having them side by side turns a vague sense that things are improving into a number.

Add the questions customers, reporters, and employees will ask

A deal creates its own question set, and it's different from the questions you normally track. Nobody asks "what are the leading vendors in this category" the week after an acquisition. They ask what happened to the thing they were using.

Write for the four groups separately. They ask in different words and they land on different pages.

Who's asking	What they type	What a bad answer costs you
Existing customers	"Is [acquired brand] still supported?"	Renewal risk and support tickets
Prospects	"[Acquirer] vs [acquired brand]"	A comparison that treats you as two vendors
Reporters and analysts	"Who owns [acquired brand] now?"	Wrong facts repeated in coverage
Candidates and employees	"Is [acquired brand] still hiring?"	Recruiting friction during integration

Fifteen to twenty-five questions covers most deals. Include the plain ownership questions, the product continuity questions, the comparison questions that used to pit the two brands against each other, and a couple about the category, to see whether the combined company shows up where neither name did before.

Tag the whole set with something like merger or deal-name. That tag is what lets you filter every chart down to the transition view later without rebuilding anything, and it's what you'll export when the deal team asks for a readout.

Then put the set on a nightly schedule. A weekly spot check is too coarse during a transition, because the interesting weeks are the ones right after an event and you only get a handful of readings before the next event lands.

Mark the announcement, the close, and every migration milestone

This step takes about a minute per event and it's the one that makes the whole record usable six months later.

Mark the announcement. Mark the close. Mark the day the acquired site redirects, the day the combined product page goes live, the day the pricing pages merge, the day the support portals consolidate, and any major piece of coverage. Each one is a candidate cause for a change in the data.

The lag is long and uneven, which is what makes the marks worth the minute. When share of voice for the new name jumps in week fourteen, somebody will ask what caused it, and the cause is usually a page that changed in week nine. You will not remember week nine. The timeline will.

Watch share of voice move from the old name to the new one

Now you read the data, and the discipline is to read the split rather than the total.

The combined total is a comforting number and it hides everything. Two names that add up to the same share they had before can mean the transition is working perfectly or hasn't started at all. What you want is the crossover; the acquired name's line coming down while the acquirer's climbs, ideally crossing on a date you can point at. A healthy transition looks like the acquirer going from 31% to 68% of category share of voice over about 22 weeks while the acquired name falls from 44% to 9%.

Check four things on a weekly rhythm, filtered to your tag.

The split between the names. This is the headline. Report it as two numbers, always, and resist anyone who wants it collapsed into one.

Whether the assistants agree. It's normal for one assistant to make the switch months before another. A name that's flipped on Perplexity and stalled on ChatGPT usually means one high-traffic source hasn't updated, which is a specific and fixable problem rather than a waiting game.

Message pull-through. Are the approved statements showing up, and are the stale ones fading? A falling old name with no rising message means assistants stopped talking about the acquired brand rather than started talking about the combined one, and those look identical on a share-of-voice chart.

Comparison answers. Ask directly whether the two are the same company. If assistants still frame you as rivals four months after close, that's the single most damaging pattern in the set, and it points straight at comparison pages you need to displace.

Set expectations with the deal team before the first readout. The first month shows the announcement bump and almost nothing structural. Real movement shows up somewhere between month three and month nine. Saying that in week one is much easier than explaining a flat chart in week six.

Trace the sources telling the deal story

At some point watching stops being useful and you need to know which pages are producing these answers. That list is shorter than people expect. Three to eight sources usually carry most of what an assistant says about a deal.

Pull the citations from the answers about the merger and sort them into four buckets.

Deal coverage. The trade and business press pieces about the announcement. These are your best asset, and they're worth grouping together so you can watch them start getting cited rather than guessing.

Comparison and review pages. Written to win the exact questions you're worried about, and updated by nobody. This is where the "still competitors" framing lives.

Directories and databases. Funding databases like Crunchbase and PitchBook, company profiles, and listing sites. They're slow to update and they get quoted with more confidence than they deserve, because a retrieval system reads three sites saying the same wrong thing as corroboration.

Your own pages. The acquired company's old site, its blog, its pricing page, its support docs. If your own material is telling the old story, that's the fastest fix on the list, and it's worth confirming assistants can read the new combined pages at all; the AI readiness audit shows which of your pages parse cleanly.

Score the outside sources before you spend outreach time. Two things decide whether a correction is worth chasing; whether AI can crawl that site at all, and whether it already gets cited for questions in your category. A well-known publication that blocks AI crawlers in its robots.txt won't change an answer no matter who you know there. Getting this ranking right is the same work as any earned media program in this channel, and connecting earned media to AI visibility goes deeper on the scoring.

Retire the old name only after the answers catch up

Eventually the acquired name is being wound down and your tracking should reflect that. The mistake is doing it on the legal date instead of the data date.

Watch for the signal first. The acquired name is under roughly 10% share of voice, the crossover has held for a month or more, and the continuity questions come back describing one company. That's when the split view stops earning its space on the chart.

Before you change anything, export it. Once the acquired name becomes an alias of your brand, both names count as you, the totals hold steady, and the separate line for the old name goes away. That's the right end state for ongoing reporting and it does erase the shape of the transition. There's no way to rebuild that chart later, so the export is the record. It also happens to be the single best slide in the integration retrospective.

Then make the flip. Move the acquired name from your competitor list to your brand aliases, and add the variants while you're there; the legal entity name, any product names that got folded in, common misspellings that show up in directory data. Directory listings quote legal names for years after the marketing name is gone.

After the flip, your reporting returns to normal. One brand, one line, one set of messages, with the deal-tagged prompts folded into the regular set or retired. The tracking that would be worth keeping is the handful of continuity questions, because customers keep asking those long after everyone internally has stopped thinking about the deal. The mechanics of the name swap itself are the same ones a rebrand goes through, and tracking a rebrand in AI answers covers them in more detail.

What good looks like a year out

The value of this workflow shows up in the integration review, not in week two. By then you should have four things.

1. A dated crossover. The week the new name passed the old one in AI answers, with the events around it marked. This is the closest thing to a market-absorption date anyone will produce, and it's defensible because the collection method never changed.
2. A closed loop on at least one wrong claim. Pick the most fixable one, usually a discontinued-product claim coming from your own support docs, correct the source, and show the claim falling out of answers eight weeks later. One documented cause and effect beats a year of charts.
3. A source list you keep. The pages that told the deal story, scored and ranked. That list is the starting point for the next announcement, and it costs nothing extra to produce.
4. A clean handoff. One brand, aliases in place, continuity questions still tracked, the transition archived as an export.

The habit is fifteen minutes a week on the same filtered view. Most weeks the numbers barely move, which is normal for a transition that runs on recrawl cycles. The point is that when someone senior asks whether the market understands the deal yet, the answer is a date and a percentage rather than a clip count and a guess.