Is Performance Max cannibalizing your branded search?

You look at the account and Performance Max is the hero — a ROAS two or three times everything else, scaling like it can't miss. It feels great until you ask the uncomfortable question: how much of that would have happened anyway? For a lot of accounts, the answer is "most of it," because PMax has quietly parked itself on top of your branded search, converting people who were already coming to buy.

This isn't a bug. It's the auction working exactly as designed — and it's the single most common way a Google Ads account tells you a comforting lie. Let's break down how it happens, how to spot it, and what to actually do about it.

And it matters more the better PMax looks. The accounts most exposed to this are the ones where PMax is the star performer, because a big, shiny ROAS is exactly what a big pool of branded demand produces. If your PMax numbers seem too good to argue with, that's the account most worth auditing — not the least.

How the cannibalization happens

Performance Max is a goal-seeking machine. You give it a target — usually ROAS or CPA — and it hunts across every Google surface (Search, Shopping, YouTube, Display, Discover, Gmail, Maps) for the conversions that hit that target most efficiently. And the most efficient conversion in your entire account, every single time, is the person who searched your brand name.

Someone typing "[your brand] running shoes" has already decided. They know you, they want you, and they were going to land on your site whether or not an ad greeted them. PMax doesn't know that context — it just sees a query with a sky-high conversion rate and a rock-bottom cost, and it does what it was built to do: it bids, it wins, and it books the sale as its own.

PMax isn't creating that demand. It's standing in the doorway and taking credit for everyone who walks through.

The result is a slow, invisible transfer. Traffic that would have arrived through organic listings or a cheap branded search campaign gets rerouted through PMax — at a higher cost per click and with the conversion credit reassigned. Your blended numbers might look fine. Underneath, you're paying a premium to buy customers you already had.

There's a second-order effect worth naming, too. Because PMax optimizes toward your ROAS target and branded conversions make that target trivially easy to hit, the algorithm learns that the cheap path is the winning path. It leans harder into brand, spends less time working the genuinely hard prospecting that grows the business, and reports back a number that says everything is wonderful. You end up rewarding the campaign for doing the one thing you didn't need it to do.

How to spot it

Cannibalization hides inside a good-looking average, so you have to go looking for it deliberately. Three signals give it away.

1. One asset group or channel with a wildly higher ROAS than the rest

Pull performance by asset group and by channel. If one slice is posting a ROAS that's three, five, ten times the rest of the campaign, that's not brilliance — that's almost always brand demand hiding inside it. Real prospecting performance is lumpy but earthbound. A single outlier that's an order of magnitude better than everything around it is the tell.

2. A jump in branded impression share

Check your branded search impression share over time. If it climbed right as you launched or scaled Performance Max, PMax is now competing in an auction you used to own outright. You're bidding against yourself — buying clicks that used to be cheap, or free, at auction prices.

3. Your own brand name in the reports

Performance Max search-term visibility is thin by design, but it's not zero. Dig into the search terms and search categories reports, plus the "insights" surfaced for the campaign. If your brand name and obvious variants keep showing up as top converting terms, you have your answer in black and white.

None of these is conclusive alone. Together, they're a pattern. This is exactly the kind of thing worth checking as part of a broader Performance Max audit — cannibalization rarely travels alone, and the same review usually surfaces two or three other leaks.

How to fix it

You have a ladder of options, from light-touch to structural. Pick the rung that matches how much control you need.

• Account-level negative keywords. Add your brand terms as account-wide negatives. Performance Max respects them, so this is the fastest way to fence PMax off your own name without touching the campaign structure.
• Campaign brand exclusions and brand lists. Google's brand exclusion feature lets you attach a brand list to the Performance Max campaign so it stops serving on your brand (and, if you want, competitor brands). It's the purpose-built lever for exactly this problem.
• Split brand into its own campaign. The cleanest fix: run a dedicated branded search campaign, exclude brand everywhere in PMax, and let each campaign do one job. Now you can see — and control — exactly what you pay for branded versus non-branded demand.

For most accounts, the right answer is a combination: brand exclusions on the PMax campaign plus a dedicated branded search campaign to catch that demand cheaply and on purpose. You're not trying to stop capturing branded traffic — you're trying to stop overpaying for it and mislabelling where it came from.

One caution before you flip everything at once: make the change deliberately and give it room to settle. Performance Max re-learns after a structural change, so expect a week or two of noisy numbers while spend redistributes. Change one lever at a time where you can, and note the date — you'll want a clean before-and-after when you measure, and a pile of simultaneous edits makes it impossible to tell which one moved the result.

How to measure incrementality afterward

Here's the part most people skip, and it's the whole point. Once you exclude brand, Performance Max's reported ROAS will drop. That's not the fix failing — that's the fix working. You've stripped out the cheap branded conversions that were flattering the number. What you're left with is closer to PMax's true incremental contribution.

So watch the right things after you make the change:

• Total sales, not campaign ROAS. If revenue holds while PMax's ROAS falls, the demand simply shifted to a cheaper channel — you saved money and lost nothing.
• Blended CAC across the account. The number that matters is what it costs you to acquire a customer overall — not what any single campaign claims in isolation.
• Branded vs. non-branded split. Now that they live in separate campaigns, you can finally read them separately — and make decisions on the non-branded number that actually reflects new demand.

The honest framing: a campaign's reported ROAS is not the same as its incremental value. Branded conversions are the cheapest in the account, so any campaign allowed to absorb them will look like a star. Take them away and measure again — that's the only way to know what PMax is really worth on top of the demand you already own.

Reported ROAS vs. incremental value

It's worth seeing the two side by side, because this gap is the entire trap:

The left column is what the platform hands you. The right column is what you can actually bank — and the only one worth scaling on.

This is the same discipline behind a proper Google Ads account audit: never trust a headline number until you've asked what it would have done without the spend. Branded cannibalization is just the most expensive place that question goes unasked.

Why an analyst catches this and a dashboard doesn't

A reporting dashboard will show you PMax's glorious ROAS and stop there — the number is technically correct, so nothing flags. Catching cannibalization takes a step of reasoning a dashboard never makes: cross-referencing impression share, spotting the outlier asset group, and asking whether the conversions are incremental at all. That's judgment work, and it's exactly what an AI media buyer is built to do — read the whole account, distrust the flattering number, and tell you where the credit is really coming from.

If you want that read on your own account — where PMax is padding its numbers with brand, and what it's actually worth without it — request a demo. Fifteen minutes, connected read-only, and you'll see the real number.