Performance Max will spend your budget on your competitors

The absolute figures below are rounded and the account is not named. The ratios are exact, and the ratios are the part that transfers.

A routine audit on a campaign nobody was worried about

The campaign was live. The budget was moving. The weekly report looked busy. Nothing in the account was flashing red, and that is what makes this one worth writing up.

I was running a search term audit on a Performance Max campaign for a premium consumer-goods brand expanding into new European markets. Strong recognition at home, close to unknown in the countries it had just launched in. I was not chasing a problem. It was a routine pass, not an investigation.

The search terms came back full of brand names. None of them belonged to the client. They were rival brands in the same category, plus a scattering of adjacent-category names with no connection to what this client sells. Over four weeks the campaign had spent around 3,000 euros and returned about 19 cents of revenue for every euro. Not 19 cents of profit. Nineteen cents of revenue, against a target of 5x.

The case for Performance Max, stated properly

Before taking it apart, the argument deserves its strongest form.

Google's pitch is that PMax sees what you cannot. It reads intent across Search, Shopping, YouTube, Display, Discover, Gmail and Maps at once, and puts budget where a conversion is most likely, in combinations no one would sit down and build by hand. For an advertiser entering a market with no keyword history and no performance data, that is genuinely attractive. You do not yet know which queries matter. PMax offers to find out faster than you can.

Sometimes it delivers exactly that. I am not going to tell you PMax does not work, and this is not an argument for switching it off. The problem is not that the machine is bad at finding signals. The problem is what it reaches for when the strongest signal available belongs to somebody else.

What the search term report showed

Sorting every query into three buckets, the client's own brand, competitor brands, and generic product terms, put most of the spend in the middle bucket. PMax had decided that people searching for other premium brands in the same category were a good audience.

Technically it was not wrong. Those people are in-market. They are also not available: they typed a specific competitor's name, which is about as clear a statement of intent as a shopper can make. The campaign was paying to put a brand nobody in that market recognised in front of people who had already decided what they wanted.

This is not a targeting mistake anyone made. It is what PMax does by default when the brand it is selling has no equity to anchor on. In a home market, PMax has strong brand signals to work with. Drop the same campaign into a country where the brand is unknown and it goes looking for the nearest strong purchase intent it can find. In that category, the nearest strong purchase intent was a competitor's name.

The same leak, running the other way

In August 2026, Nils Rooijmans, who writes a daily Google Ads newsletter, answered a reader question that describes this mechanism pointing in the opposite direction: "What to do if a PMax campaign overperforms because of brand keywords?"

Exclude branded searches via Brand exclusions and negative keywords. That way, PMax will not be able to go after these branded clicks with high conversion value that hide the under-performance of non-branded searches in your PMax.

Read those two cases together and they are the same defect. PMax does not separate branded demand from non-branded demand unless you make it. Which direction it hurts you depends only on whose brand is nearby.

If your brand is weak in that market, PMax reaches for competitors and your ROAS collapses, which at least is visible. If your brand is strong, PMax harvests your own brand searches, the ones that would have converted anyway, and the campaign average looks excellent. That version is worse, because a number that looks excellent is a number you scale.

What this means for your account

A single PMax ROAS figure is usually an average of two different businesses: demand you created and demand you captured. Averaged together, one hides the other, and you cannot tell which one your budget increase is actually funding.

Three things follow, and the third is the one that catches people.

• New market launches amplify it. Low brand recognition means fewer brand signals to anchor on, so the drift toward competitor terms happens faster, in exactly the campaigns with the least historical data to sanity-check against.
• Nothing surfaces it for you. PMax publishes a search term report, but it does not raise a flag when its spend drifts. The report is the only window, and it only works if somebody opens it.
• A healthy-looking PMax campaign is not evidence of anything yet. If you have never split branded from non-branded inside that campaign, a strong ROAS and a weak one are equally uninformative. You are reading a blended number.

There is a second-order cost too. If you are running a dedicated brand Search campaign, an unguarded PMax campaign competes with it for your own brand queries, and you end up paying twice for traffic you already had.

What to do this week

None of this needs a new tool or a budget conversation. It is four passes through the interface and one recurring appointment.

1. Pull the search term report on every PMax campaign. Open the campaign, go to Insights, then the search terms view. Sort every query into three buckets: your brand, competitor brands, generic. If the middle bucket has meaningful spend in it, you have found your answer, and you found it in about ten minutes.
2. Apply brand exclusions. Open the PMax campaign, Settings, then Brand exclusions (on some accounts it sits under Additional settings then Brand safety). Choose Exclude specific brands and add them. Two different jobs here: add your own brand to every non-brand PMax campaign so it stops eating your brand traffic, and add the competitor brands you just found so it stops funding theirs. If a brand is not in Google's list, add it as a custom brand name.
3. Back the exclusions with negative keyword lists. Tools, then Shared library, then Negative keyword lists. Build one list for your own brand terms and one for the competitor terms, including misspellings and abbreviations, and apply them. This step is not optional, for the reason in the next section.
4. Separate brand into its own Search campaign. Brand queries belong in a dedicated Search campaign where you can see them and control them. PMax handles discovery and product-level traffic. That split is what gives you a non-brand number you can actually steer toward.
5. Put a weekly search term review in the calendar. Weekly while a campaign or a market is new, monthly at the absolute minimum after that. This is not a one-time fix. Remove the guardrails and PMax will find new places to spend.

What this does not settle

If you are running Performance Max, open the search term report before you close this tab and count how many of your top twenty queries carry a brand name that is not yours. What number did you get, and was it a campaign you would have called healthy that morning?

If the answer turns out to be uncomfortable, the fix is in the list above and it costs nothing but an afternoon. If you would rather have someone go through the account with you, that is what the first pass of my process is, and there is a related piece here on why the popular fix is not always the one that moves anything.