Does PMax Cannibalize Your Search Campaigns? How to Measure It

Does Performance Max cannibalize your Search campaigns? Short answer: it can, and in accounts with meaningful brand demand it usually does by default. Cannibalization here means PMax serving on queries your existing Search campaigns - above all your brand name - would have caught anyway, at whatever price the PMax auction produced, and then booking the conversion as its own. The money is not always wasted outright; sometimes the same click just got bought through a different door. But the credit is always misassigned, and that is the expensive part: PMax looks brilliant, your brand campaign looks like it is fading, and every budget decision built on those numbers inherits the error.

The overlap is not a rare edge case. Optmyzr studied 503 accounts running both Search and Performance Max and found query overlap between the two in 91.45% of them. That figure measures presence of overlap, not its size - more on the hedging below - but it settles one question: if you run both campaign types, the burden of proof is on the claim that your account is the exception.

This guide is the measurement half of the PMax story. If you have not found the search terms report yet, start with where PMax search terms live and how to triage them; if you already know what you want to block, the mechanics are in our PMax negative keywords guide. Here we stay on one question: how much of your PMax performance is actually demand you already owned - and how to find out with numbers instead of vibes.

What “cannibalization” actually means here

Three distinct things hide under the word, and they deserve different reactions.

1. Brand cannibalization - the expensive one. Performance Max optimizes toward conversions, and the cheapest conversions in any account belong to people already searching for you by name. Left unguided, PMax gravitates to those queries, harvests near-certain conversions, and reports them as machine-found wins. If a brand Search campaign exists, the two campaigns are now fishing the same tiny pond; if none exists, PMax is often buying clicks that organic search would have delivered free. Either way, “PMax ROAS” is now partly a measurement of how strong your brand is, not how good the campaign is.

2. Query overlap with non-brand Search. PMax also serves on generic queries your Search keywords already cover. This is partly structural - PMax includes Search inventory by design - and it is not automatically bad. The question is which campaign wins the query, and at what cost per conversion.

3. The attribution shell game. Even when total account conversions stay flat, cannibalization moves credit between campaigns. A brand Search campaign that loses traffic to PMax reports declining volume; PMax reports growth. Nothing improved. Nothing necessarily got worse. But your reporting now tells a false story, and false stories get budgets reallocated toward them.

One mechanical clarification, because it confuses almost everyone: within a single account, Search and PMax do not both show an ad for the same query at the same time - Google picks one. As of mid-2026, Google’s stated priority rules are that a Search campaign wins the query when it has an eligible exact match keyword identical to the search term; otherwise, the ad with the higher Ad Rank enters the auction, from whichever campaign. The trap is the word eligible: a brand Search campaign that is budget-capped, or whose keyword loses eligibility at that moment, silently hands the query to PMax. So cannibalization is not “paying twice for one search” - it is the wrong campaign winning your own query, at different economics and with different attribution, without telling you.

Why you cannot see it by default

If cannibalization were visible, it would not be a topic. It is invisible in standard reporting for four compounding reasons.

Blended reporting. A PMax campaign reports one set of numbers across Search, Shopping, YouTube, Display, Discover, Gmail and Maps. A conversion from someone who typed your company name into Google sits in the same ROAS as a conversion from a cold YouTube viewer. Nothing in the campaign-level view distinguishes harvested demand from created demand.

The report exists, but the labels do not. Since 2025 Google shows you actual PMax search terms natively (here is where the report lives). What it does not do is mark which rows are your brand, which rows a Search campaign also bought, or which rows would likely have converted anyway. It hands you a few thousand unlabeled rows and walks away.

Part of the spend is a black box. Like every search terms report since Google’s privacy thresholds arrived, low-volume queries are aggregated away. A real share of your PMax spend belongs to rows you will never see. Any brand-share number - ours, yours, anyone’s - is honestly a share of the visible portion, and a measurement that does not say so is overclaiming.

The damage shows up in the wrong place. The visible symptom of PMax brand cannibalization is usually a Search campaign chart: brand impressions drifting down, brand CPC drifting up, “brand demand seems to be declining.” Nobody investigates a thriving PMax campaign to explain a fading Search one - which is exactly why this pattern survives audits.

How to measure it: the manual method

Everything below is doable in Google Ads plus a spreadsheet, in an afternoon. You are building four numbers: your brand share of visible PMax spend, your brand share of visible PMax conversions, your Search-overlap cost, and your black-box share. Decisions come after the numbers, not before.

Step 1 - Write down your brand inventory

List every string that means “us”: company name, product names, your domain with and without the TLD, common misspellings, spaced and joined variants (“keywordninja”, “keyword ninja”). Be explicit and keep the list - the whole measurement depends on this definition, and you want it visible so you can challenge it later, not buried inside a formula.

Step 2 - Pull the PMax search terms report

Open the campaign’s search terms report (the exact click path is in the search terms guide) and export at least 30 days - long enough to smooth out day noise, short enough that the campaign has not changed identity mid-window. Note the campaign’s total cost for the same window from the campaign view; you need it in step 4.

Step 3 - Isolate brand and compute the shares

Flag every exported row that matches your brand inventory (a case-insensitive “contains” pass catches most of it; eyeball the remainder). Then compute:

• Brand share of visible spend = brand-row cost ÷ total visible term cost
• Brand share of visible conversions = brand-row conversions ÷ total visible term conversions

The second number is usually the shock. Accounts routinely discover that a large slice of “PMax conversions” trace back to people who searched the company name - we have seen accounts on our platform where a great-looking PMax campaign was roughly 90% brand demand on the visible terms, with the machine-found incremental traffic amounting to a rounding error. That is one account’s pattern, not a benchmark; the entire point of this exercise is to get your number.

Step 4 - Account for the black box

Divide total visible term cost by the campaign’s total cost for the window. Whatever is left over is spend no visible query explains - your black-box share. If 60% of spend is visible and 40% is not, your brand share from step 3 describes the 60%. Say it that way, to yourself and to anyone you report to. Brand behavior inside the hidden portion is unknowable; assuming it matches the visible portion is a guess, and it should be labeled as one.

Step 5 - Run the Search overlap analysis

Export your Search campaigns’ search terms for the same window. Normalize both exports (lowercase, trim whitespace) and intersect them: which exact queries did both Search and PMax buy? For each shared query, compare cost per conversion on each side. Three patterns emerge:

• Search wins the economics. Same query, cheaper conversion in Search. Every PMax click on it is leakage - this is your negative-keyword shortlist.
• PMax wins the economics. Less common, but real - sometimes PMax’s auction entry genuinely buys the same query cheaper. Knowing this beats assuming either direction.
• PMax-only brand terms. Brand queries PMax bought that your brand campaign never saw - often long-tail brand-plus-product variants. These are the genuinely interesting rows: either your brand campaign has coverage gaps, or PMax is expanding on your name in ways you never sanctioned.

Step 6 - Apply holdout logic for the incrementality question

Steps 1-5 measure overlap. They cannot answer the deeper question - would those conversions have happened without PMax spending on them? - because campaign-level metrics only ever show reallocation of credit. The honest instrument is a holdout:

• Time-based holdout (works for most accounts): exclude brand from PMax for two to four weeks, and watch total account brand conversions - brand Search plus PMax plus organic if you track it - against a comparable baseline period. Not either campaign’s own numbers; the total. If the total holds while PMax spend drops, the excluded spend was non-incremental. If the total sags, PMax was capturing demand nothing else caught.
• Geo split (for larger accounts): apply the exclusion in a subset of comparable regions and difference the totals. Cleaner inference, more setup.

Two warnings that keep this honest: changing exclusions disturbs PMax’s learning, so expect turbulence and do not judge the first week; and never run a holdout across an obvious seasonality boundary and attribute the difference to the test.

What the data says - and how hard to lean on it

The one rigorous public number in this space is Optmyzr’s: across 503 accounts running Search and Performance Max together, 91.45% had query overlap between the two. Credit where due - it is a real study with a stated sample, which is more than most claims in this genre can say.

Now the hedging, because the number gets misquoted in both directions. It says overlap exists in roughly nine accounts out of ten. It does not say 91% of PMax spend is cannibalized, or that overlap costs those accounts money, or anything about your account’s overlap size. An account where PMax and Search share a handful of low-cost queries and one where PMax’s visible spend is mostly the advertiser’s own name both count as “overlap” - and they need opposite responses.

The honest synthesis: overlap is near-universal, its cost is account-specific, and no published average substitutes for the six steps above. Treat anyone quoting a universal “X% of PMax is wasted on brand” figure the way you would treat an unlabeled axis on a chart.

The judgment call: when overlap is fine and when it is theft

Measurement produces a number. What to do about it is a judgment call, and pretending otherwise is how tools end up auto-“fixing” accounts into worse performance. The framework we use:

Overlap is often fine when:

• You run no brand Search campaign and competitors bid on your name. Someone will win those auctions; PMax defending your own SERP can be a legitimate choice - as long as it is a choice, priced and reviewed, not a default you never noticed.
• PMax catches brand long-tail your brand campaign misses. Rigid exact-match brand campaigns leak brand-plus-product queries; PMax picking them up is coverage, not theft - though the better fix is usually widening the brand campaign.
• The shared generic queries convert cheaper through PMax. If step 5 says PMax wins the economics on a query, taking it away out of principle costs you money.

Overlap is usually harmful when:

• A dedicated brand Search campaign exists and PMax brand clicks convert at a higher cost. Same demand, worse price, muddier attribution - pure leakage.
• Reported PMax ROAS is driving budget decisions. If brand demand is inflating the number, you are systematically shifting budget toward the campaign that harvests and away from campaigns that create - the compounding version of the error.
• Your brand campaign is budget-capped. Remember the eligibility trap: a limited brand campaign silently forfeits queries to PMax at whatever the PMax auction charges.

The right tool for each verdict. If the verdict is “keep PMax off my brand,” use Google’s brand exclusions - brand lists attached in campaign settings, available for PMax since 2023, with Google maintaining misspelling and variant coverage, and they do not consume your negative-keyword capacity. If the verdict is “block these specific observed queries,” use negative keywords - since March 2025 the cap is 10,000 per campaign, so rationing is no longer an excuse (full mechanics here). Brand exclusions are the blanket; negatives are the scalpel.

And through all of it, keep the incrementality frame: judge changes at the account level. Any intervention that merely moves conversions between campaign columns changed your reporting, not your business.

How Keyword Ninja measures it - and what it refuses to do

Everything above works manually, and we just gave you the whole method. What our Performance Max suite changes is that the measurement runs continuously and arrives labeled - with some deliberate refusals built in, because this topic is full of tools quietly overclaiming.

• A brand split with the definition shown. We split each PMax campaign’s visible spend and conversions into brand and non-brand using one brand matcher, and we show you the exact seed terms counted as your brand - so you can see the definition and challenge it, instead of trusting a black-box classifier with the most important variable in the analysis.
• Real numbers and estimates, never mixed. Where Google exposes real per-term cost, we use it and say so. Where only matched-category data exists - which carries conversions but no cost at all - we report conversions and decline to invent a cost share. Two sources, each labeled, never blended into one plausible-looking lie.
• The black box, printed next to every brand number. Each brand share ships with the visibility figure it depends on - what fraction of the campaign’s real spend is even visible at term level. A brand share without its black-box share is a number quoted out of context, so we do not render one without the other.
• Labels, never silent filters. When our waste meter attributes PMax spend to overlap, it distinguishes spend overlapping your Search campaigns from spend that is purely your own brand name - and in both cases it relabels and shows. We never quietly exclude traffic, never auto-apply brand exclusions, never reclassify your conversions behind your back. The verdict card says what we found; the account stays exactly as you left it.
• One click if - and only if - you decide. If your verdict is to shield the brand, one action adds the observed brand queries (the exact strings your account was served on, not the raw seed words, which would over-block) as exact campaign-level negatives across your enabled PMax campaigns - deduped, capped, and with every single addition individually undoable from the change log. A bulk action you cannot reverse row-by-row is a trap; ours is N reversible rows.

Stated plainly, what it cannot do: it cannot see the queries Google hides (nothing can - we measure that blind spot instead of pretending to open it), and it cannot prove incrementality for you - a holdout remains the gold standard, and a dashboard that claims otherwise is selling something. What it does is make the measurement permanent and the response reversible.

FAQ

Does Performance Max steal from Search campaigns?

It can. Within one account Google serves only one of your ads per query - a Search campaign wins when it has an eligible identical exact match keyword, otherwise the higher Ad Rank enters, which is often PMax. The practical result, especially for budget-capped brand campaigns, is PMax absorbing queries your Search campaigns previously won - visible as PMax growth paired with unexplained Search decline. Whether that is “stealing” or acceptable substitution depends on the economics per query, which is what the measurement method above establishes.

Is PMax brand cannibalization always bad?

No. If you have no brand Search campaign and competitors bid on your name, PMax defending the brand SERP can be rational. Overlap on generic queries can even be profitable when PMax buys them cheaper. Cannibalization is bad when the same demand gets bought at a worse price, or when inflated PMax numbers drive budget decisions. The failure mode is not overlap - it is overlap you never measured and never chose.

How do I stop Performance Max from bidding on my brand?

Two instruments. Brand exclusions - brand lists attached in PMax campaign settings, available since 2023 - are the purpose-built tool: Google maintains variant and misspelling coverage, and they do not use up negative-keyword capacity. Negative keywords - up to 10,000 per campaign since March 2025 - are for blocking the specific observed queries you decide are leakage. Most accounts that exclude use brand lists for the blanket and negatives for stragglers the list misses; the PMax negatives guide walks through every level.

Do brand exclusions actually work?

Broadly yes, with edges worth knowing. They govern search-shaped PMax traffic - a YouTube impression has no query to exclude - and Google’s variant coverage, while good, is not instantly exhaustive, so verify by re-checking the search terms report a couple of weeks after applying them. Expect some learning turbulence right after the change, and judge results against total account brand conversions, not either campaign alone.

How much Search-PMax overlap is normal?

There is no honest universal number. Optmyzr found overlap present in 91.45% of 503 accounts - so having some is the norm - but the size of overlap varies with brand strength, campaign structure, and budget caps, and no published average predicts your account. A strong-brand retailer and a no-name startup can both be “normal” at wildly different overlap shares. Measure your own; it takes an afternoon manually, or runs continuously with tooling.

The one-sentence version of this whole guide: PMax cannibalization is a measurement problem before it is a settings problem, and every fix applied before measuring is a guess. Our free Google Ads audit runs the brand-overlap measurement on your own account - read-only, PMax included, with the black-box share stated - and the Performance Max suite keeps the split, the labels, and the reversible one-click response running after the audit. Measure first. Then decide. In that order.