If you run an operating store on Google Ads, "exclude keywords" is the wrong mental model to start from. Performance Max is not a keyword campaign. It serves across Search, Shopping, Display, YouTube, Gmail, Discover, and Maps from your feed and asset groups, so the thing you are actually trying to suppress depends on which surface the waste is coming from.
Most ranking guides on this topic walk you through clicking "add negative keyword" and stop there. That misses the part that moves profit. This article covers all three exclusion controls, where each one lives, and — the part everyone skips — the arithmetic that tells you whether an exclusion is helping or just shrinking your volume.
The three ways to exclude in Performance Max
Negative keywords (the search-query filter)
Negative keywords stop your ads from showing on searches that contain the term. The critical limitation almost no beginner guide states plainly: in Performance Max, negative keywords only affect the Search and Shopping inventory, not Display or YouTube (Search Engine Land). If your wasted spend is a YouTube placement, a negative keyword does nothing.
For a long time PMax capped negatives at a near-useless 100 per campaign. Google raised that limit to 10,000 negative keywords per campaign in early 2025 (Search Engine Land), which is what made genuine query hygiene possible instead of a token gesture.
Brand exclusions (the cannibalization fix)
Brand exclusions are a different tool entirely, and for most operators they matter more than the negative list. PMax chases the cheapest conversions it can find, and your own branded searches are the cheapest conversions in the account. Left unchecked, PMax serves on "yourstorename mugs," claims a conversion the customer was already going to make, and reports a gorgeous ROAS while adding nothing incremental.
Account-level brand exclusions, available since 2024 and covering brand variants and common misspellings (AdNabu), are how you stop that. This is a settings control, not a keyword you type into the negative box.
Placement and audience exclusions (the Display and YouTube side)
When the waste is a junk mobile-app placement or an irrelevant YouTube channel, you exclude the placement, not a keyword. These controls shape where PMax serves on the non-search surfaces, and they are the right lever when your channel report shows spend leaking into Display you never wanted.
Where to add each one, and the real limits
Campaign-level negatives live under your PMax campaign in "Audiences, keywords and content," then "Search keywords." You can also build negative keyword lists in the shared library and apply one list across multiple campaigns, which is how you keep a standard junk list consistent. Account-level negatives sit in account settings and should be reserved for universal exclusions you never want on any campaign.
Brand exclusions are set separately in the campaign's additional settings, under brand exclusions — the same place covering your branded variants (Astraloop). Keep these two mentally separate: a negative keyword filters an arbitrary query, a brand exclusion protects a specific asset you already own.
When excluding keywords protects profit — a worked example
Exclusions earn their keep only when the spend they remove was genuinely unprofitable. To know that, you need your break-even ROAS, which comes from your margin, not from a benchmark blog.
Say you sell a mug at a $34 price. Your base and fulfillment cost is $14, your shipping share is $4, and processing fees run about $1.30. That leaves a contribution of $34 − $14 − $4 − $1.30 = $14.70, or a 43% contribution margin. Your break-even ROAS is 1 ÷ 0.43 = 2.3x. Below that, every order loses money.
Now suppose your PMax search-terms report shows the campaign spent on "free mug svg template" — people hunting for a free download, not buyers. Say that query pulled 300 clicks last month at a $0.90 cost per click. That is 300 × $0.90 = $270 in spend producing essentially zero orders.
Excluding that one query doesn't just save $270. It removes 300 zero-value clicks from the data your smart bidding learns on, which lets the system re-solve toward buyers. That is the real return on a good exclusion: cleaner signal, not just saved pennies. The same profit-first logic underpins why ROAS alone hides your true Google Ads economics.
When excluding keywords quietly costs you volume
Here is the trap the step-by-step guides never warn you about: a broad-match negative keyword blocks far more than you think. Add "cheap" as a broad negative and you also kill "cheapest custom mug" and "cheap gift under twenty dollars" — queries that can convert fine. The negative keyword looked like a cost cut; it was a volume cut.
Three rules keep exclusions from backfiring. First, prefer phrase or exact negatives for anything ambiguous, and reserve broad negatives for words that are unambiguously junk ("free," "jobs," "pdf"). Second, exclude from the search-terms report, not from your imagination — remove queries you can see wasting money, not queries you assume are bad. Third, give the campaign one to two conversion cycles to recalibrate after a batch of exclusions before you judge the result; a dip right after a change is the system re-learning, not a failure.
Chasing exclusions for their own sake buys you the same false comfort as chasing impression share. Both feel like control. Neither is a goal.
The brand-exclusion decision most operators get wrong
The signature symptom is a PMax campaign reporting a spectacular ROAS while your total business orders stay flat. That is almost always brand cannibalization: PMax absorbed branded searches your organic listings or a dedicated brand campaign were already winning.
The cost mechanism is stark. As the retail-Shopping specialists at smec put it, unprotected PMax can have you paying "$1.50 for a click that you could have bought for $0.20" (smec). The fix is the brand exclusion plus a cheap, dedicated brand Search campaign, and then judging PMax on non-brand performance and total-business results — never on the inflated blended number.
This is the single highest-leverage exclusion for a store with any brand recognition. It is also the one the keyword-focused guides bury or skip, because it is not a keyword at all.
How to find the keywords worth excluding
The honest method is unglamorous: open the PMax search-terms report, sort by spend, and read the queries that cost money and produced no orders. Those are your negatives. Cross-reference against Merchant Center so you are not excluding a query that maps to a healthy SKU.
The market makes this ongoing, not one-and-done. WordStream's 2025 benchmark data found roughly 87% of industries saw year-over-year cost-per-click increases, averaging about a 12.9% rise (Search Engine Land). As clicks get more expensive, a query you tolerated at a low CPC can quietly cross your break-even line. With retail-category search CPCs averaging around $4.14 in LocaliQ's 2026 data (LocaliQ), a single bad broad query can burn real money fast — which is why the audit is a monthly habit, not a launch task.
If you want the full framework for reading these reports and setting targets from margin rather than guesswork, start with our Google Ads economics guide. Operators who have outgrown doing this by hand often weigh bringing in a Google Shopping ads partner to run the hygiene on a cadence.
Where Victor fits
Reading the search-terms report, reconciling each query against true per-order profit, and catching brand cannibalization before it compounds is exactly the recurring work that gets dropped when you are running the store. PodVector AI's Victor is an AI employee that connects to your Shopify store, Meta Ads, and Google Ads as a full operator, computes true per-order profit, and surfaces where spend is losing money. Every write action — including any change to an exclusion — is approval-gated, so Victor proposes and you approve before anything executes. You can put Victor to work on your account here.
FAQs
Do negative keywords in Performance Max work across every channel?
No. Negative keywords in PMax only filter the Search and Shopping inventory (Search Engine Land). If your wasted spend is on Display placements or YouTube, you need placement or content exclusions instead — a negative keyword will not touch those surfaces.
How many negative keywords can I add to a Performance Max campaign?
Google raised the limit to 10,000 negative keywords per campaign in early 2025, up from the old cap of 100 (Search Engine Land). You can also apply shared negative keyword lists across multiple campaigns for consistency.
What is the difference between a negative keyword and a brand exclusion?
A negative keyword stops your ad from showing on searches containing a term you choose. A brand exclusion is a separate account-level setting that stops PMax from serving on your own branded queries, including variants and misspellings (AdNabu). The brand exclusion is the one that fixes cannibalization, and for most stores it matters more.
Will excluding keywords improve my ROAS?
It can, but be careful what you are measuring. Removing genuinely wasteful queries cleans the data your bidding learns from, which helps. But adding a brand exclusion will often lower your reported PMax ROAS precisely because it removes the cheap branded conversions that were inflating it — and that is the system working correctly, not getting worse.
How do I know which keywords are actually hurting me?
Pull the PMax search-terms report, sort by spend, and find queries that cost money and produced no orders. Compare each against your break-even ROAS, which is 1 ÷ your contribution margin. A query is only "wasteful" if it is costing more than the profit it returns — a number you can only judge against your own margin, not a benchmark.
Should I exclude keywords using broad match?
Rarely. Broad negatives block every search containing the word, so one careless broad negative can silently kill converting queries you wanted. Use phrase or exact negatives for anything ambiguous, and reserve broad negatives for words that are unambiguously junk for your store, like "free," "jobs," or "pdf."