Can I add negative keywords directly to a Performance Max campaign?

Google has expanded campaign-level negative keyword controls for Performance Max, but what is available and how large a list can be has changed more than once, and rollout has not always been simultaneous for every account. Open your own campaign to see what it exposes and check the current Google Ads documentation before planning a structure around it.

What is the difference between a negative keyword and a brand exclusion?

A negative keyword blocks matching against a string you type. A brand exclusion attaches a brand list, and the campaign stops serving against searches Google associates with that brand, which normally includes misspellings and variants you would never think to list. For your own brand or a competitor's, the exclusion list is the more complete tool.

Why can I not see individual search terms in Performance Max?

Performance Max reports query data as search categories on the Insights page rather than as a full term-by-term list. You get themes and their trend, not every string. That is why exclusion work here is thematic and why your first negatives in a category should be narrow rather than broad.

Will negative keywords stop my ads showing on irrelevant YouTube videos?

No. Keyword negatives apply to the query-driven side of the campaign. Where your ad renders on video, display or apps is governed by placement and content-suitability settings at account level, which is a separate control surface with its own limits.

How long should I wait before judging the effect of an exclusion?

Excluding a category changes the pool the campaign draws from, so the days immediately afterwards are re-learning rather than a verdict. Set the change, note the date, and compare a full period against the equivalent period before it instead of reading day-to-day movement.

Negative keywords for Performance Max

Negative keywords for Performance Max campaigns do exist, but they are a different instrument from the negatives you run on a Search campaign. In Search you watch a full query report and cut terms one by one. In Performance Max you are working with coarser reporting, a campaign that also serves Shopping, Display, YouTube, Discover, Gmail and Maps, and an exclusion system that only bites on the query-driven part of that inventory. Knowing where the lever reaches, and where it does not, is the difference between useful hygiene and a wasted afternoon.

Where negative keywords for Performance Max actually live

There is no single box. Depending on how your account is set up you are dealing with three separate mechanisms, and they do not behave the same way.

Account-level negative keyword lists. Applied across everything you run, these are the blunt instrument for terms you never want to pay for anywhere: job-seeker language, adult or violent modifiers, competitor names you have decided not to chase, anything legally awkward for your category.
Negatives set on the Performance Max campaign itself. Google has been expanding what advertisers can set directly at campaign level, and both the interface and the ceilings on list size have moved more than once. Open the campaign, look at what your own account currently exposes, and confirm the limits in the official Google Ads documentation rather than trusting a blog post from last year.
Brand exclusion lists. A separate object from negative keywords. Instead of matching literal strings you attach a brand list, and the campaign stops serving against searches Google associates with that brand, including misspellings it recognises on its own.

Treat these as three tools with three jobs. Dropping your own brand name into a negative keyword list is not the same as applying a brand exclusion, and the second usually catches variants the first one misses.

Why the Search playbook does not transfer

On a Search campaign the loop is tight: read the search terms report, spot a term that spent and did not convert, negate it, move on. That loop assumes you can see the term. Performance Max gives you a search categories view on the Insights page instead, which groups queries into themes and shows how each theme is trending. You are excluding a bucket, not a string.

That has a consequence people underestimate. When you negate a term on Search, you know precisely what you switched off. When you act on a category in Performance Max, you are inferring the underlying queries from the category label, so your negatives should start narrow and specific rather than broad. A broad negative sharing a word with a converting theme can silence far more than you intended, and because the campaign redistributes budget on its own, the damage surfaces as a general softening rather than an obvious drop in one place.

Sizing what a wasteful category costs you

Before you touch anything, size the problem. Take a single 30-day window and hold it for the whole exercise. Say the campaign spent 40,000 TL and returned 100,000 TL, so ROAS is 100,000 ÷ 40,000 = 2.5x. In the search categories report, one informational theme accounts for 6,000 TL of that spend and 3,000 TL of the revenue. That theme is running at 3,000 ÷ 6,000 = 0.5x, and it is 6,000 ÷ 40,000 = 15% of campaign spend.

Strip it out on paper: 40,000 - 6,000 = 34,000 TL of spend against 100,000 - 3,000 = 97,000 TL of revenue, which is 97,000 ÷ 34,000 = about 2.85x. That is roughly 14% above the 2.5x you started with, from a single exclusion.

Now the caveat, because this is where the exercise usually gets oversold. Removing a category does not hand you 6,000 TL back. Performance Max pushes most of that budget into whatever else it can find, and the return on the redirected spend is unknown until you run it. So 2.85x is a ceiling for the effect, not a forecast, and the gap has to be wide enough to survive that discount before the change is worth making. If you want to run the same before-and-after on your own figures, the ROAS calculator handles the division.

Brand traffic and the number it moves

The most common reason advertisers reach for exclusions here is brand. The campaign will happily absorb searches for your own name, those searches convert cheaply because the person already decided, and reported ROAS rises for reasons that have little to do with what the automation found. Apply a brand exclusion and that traffic leaves, so the reported figure falls. The fall is not a regression. It is the number becoming honest about the campaign's prospecting work.

Whether brand searches should be bought at all, and by which campaign, is a separate strategic argument with real cases on both sides. The mechanical point here is narrower: while brand sits inside Performance Max you cannot read the campaign as a prospecting channel and you cannot compare it fairly against anything else. Decide where brand belongs, apply the exclusion, and record the date so the break in your reporting line has an explanation attached to it.

What you genuinely cannot exclude

Non-search inventory. A negative keyword does nothing about a YouTube channel or a mobile app where your ad rendered. That is a placement and content-suitability problem, handled in account-level settings, not in a keyword list.
Matching driven by your own data. Performance Max reads your product feed, your landing pages and your asset text to decide what to show. If a product title or a page is pulling the wrong intent, negatives are a bandage over a source problem. Repairing the input is the real fix, which is why feed hygiene and exclusion work belong in the same review session; the piece on product feed ads covers that layer.
Anything you already paid for. Negatives are not retroactive. They shape future auctions only, so the spend that triggered the change is gone either way.
Full term-level transparency. Whatever query detail your account exposes today, plan for less of it than a Search campaign gives you and build the review around themes rather than strings.

A review rhythm that respects the learning phase

Exclusions are edits to a system that is continuously re-optimising, so cadence matters as much as content.

Review search categories on a fixed schedule rather than after every bad day. A weekly glance and a monthly decision session works for most accounts.
Batch the changes. Twenty negatives added across three sessions in one week makes it impossible to attribute whatever happens next to any of them.
Write down each addition with a date and a one-line reason. Six months on nobody remembers why a term is blocked, and unexplained lists get deleted wholesale by the next person in the account.
Keep one shared account-level list for terms that apply everywhere and reserve campaign-level negatives for campaign-specific noise. Duplicating the same fifty terms across eight campaigns guarantees they drift apart.
After a meaningful exclusion, give the campaign time before judging it. You changed the pool it draws from, so the first few days are re-learning, not a result.

Two habits keep this from becoming busywork. First, check whether a category you want to block is genuinely unprofitable or merely upper-funnel, because a theme that never converts on a last-click view may still be doing work. Second, if you are mapping which targeting levers still exist inside heavily automated campaigns, the AI targeting overview covers that side of the platform, and the wider piece on AI for Performance Max places exclusions alongside everything else the campaign type decides for itself.

Search term hygiene in Performance Max is not the surgical work it is on Search, and pretending otherwise leads straight to disappointment. Done properly it is a slower, coarser discipline: size the waste, exclude the clearest cases, repair the inputs that caused the mismatch, and keep your expectations aligned with what the exclusion system can actually reach.