How negative keywords actually work in Performance Max

Performance Max is the campaign type that gives operators the least control over negatives, and the one where bad negatives have the most opaque consequences. Almost everything in this section comes with a caveat.

This piece covers what the controls actually do, where the gaps are, and the working approach for keeping a Performance Max campaign from drifting into junk traffic without getting blocked from running at all.

The four levers Performance Max gives you

Performance Max does not have a true "negative keywords" screen the way Search campaigns do. What it has, as of early 2026, are four levers that together approximate negative-keyword behavior.

Account-level negative keyword lists. Added in late 2024. Up to a thousand keywords in a single list, applied to every campaign in the account, including Performance Max. This is the closest thing PMax has to a normal negative list.

Brand exclusions. A specific list of brand names that PMax should not show ads against. This is for queries that contain another company's brand. Useful, but limited to brand-style mismatches; it does not handle generic junk like "free" or "salary."

Negative campaign-level keywords (limited). As of late 2024, you can request negative keyword support at the campaign level for Performance Max from your Google account rep. The feature is slowly rolling out and not available on every account. If your account has it, you get a much closer approximation of normal negative behavior.

Search themes. Positive signals about what queries you want to bid on. The absence of a theme is not the same as a negative; PMax can still match queries that aren't on your search themes list.

Why this matters: the search categories report

The closest thing PMax gives you to a search terms report is the search categories report. It does not show individual queries. Instead, it shows category-level groupings: "ac repair," "plumbing services," "insurance quotes." The category labels are Google's own categorization and you cannot drill into them to see the individual queries that fed each category.

This is where the gap with Search becomes painful. In a Search campaign, you can see "free ac repair" as a row and add "free" as a negative. In Performance Max, you see the category "ac repair" as a row and you do not know whether the queries inside are mostly "free ac repair" (block-worthy) or "emergency ac repair" (keep-worthy). The category report tells you the spend went to "ac repair." It does not tell you which "ac repair."

The workaround is to assume that the same junk patterns that show up in your Search campaigns are also showing up in your PMax campaigns, and to apply the same universal blocks at the account level. This is imperfect; it catches the obvious junk while leaving the campaign-specific junk untouched.

The working approach for PMax negatives

Given the gaps, the practical approach is layered. None of the layers is sufficient on its own; together they get most of the way to where Search negatives get on their own.

Layer one: account-level negative list. Apply the universal junk blocks (free, cheap, diy, salary, jobs, schools, training) at the account level. This catches the most common waste across both Search and PMax in one place.

Layer two: brand exclusions. If you do not want to show against competitor brand searches, populate the brand exclusion list. This is the only way to handle competitor-brand queries in PMax.

Layer three: campaign-level negatives (if you have access). If your account has the limited rollout of campaign-level negatives in PMax, use them for blocks that are PMax-specific, like geo blocks or service-category blocks that do not apply to your other campaigns.

Layer four: search themes. Use positive search themes to bias the campaign toward the queries you actually want. This is not a negative; it is a hint. PMax can still match outside the themes, but well-chosen themes shift the distribution toward better queries.

What to do when search categories surprise you

A common pattern in PMax: the search categories report shows substantial spend going to a category that does not match any of your search themes or your business. Examples we have seen include "personal loans" appearing under a home services campaign, or "insurance quotes" appearing under a home goods campaign.

These surprise categories almost always indicate that PMax has expanded its matching beyond your apparent intent. The fix path is awkward. You cannot directly block the category; you can only:

• Refine your asset group's signals (audience, demographics, themes) to push PMax away from it.
• Add specific keywords from the surprise category to your account-level negative list, if you can guess what queries are inside the category.
• Pause the asset group and rebuild with tighter signals.

None of these are clean. The cleanest move when PMax is wandering is sometimes to split out the most important parts of the campaign into regular Search campaigns, where you have full negative-keyword control.

When Performance Max is and is not a good fit

PMax is a good fit when:

• The product or service has clear visual signals that benefit from showing across multiple inventory types (display, video, shopping).
• The audience is broad enough that automated bidding has signal to learn from.
• The account has access to first-party audience signals (customer lists, conversion data) that PMax can lean on.
• The operator is willing to trade granular control for automation and broader reach.

PMax is a poor fit when:

• The audience is narrow and the obvious junk patterns are well-known. A regular Search campaign with strong negatives outperforms PMax in this case.
• The conversion volume is too low to feed PMax's bidding. Below about thirty conversions a month, PMax starts to bid speculatively and waste budget.
• Brand safety matters and you need to control specific match exclusions. PMax's brand exclusion list is fine but coarser than Search-level controls.
• The operator's time budget for monitoring is small. PMax surprises happen; if nobody is watching the search categories report, the surprises become silent leakage.

For most small service businesses, the right move is a mix: Search campaigns for the high-intent commercial queries where every negative matters, and a Performance Max campaign for broad-reach prospecting where automation can find conversions at scale. The Search campaigns get the careful negative-list treatment; PMax gets the layered universal-block approach.

What is changing, and what to expect next

Google has been steadily improving PMax's negative-handling since the campaign type launched. The 2024 additions of account-level negative lists and the limited campaign-level negative rollout were the biggest improvements yet. The trajectory suggests more granular controls are coming.

Things to watch for in 2026 and beyond:

• Negative search themes. A positive-only search themes feature is asymmetric. Adding negative search themes would let operators block whole categories without enumerating individual queries.
• Search-terms-level visibility in the categories report. Even partial query-level data inside categories would close most of the diagnostic gap.
• Wider rollout of campaign-level negatives for PMax. Currently limited; making it standard would put PMax much closer to feature parity with Search.

In the meantime, the layered approach is the working version. Universal blocks at the account level. Brand exclusions per campaign. Search themes as positive bias. Vigilant monitoring of the search categories report.

PMax is improving, but it is not yet a campaign type that runs well unattended. The ones that do run well have someone paying attention to the gaps.