Performance Max Search Terms: What You Can See and Control

Pmax hides the raw query report, but you still have real levers. Here is exactly what is exposed, what is not, and how to steer spend anyway.

October 8, 2026 · QueryCut

Performance Max does not give you a traditional search terms report. What you get instead is a Search categories view that buckets queries into themes, shows you clicks and impressions, but hides conversions and spend per category. You cannot add negative keywords through the normal UI either, but you can control irrelevant traffic through account-level negatives (request access from your Google rep), brand exclusion lists, and asset-level signals. The practical answer: stop trying to prune individual terms the way you would in Search, and instead manage Pmax at the theme and audience-signal level.

What the Search categories report actually shows, and what it hides

Go to the Pmax campaign, then Insights and reports > Search terms (some accounts still label it Search categories under the Insights tab). You will see named theme buckets with clicks and impressions. That is roughly the whole picture Google gives you. There is no cost column, no conversion column, and no per-category CPA. You are looking at volume, not value.

This matters because the first instinct of a Search manager is to find the expensive junk and kill it. In Pmax you cannot see which category is expensive. A category showing 4,000 impressions and 60 clicks might have driven your best conversions or burned budget on a competitor's brand name, and the report will not tell you which. You are navigating with half the dashboard covered.

The second thing to understand: these categories are not literal search terms. They are Google's grouping of many related queries under a human-readable label. So when you see a theme you dislike, you are not looking at one query you can negate. You are looking at a cluster, and the lever that affects a cluster is different from the lever that affects a single term.

The mechanism: why most terms do not matter and themes do

Pmax runs a single budget across Search, Shopping, Display, YouTube, Gmail, and Discover, and the bidding system allocates in real time toward whatever predicts conversions. The search side is not a keyword auction you control. It is closer to Dynamic Search Ads: Google reads your landing pages, assets, and audience signals, then decides which queries to chase.

Because of this, the signal that actually moves spend is thematic, not per-query. A single odd search term carries almost no budget and self-corrects as the model learns. What costs you real money is a whole theme the model has latched onto incorrectly, usually because your feed or landing content looks relevant to it. If you sell commercial espresso machines and a theme like home coffee makers shows heavy volume, that is the model misreading your assets, and it will keep spending there until you change the input.

This is the core reframe. In Search you hunt individual converters and wasters. In Pmax you ask one question: is there a theme here that does not belong, and if so, what in my inputs is telling Google it does? The fix lives upstream in the feed, the final URLs, and the exclusions, not in a negative keyword box.

The levers you actually have

You have fewer controls than Search, but the ones you have are blunt and effective when used on themes rather than single queries.

Account-level negative keywords are the big one. These are not available in the standard interface for Pmax; you request them through your Google Ads rep or support, and they apply across the whole account. Use them for whole categories of intent you never want (job seekers, free, DIY, competitor brands), not for one-off terms. Brand exclusions are a separate, self-serve control under the campaign settings that let you stop Pmax from spending on your own brand or specified other brands. Turn brand exclusions on by default unless you have a deliberate brand-defense reason, because Pmax will happily eat cheap brand clicks and inflate your reported ROAS without adding incremental sales.

Then there are the indirect steering tools: audience signals (which nudge, not restrict, who the model targets), the content and specificity of your text and image assets, and your final URL expansion setting. Turning off final URL expansion, or restricting it to a URL set, is one of the most underused controls. It stops Pmax from sending traffic to pages you did not nominate, which is often where off-theme queries come from.

• Account-level negatives: request from your Google rep, use for whole intent categories.
• Brand exclusions: self-serve in settings, turn on unless you are deliberately defending brand.
• Final URL expansion: switch off or restrict to stop off-theme page targeting.
• Audience signals: a nudge toward likely converters, not a hard filter.
• Asset relevance: vague assets invite broad, off-theme matching.

Decision rules for acting on a theme

Since you cannot see cost or conversions per category, you have to triangulate. First, check whether a questionable theme is also a high-volume theme. A theme with a handful of impressions is noise and not worth a negative; the account-level negative list is a shared resource and over-stuffing it creates maintenance debt and risks blocking valid future queries. Reserve negatives for themes that are both clearly off-intent and clearly high-volume.

For timing, do not touch a new Pmax campaign's search themes for the first two to three weeks or until it has cleared the learning phase. Early on, the model explores broadly on purpose, and themes that look wrong in week one often vanish by week three as it finds your converters. Cutting too early can strangle exploration the model needed to find cheap conversions. If an off-intent theme is still high-volume after three weeks and after your overall CPA has stabilized, that is when you add the negative.

A useful heuristic for the whole campaign: if your blended Pmax CPA or ROAS is at target, resist the urge to prune themes you merely dislike. You are optimizing against a signal you cannot fully see, and a theme that looks irrelevant to you may be quietly cheap and converting. Only intervene when the theme is both strategically wrong (a product line you do not sell, a region you do not serve) or when total performance is off target and you are hunting for the cause.

The failure mode: pruning a black box you cannot measure

The classic Pmax mistake is treating the Search categories report like a Search terms report and reactively negating every theme that looks ugly. The trap: you cannot see whether that theme was converting, so you may be removing your cheapest conversions and then watching CPA climb with no idea why. Because the removal happens at account level and the loss shows up as a blended number, the cause and effect are nearly impossible to trace after the fact.

The second trap is brand exclusions in reverse. Teams leave brand traffic in Pmax because the reported ROAS looks great, not realizing the campaign is harvesting conversions that would have happened anyway through free organic or dedicated brand Search. The honest limit here: Google does not expose incrementality in Pmax, so you cannot prove from the interface whether a brand-heavy Pmax is adding sales or just claiming credit. The defensible default is to exclude brand from Pmax and run brand on a tightly controlled Search campaign where you can see the terms.

Be honest about what you cannot know. You cannot see per-theme spend or conversions. You cannot negate a single query. You cannot fully separate Pmax's incremental contribution from what channel mixing is crediting to it. Tools like QueryCut can surface and cluster the terms that are exposed and suggest negatives faster than manual review, but no tool can report the cost and conversion data Google chooses not to share. Plan your controls around that gap rather than pretending it is not there.

Questions people also ask