Performance Max gives you two controls that both claim to shape which queries you pay for, and they can quietly cancel each other out. Search themes tell the campaign which searches to treat as relevant; negative keywords tell it which searches to refuse. When a theme and a negative point at the same query, the negative wins — every time, with no warning in the interface. That single precedence rule is the whole of this post, and getting it wrong is how an account manager ends up with a carefully chosen theme that spends almost nothing and no obvious reason why.
This is not the same problem as brand exclusions, channel coverage, or the general mechanics of adding negatives to Performance Max — those are covered elsewhere. The gap here is narrower and sneakier: two of your own inputs disagree, one silently overrules the other, and the UI never tells you the collision happened. If you have set search themes to steer a campaign and the traffic has not followed, the first thing to check is not the theme — it is whether one of your negatives is standing in the doorway. The rest of this post explains why the negative wins, how to find the conflicts by hand, and how to decide which of the two contradicting signals is actually the mistake.
What each control actually does
A search theme and a negative keyword are different kinds of instruction, and the difference is the reason one beats the other. A search theme is a signal: you hand Performance Max a phrase — say, “commercial refrigeration repair” — and the campaign treats it as evidence about the kind of query you want to be considered relevant, then matches loosely around it in the keywordless way it handles most of its traffic. It is an input to a model, not a rule, and it is deliberately fuzzy. Google frames search themes as information that supplements automated targeting rather than a hard control, which is exactly why they cannot force anything through.
A negative keyword is the opposite in kind: it is a rule, a hard exclusion that removes a query from eligibility before any model gets a say. It uses strict match-type logic — broad, phrase, or exact — and a broad-match negative blocks any query containing those words in any order. Because one is a soft signal and the other is a hard gate, there is no symmetry to the contest: a signal cannot reinstate traffic that a gate has closed. Understanding that the two are not peers — that a theme is a suggestion and a negative is a veto — is the mental model that makes the precedence rule obvious rather than surprising. For the full mechanics of how PMax negatives are scoped and applied, the Performance Max negative keywords walkthrough is the companion to this one.
Why the negative always wins
When a theme and a negative target the same query, eligibility is decided in the order that matters: exclusions are applied first. The negative keyword removes the query from the auction entirely, and the search theme never gets the chance to argue for it, because there is no longer a query there to be relevant to. This is not a tiebreak or a weighting where a strong enough theme might overcome a weak negative — it is a gate that closes before the signal is read. A theme pointed at an excluded query is simply pointed at nothing.
That is why the failure is silent. You add the theme, the campaign accepts it, the interface shows it set and active, and nothing flags that a negative three lists away is cancelling it. The theme looks like it is working because the UI has no concept of “this input is being overruled by that one.” The only place the contradiction surfaces is in the outcome: the queries the theme described never show up in the search terms report, and the spend you expected to follow the theme never appears. The report is privacy-thresholded and will never reconcile to the last query, but a theme that produces a near-total absence of related terms is a strong signal that something upstream is blocking them — and a negative is the usual culprit.
Finding the conflicts by hand
There is no screen in Google Ads that lists theme-versus-negative conflicts, so you assemble the cross-check yourself, and it is worth doing as a deliberate pass rather than hoping to notice it. Start by listing every search theme you have set, across every asset group. Then list every negative that can act on the campaign: the campaign-level negatives, and — this is the part people forget — every account-level negative list applied to it. A conflict you cannot see is almost always a negative living on a shared list you set up months ago and no longer think about, quietly excluding a theme you added last week.
For each theme, take its head terms and ask whether any negative would block a query that theme describes, matching the negative the way Google will: a broad-match negative for “cheap” kills every query containing the word “cheap,” so a theme built around budget-conscious buyers dies against it instantly. This is the same evidence-first discipline you use when mining the search terms report for negatives, run in reverse: instead of scanning real queries for words to block, you are scanning your block list for words that strangle a theme. If a theme is set and its queries are absent from the search terms report, treat a negative as the prime suspect before you touch anything else about the theme.
Deciding which signal is the mistake
A conflict is not automatically an error to be deleted — it is a question you have to answer: which of these two did I actually mean? Sometimes the negative is right and the theme is the accident, because you added a theme that overlaps a kind of traffic you long ago, correctly, decided not to pay for. In that case the fix is to drop the theme and let the negative stand; the account is behaving exactly as your earlier self intended. Adding themes without checking them against existing negatives is how you end up re-opening doors you deliberately closed.
The other case is that the theme is right and the negative is now too blunt. Here you narrow the negative rather than deleting it outright: tighten a broad-match negative to phrase or exact so it stops catching the theme's queries, or scope it to a different campaign where the exclusion still belongs. What you must not do is the thing that feels like compromise — leave both in place and assume they will sort themselves out. They will not; the negative will keep winning, and the theme will keep spending nothing. Resolve every conflict in one direction or the other. This precedence check belongs in the same routine as the broader 2026 Performance Max steering controls audit and the related brand exclusions versus negatives decision — all three are versions of the same habit: when two of your own controls can contradict each other, find the collision before the spend report finds it for you.
Related reading
For the full mechanics of adding and scoping negatives in Performance Max, see Performance Max negative keywords. For the complete 2026 control surface search themes sit within, see the 2026 Performance Max steering controls, and for the adjacent question of brand inclusions versus exclusions, see brand exclusions versus negative keywords.