If you have a Performance Max campaign live for more than 30 days and its Negative keywords tab is still empty, you are leaving the account-level list to do all the work — and the account-level list is capped at 1,000 entries shared across every campaign. The campaign-level slot is the place where most of your day-to-day waste blocking now belongs.

Why this matters

Most accounts audited in the first half of 2026 still rely only on the account-level negative list because the campaign-level slot rolled out quietly and was not switched on retroactively. The default state for every PMax campaign created before January 2026 is an empty campaign-level tab.

That default has two business consequences:

You burn the account-level cap on per-campaign noise. The 1,000-entry account ceiling fills up fast when you push every PMax-specific competitor or long-tail waste term into it [1]. Once the cap is full, you start trading older negatives for newer ones, and protection drifts.
You cannot tailor exclusions per campaign. A brand variant that should block your generic PMax must also block your branded PMax if it lives at account level — there is no surgical option. The campaign-level slot exists exactly to break that trade-off.

The conceptual mechanics (10,000-per-campaign cap, Search/Shopping inventory only, PMax Experiments interplay) live in the sibling glossary [[pmax-negative-keywords-campaign-level]]. This finding is purely about whether you have actually used the slot on each PMax campaign that has earned the protection.

How to verify

Open the PMax campaign in Google Ads.
Navigate Campaign → Keywords → Negative keywords.
Check the count at the campaign level (not the account-level or shared-list tab).
Flag the campaign if it has been live > 30 days, spent > a meaningful threshold for your account, and the campaign-level count is 0 or trivially small (e.g. < 20 entries).
Cross-check the Search terms view (Insights → Search terms) for the last 90 days — anything sorted by spend with > $50 lifetime on irrelevant or competitor queries is a coverage gap that the empty slot is failing to catch.

If the campaign-level count is healthy but the same variants are duplicated in both account-level and campaign-level lists, that is a separate drift-risk finding — keep ownership clean (account-level = global noise, campaign-level = per-campaign tailoring).

How to fix

Pull the Search terms report. PMax campaign → Insights → Search terms → last 90 days, sorted by spend.
Bucket the spend. Industry-standard buckets for a starter list: brand safety (profanity, adult, violence), competitor brands and variations, price-intent for premium brands ("cheap", "free", "discount", "clearance"), job-seeker terms (if not recruiting), DIY/tutorial terms (if selling finished products), and informational intent ("what is", "definition", "guide") [3].
Seed the campaign-level slot. Campaign → Keywords → Negative keywords → +. A well-structured initial list across the buckets above typically runs 400-800 keywords and gives immediate protection against the obvious-waste long tail [3].
Apply the 50-impression threshold for ongoing additions. Do not negative a search term with fewer than 50 impressions in the last 90 days unless the irrelevance is obvious on its face — below that threshold, the signal is too noisy to act on [3].
Choose match types deliberately. Use phrase match for competitor names, exact match for surgical blocks (specific brand misspellings you want stopped cold), broad match only when the term is unambiguous ("free", "cheap") and you accept the broader collateral block. Account-level and campaign-level negatives both honour broad/phrase/exact, with no close-variant matching [1].
Decide where each new negative lives. Global noise (industry-wide profanity, account-wide brand-safety list) → account-level. Per-campaign noise (this campaign's specific competitor list, this campaign's irrelevant long-tail) → campaign-level. Do not double-list.

Common gaps

These are the patterns that most frequently show up as findings against this rule:

Brand-only PMax with an empty campaign-level slot. Even brand-only PMax campaigns benefit from a small, targeted campaign-level list (misspellings of competitor brands that triggered, "free trial" of competitor product, etc.). Empty here is almost always neglect, not by design.
Generic PMax with brand negatives stuffed into the account-level list. This blocks your branded PMax from catching the same query. Move the brand variants to the generic PMax's campaign-level slot and keep the branded PMax clean.
Shared library exclusion list attached, campaign-level slot empty, and a separate "global" account-level list also in play. Three overlapping surfaces, none of them clean. Decide on the policy: shared list for cross-campaign noise, campaign-level for per-campaign tailoring, account-level only for true account-wide blocks.
Campaign-level slot has < 20 entries despite 6+ months of spend. That is a token gesture, not a seeded list. Schedule a 60-90 minute seeding session against the 90-day search terms report and ship the 400-800 starter list in one pass.

Edge cases

Brand-new PMax campaign (< 30 days live). Do not seed aggressively from day one — the search terms report has not accumulated enough volume to tell you what is actually wasting spend. Seed a small brand-safety baseline (20-50 obvious terms) at launch, then re-audit at the 30-day mark.
PMax serving very low spend (< $500/month). The 400-800 starter list is overkill — a 30-50 entry brand-safety + obvious-competitor list is sufficient until the spend justifies more time investment.
PMax Experiment running on the same campaign. If you are A/B testing a campaign-mix change via Experiments, hold negative-list changes flat for the duration of the experiment window so you do not contaminate the read. See [[pmax-negative-keywords-campaign-level]] for the Experiments interaction.
You hit the 1,000 account-level cap. Do not solve that by adding more account-level entries — audit the existing list for terms that should be moved down to campaign-level, then prune.

[[pmax-negative-keywords-campaign-level]] — what campaign-level negatives are, the 10K cap, rollout timeline, Search/Shopping inventory scope, and PMax Experiments interaction.
Negative keywords — match types, account vs campaign scope, and the shared library workflow.
Fix: Search terms report shows irrelevant queries — the upstream signal that should drive seeding.
Fix: Negative keywords missing — sibling finding for Search campaigns with no negatives attached.
Performance Max brand cannibalization — why brand variants belong in the generic-PMax campaign-level list, not at account level.

Sources

Google Ads Help — About account-level negative keywords — confirms the 1,000-per-account cap, scope across Search/PMax/Shopping/App/Smart, broad/phrase/exact match support, and lack of close-variant matching.
Google Ads Help — Negative keywords in Performance Max campaigns — canonical reference for adding campaign-level negatives in PMax, Search/Shopping-inventory-only scope, and how account-level negatives apply across PMax.
Groas — Performance Max Negative Keywords 2025: Complete Guide to the 10,000 Keyword Limit — rollout timeline (December 2024 initial, March 2025 cap raise to 10,000, August 2025 shared-list support), 400-800 starter-list recommendation across brand-safety/competitor/price/informational buckets, and the 50-impression threshold for ongoing additions.