PMax campaign-level negative keywords

Campaign-level negative keywords landed in Performance Max in Q1 2026 — the first time you can scope a negative to one PMax campaign instead of the whole account. Combined with PMax Experiments, the brand-cannibalization playbook changes.

What it means and benchmark

Before Q1 2026, the only way to keep a Performance Max campaign off a specific query was an account-level negative keyword (blunt — hits every campaign) or a brand exclusion list (only matches what Google identifies as "branded"). Campaign-scoped negatives close that gap: you can now exclude long-tail wastage from one PMax campaign without disturbing the rest of the account, and you can run a brand-only PMax alongside a generic PMax without account-level negatives nuking both.

Benchmark for a healthy PMax setup in 2026:

• ≤ 1% of PMax Search-slice spend going to queries you would call irrelevant or competitor-only.
• ≥ 1 campaign-level negative list attached to every PMax campaign that has been live > 30 days.
• Brand-only PMax and generic PMax separated when total branded-query monthly spend > $5k — generic side carries the brand variants as campaign-level negatives.

Top causes of brand/generic bleed in PMax

1. No campaign-level negatives attached — most accounts audited in May 2026 still rely only on the account-level list because the campaign-level UI rolled out silently. The default state for every PMax campaign created before January 2026 is empty.
2. Brand exclusion list misses variants — Google's brand-exclusion matching is opaque; misspellings, acronyms, and competitor + brand combos slip through. Campaign-level negatives let you belt-and-brace exactly the variants the search terms report shows.
3. Single PMax campaign mixing brand and generic intent — when one PMax carries both branded asset groups and generic asset groups, account-level brand negatives are unusable (they would starve the brand asset group). Splitting into brand-only + generic-only PMax campaigns is now the recommended pattern.
4. No PMax Experiment to measure the lift — teams add negatives, then attribute every subsequent ROAS swing to the change without a control. The Q1 2026 Experiments feature exists exactly to isolate this.

Diagnostic checklist

Fix paths

1. Pull the PMax search terms report. Open the PMax campaign → Insights → Search terms for the last 90 days. Sort by spend. Bucket queries into brand / competitor / irrelevant / wanted. Anything > $50 lifetime spend in the first three buckets is a campaign-level negative candidate.
2. Add campaign-level negatives. Campaign → Keywords → Negative keywords → click +. Add up to 10,000 — but start with the 30-100 highest-spend wastage queries. Use phrase match for competitor names, exact match for specific brand misspellings you want surgically blocked, broad only when the term is unambiguous (e.g. "cheap", "free").
3. Split brand-only from generic PMax when budgets justify it. If branded-query spend exceeds $5k/month: create a generic PMax with brand variants as campaign-level negatives, and a brand-only PMax with brand variants in the asset group themes. This lets you bid differently per intent and prevents the generic PMax from poaching cheap brand clicks.
4. Run a PMax Experiment to measure lift. Campaigns → Experiments → New experiment → Performance Max. Use the "campaign mix" experiment type to test current setup (control) vs new negative configuration (treatment) on a 50/50 split for 14-21 days. Confirm ROAS or CPA delta is statistically meaningful before rolling out.
5. Re-baseline and document. Screenshot Search-slice spend, brand Search impression share, and account MER on the change date. Re-check at day 14 and day 30 of the Experiment.

When to escalate

Escalate the negative-keyword + Experiment workflow to a structural account redesign when any of these conditions hold:

1. Brand and generic PMax cannot be cleanly separated because the product catalog itself doesn't carry brand attribution (resellers, marketplaces). Negatives alone won't fix the underlying intent overlap — restructure the feed or move brand traffic to Search.
2. Campaign-level negatives plus brand exclusions both attached, yet Search-slice spend on brand variants remains > 2% after a 21-day Experiment. The matching engine is overriding your exclusions — open a Google rep escalation, do not keep stacking negatives.
3. PMax Experiment shows no statistically significant ROAS delta after 21 days at ≥ 50% spend share. Either the negatives weren't biting (Search slice was already clean) or PMax is recovering revenue elsewhere; revisit channel-level performance via the PMax channel performance timeline before adding more negatives.

Methodology note — how Whitead diagnoses PMax negative keyword hygiene. The audit looks at four signals together, not one. (1) Presence: is a campaign-level negative keyword list attached to each PMax campaign live > 30 days, or only the account-level list? Empty campaign-level slot on a mature PMax = flag. (2) Coverage: cross-reference the last 90 days of PMax search terms against the campaign-level negatives — any wastage query with > $50 lifetime spend that is not blocked counts as a coverage gap. (3) Separation: when monthly branded-query spend exceeds $5k, check whether brand-only and generic PMax are split. Single mixed campaign = structural flag, not just a negatives flag. (4) Measurement discipline: was a PMax Experiment used to validate the negative-keyword change, or was it shipped blind? Blind ships make rollback decisions impossible to ground. Fix priority is sequential — coverage gaps ship same-day (20 minutes per campaign), separation is a 2-week project, and any change ≥ 30 negatives should be paired with a PMax Experiment so the team has a defensible answer when ROAS moves the following month.

Related concepts

• Performance Max — the campaign type and its 2025-26 transparency surface.
• Performance Max brand cannibalization — the canonical brand-bleed playbook; campaign-level negatives are the new tool in that fix kit.
• Negative Keywords — match types, account vs campaign scope, and the shared library workflow.
• Brand search cannibalization — the broader cross-channel problem of which PMax brand-bleed is one instance.
• PMax channel performance timeline — diagnose where PMax spend is actually going before adding negatives.
• PMax low ROAS diagnostic — broader troubleshooting flow when negatives alone don't fix the ROAS gap.
• Fix negative keywords missing — the rule-tied finding for Search campaigns with no negatives attached.

Sources

• Google Ads Help — Negative keywords in Performance Max campaigns — canonical reference for campaign-level negatives in PMax, Search/Shopping-only inventory scope, and how to add them via Keywords → Negative keywords.
• Google Ads Help — Unlock more visibility and control in Performance Max — 2026 PMax transparency and control rollout including campaign-level controls.
• Search Engine Land — Google Ads expands PMax negative keyword limits (2025-03-11) — confirms the cap raised from 100 to 10,000 per campaign.
• PPC Land — Google raises negative keywords limit to 10,000 for PMax campaigns (2025-03-11) — corroborating coverage with Ginny Marvin quote on cap rationale.
• Practical Ecommerce — 3 Performance Max updates for 2026 (2026-02-06) — covers PMax Experiments rollout including campaign-mix experiment types for testing changes against current setup.
• Search Engine Land — Performance Max guide — Asset Experiments workflow within PMax via Campaigns → Experiments.