Every account I inherit has the same artifact: one giant negative keyword list, built over years, with no logic to it. Someone added "free" in 2021. Someone else added a competitor name in 2022. A junior hire added forty job-title terms after a bad week of leads. Nobody ever removes anything, because nobody remembers why it's there. By year three, that list is actively fighting the account it was built to protect.
This matters more now than it did five years ago. Broad match and Smart Bidding are designed to find query variations you'd never think to bid on manually. That's the whole value proposition. A sloppy negative list doesn't just block junk — it silently excludes legitimate queries that happen to share a word with something you blacklisted in a different context two years ago. You end up paying for the reach of broad match while structurally preventing it from working.
Why the single-list approach fails as you grow
A flat negative list has no concept of context. "Jobs" might be a correct exclusion for your lead-gen campaign and a false exclusion for your careers-page remarketing campaign. "Cheap" might be irrelevant noise for a premium SaaS brand but exactly the query a discount-tier product line should be bidding on. When every negative sits in one undifferentiated pile applied everywhere, you're forcing every campaign to inherit every other campaign's assumptions.
The failure compounds with account growth. New campaigns get launched into an account that already has years of accumulated negatives silently suppressing their reach from day one — before you've even run a single search term report on them. Nobody audits this because negative lists feel like a "set it and forget it" hygiene task, not a structural decision.
One shared negative list, built ad-hoc over years, applied uniformly across every campaign regardless of product line, match type, or funnel stage — blocking queries nobody remembers the reason for.
A tiered structure — account-level exclusions for true universals, campaign-level lists for context-specific exclusions, and a separate, lighter-touch list for Performance Max — each reviewed on its own cadence.
The three-tier structure that actually holds up
I run negatives on three distinct levels, and I keep them that way deliberately rather than collapsing them for convenience.
- Account-level shared list: reserved for terms that are universally irrelevant regardless of campaign — your actual competitor brand terms if you're not running a conquesting campaign, recruitment terms, "free," "DIY," "refund," terms tied to a product line you've fully discontinued. Short list, high confidence, rarely touched.
- Campaign-level lists: this is where most of the judgment lives. A campaign selling enterprise software shouldn't inherit the same exclusions as a campaign selling the entry-tier plan. A prospecting campaign and a remarketing campaign should not share exclusions either — intent is different, so the same query can be junk in one and a conversion in the other.
- PMax exclusions: handled separately and more conservatively, since Performance Max doesn't give you search term-level visibility the way standard Search does. I only add account-level negatives here (PMax only respects the account-level list, not campaign-level lists), and I keep that list deliberately short — adding too aggressively here removes Smart Bidding's ability to explore inventory you can't see it exploring in the first place.
If you're not already pulling real search term data to populate campaign-level lists on a schedule, that's a separate discipline worth fixing first — I cover the process in The Search Term Mining Framework That Catches Wasted Spend Before It Compounds. The architecture here only works if the inputs feeding it are current, not three years stale.
A negative keyword list isn't a one-time cleanup — it's infrastructure, and infrastructure needs a maintenance schedule or it starts working against you.
The audit that catches silent damage
Twice a year, I pull the full account-level shared list and check every term against current campaign structure. The questions I ask: Is this still a real competitor, or did they rebrand? Is this product line still discontinued, or did it relaunch? Is this term blocking a query pattern that a newer campaign actually wants? I've found account-level negatives from a campaign that no longer exists still quietly blocking impression share on a campaign launched eighteen months later — nobody connected the two because they were never in the same room when both decisions were made.
The other check I run is cross-referencing negatives against Smart Bidding performance. If a campaign on Target CPA or Target ROAS is underdelivering on impression share with budget left unspent, negatives are one of the first places I look — alongside the usual Lost IS: Budget vs. Rank diagnosis. A list that's too aggressive produces the same symptom as a bid strategy that's too conservative, and it's easy to misdiagnose one as the other.
- ✓︎Account-level list contains only true universals — not campaign-specific decisions that drifted upward
- ✓︎Each campaign has its own list reflecting its actual intent and funnel stage, not an inherited copy-paste
- ✓︎PMax exclusions are limited to the account-level list and kept short, since PMax can't see campaign-level negatives
- ✓︎Every negative has a traceable reason tied to real search term data, not institutional memory
- ✓︎List reviewed on a fixed schedule, not only when someone notices a problem
- Does Performance Max respect campaign-level negative lists?
- No. PMax only applies account-level negative keyword lists. Campaign-level exclusions you've built for Search campaigns have no effect on PMax, which is why the account-level list needs to be curated with PMax's limited visibility in mind.
- How often should I review negative keyword lists?
- I review account-level lists twice a year and campaign-level lists monthly, tied directly to search term report pulls so exclusions are based on current query data rather than guesswork.