Account-Level Negative Keyword Lists vs Campaign-Level

Most guides to negative keywords describe three levels — account, campaign, ad group — note that shared lists exist, and stop. That leaves the question people actually have unanswered: when two layers disagree, which one wins?

The honest answer is that they never disagree, because negatives have no mechanism for disagreement. There is no "allow" in Google Ads. A negative keyword can only subtract. So the layers combine, and the union of all of them is what applies. Once you internalise that, the real decision stops being which level wins and becomes which level can I get out of later — and that question has a much more interesting answer.

The four layers, and how they combine

Google Ads lets you exclude a search term in four places. Each applies to a different slice of the account, and the slices overlap rather than override.

Account-level negative keywords sit in Tools → Shared library → Exclusions lists, under the account-level tab. They apply to every eligible Search and Shopping campaign in the account, now and in the future. You do not attach them to anything. They just apply. New campaigns inherit them automatically, which is the point and also the trap.

Shared negative keyword lists also live in the shared library, but you attach them to specific campaigns. One list can be attached to fifty campaigns; one campaign can have several lists attached. Edit the list once and every attached campaign changes. Detach it from a campaign and that campaign stops applying it, immediately, without touching the list itself.

Campaign-level negative keywords are typed into a single campaign and apply only there. Nothing else sees them.

Ad group-level negative keywords apply to one ad group. Their most common legitimate use is not blocking waste at all — it is steering traffic between ad groups in the same campaign, so the "running shoes" ad group does not eat queries meant for the "trail running shoes" ad group.

Now the combination rule. At auction time, Google collects every negative that applies to the ad group being considered: its own ad group negatives, its campaign's negatives, every shared list attached to that campaign, and the account-level list. If the search term matches any one of them, the ad does not show. There is no priority order, no "more specific wins", no way to carve an exception. If repair is an account-level negative and your new campaign is a repair campaign, that campaign is dead on arrival and nothing you type inside it will revive it.

This is the single most expensive misunderstanding in the whole topic. People add an account-level negative in year one, hire a new manager in year three, launch a campaign that mysteriously gets no impressions for its best keyword, and spend two days checking bids and budgets. The exclusion list is not in the campaign. It is not in the change history they are looking at. It is three menus away in a part of the interface nobody opens.

Which layer wins when they overlap

None of them. A term blocked anywhere is blocked everywhere that layer reaches. You cannot un-block it at a lower level.

It is worth being specific about what this rules out, because people keep trying.

You cannot put cheap on the account-level list and then "allow" it in your clearance campaign by adding cheap as a keyword there. Negatives beat positives, always. Your clearance campaign can have cheap running shoes as an exact match keyword with a high bid and a perfect landing page, and it will still never serve, because the account-level cheap excludes the query before the auction.

You cannot make a campaign-level negative "weaker" than an account-level one. There is no strength.

You cannot use a broad negative at account level and a narrower exact positive below it to claw traffic back. Same reason.

What you can do is move the negative to a layer that does not touch the campaign you want to protect. That is the only tool. If cheap needs to block ten campaigns and allow one, it does not belong at account level — it belongs on a shared list attached to those ten. The decision about which "free" and "cheap" queries are actually worth blocking is a separate judgement, but once you have made it, the layer you put it on is what determines whether you can change your mind cheaply.

One genuine conflict case does exist, and it is not between layers. It is between a negative and your own keyword. If you add the exact negative [plumber near me] and you also bid on plumber near me as an exact keyword, Google will flag it in the interface as a conflicting negative. That is a warning, not a resolution — the negative still wins. The report lives under the keyword view's conflict column and is worth checking after any bulk upload, because a bulk upload is exactly how that mistake happens.

What account-level negatives cover — and what they quietly skip

Account-level negative keywords apply to Search and Shopping campaigns. They do not apply to Performance Max. That exception is large enough to undermine the whole "set it once at account level" strategy for most modern accounts.

Performance Max accepts campaign-level negative keywords only — up to 10,000 of them per campaign — and accepts no shared negative keyword lists at all. So a term you excluded once at account level and considered handled is still live in every PMax campaign you run. If PMax is 60% of your spend, your account-level list is covering 40% of your money.

This is the part of the topic that the general guides get wrong by omission. They were written when Search was the account, and they describe a hierarchy that PMax does not join. Practically it means every negative you care about has to exist twice: once at account or shared-list level for Search and Shopping, once as a campaign-level negative inside each PMax campaign. Duplicating a 400-term list across six PMax campaigns by hand is not a task anyone does twice, which is why most accounts have a well-tended Search exclusion list and PMax campaigns with eleven negatives in them. The 10,000-term cap and how to spend it matters more than it sounds, because that budget is per campaign and you are now maintaining it per campaign too.

Two more coverage gaps worth stating plainly:

Display, Video and App campaigns do not use negative keywords the way Search does, and they report no search terms at all. campaign_search_term_view returns rows for Search, Shopping and Performance Max. Nothing else. So no layer of negative keyword list is doing anything for your Display spend, and no amount of list hygiene will tell you where that money went. For those campaigns the lever is placement exclusions, with all their limits.

Standard Shopping does accept negative keywords and shared lists, like Search — but they behave differently from Search in a way that catches people out, because Shopping has no keywords of its own to conflict with and your negatives are the only query control you have. That is covered in more detail in how Shopping negatives differ.

Match types stack too, and broad negatives stack worst

Each of the four layers can hold broad, phrase or exact negatives, and the match type is what decides how much collateral damage a layer causes. Broad negatives at account level are the most dangerous object in a Google Ads account.

A broad negative keyword blocks a search term only if the term contains all of the words, in any order. free trial as a broad negative blocks "free trial software" and "trial free crm". It does not block "free software" — only one of the two words is present.

That sounds safe. The problem is scale. A single-word broad negative like free or jobs or used blocks every search term that contains that word anywhere, forever, across every Search and Shopping campaign in the account. used at account level kills "used car dealer near me" for a used car dealer who later opens a second campaign. And broad negatives behave unpredictably across match types in ways that do not mirror how positive broad match works — a positive broad keyword can match a synonym or a related concept, a broad negative cannot, so the symmetry people assume is simply not there.

The rule that survives contact with real accounts: the higher the layer, the tighter the match type. Account level should be phrase or exact only, and single-word broad negatives should not be there at all. A campaign-level negative can afford to be broader, because its blast radius stops at the campaign and you can read that campaign's search terms the next morning to see what it cost you.

Phrase negatives are the usual compromise. plumber salary as a phrase negative blocks "plumber salary uk", "average plumber salary", "how much is a plumber salary" — all the job-seeker and salary traffic that will never convert — without touching anything that mentions a plumber for other reasons.

Choosing a layer: a decision you can actually apply

Put a negative at the highest layer where you would never want an exception, and no higher. Three questions get you there.

Would this term be wrong for every campaign in this account, including ones that do not exist yet? If yes, account level is correct. The list that belongs here is short and boring: obscene terms, "salary", "jobs", "how to become a", "wikipedia", "torrent", "free download" for a business that sells nothing free. Terms whose irrelevance is a property of your business, not of a campaign. In accounts we audit, this list is usually healthy at somewhere between 50 and 300 terms. Past that it has started absorbing campaign decisions and will cause an unexplained zero-impression campaign within a year.

Would this term be wrong for most campaigns but right for one or two? Shared list. Competitor names are the textbook case: you want them out of generic Search and Shopping, and you may well want a dedicated conquesting campaign that bids on exactly those names. Attach the list everywhere except there. The block-or-keep call on competitor terms is worth making deliberately rather than inheriting from an old list. Shared lists are also the right home for industry starter sets — the per-industry lists worth beginning with are attachments, not account-level commitments.

Is this term wrong because of what this specific campaign sells? Campaign level. A campaign for commercial HVAC should block "residential", a campaign for residential should block "commercial", and neither block makes sense anywhere else in the account.

Ad group level is the fourth answer and the rarest. Use it for traffic routing inside a campaign, not for waste. If you find yourself adding waste negatives at ad group level, the same term is probably wasting money in three other ad groups you have not looked at yet.

A fifth consideration cuts across all of them: who else touches this account. On agency accounts the layer choice is partly a documentation choice, because an account-level negative is invisible to whoever inherits the account unless they go looking. Keeping negative keyword work legible across client accounts usually means preferring named shared lists over account-level entries, even when account level would technically do the job. A list called "Jobs & salary — all clients" explains itself. An entry in the account-level exclusion tab explains nothing.

The reversibility question nobody asks first

The layer you choose determines how expensive it is to be wrong, and you will be wrong about some of these.

Detaching a shared list from a campaign takes one click and reverses instantly. Deleting a campaign-level negative takes one click. Deleting an account-level negative also takes one click — but finding out that it is the cause of your problem takes days, because nothing in the campaign view points at it. Reversibility is not just about the undo action. It is about how fast you can diagnose that the undo is needed.

This is why a negative that blocks a term which later converts is worse than a negative that blocks nothing. You never see the conversion that did not happen. There is no report of ads you failed to show. The only signals are indirect — impressions falling on a keyword you still bid on, a campaign whose search term list gets shorter every month, a sudden drop in conversions with stable spend. Those are the signs that negatives are hurting performance, and they all take weeks to read.

Two habits make the cost bearable. First, date and label everything. Shared lists can be named; use the name to say when and why. Second, check the search terms report after a bulk change rather than a month later — how often to look depends on spend, but the week after a 200-term upload is not the time to be on holiday.

PPCleaner's own position on this is deliberate. It adds negatives at campaign level, not account level, including inside Performance Max campaigns where campaign level is the only option available. That means more rows than an account-level entry would need, and it is the trade we chose: a campaign-level block is visible in the campaign that it affects, and reversible in one click for 90 days. It also never blocks a term that converted in the last 90 days, which removes the most expensive class of mistake from the list but not all of them. It does not touch your account-level exclusion list, and it will not tell you that a term you are wondering about is already blocked there — that is still a manual check, and on an inherited account it should be the first one you run.

What to check in an account you have just inherited

Before you add a single negative to an account you did not build, read the ones already there. Four places, in this order.

Account-level exclusion list. Tools → Shared library → Exclusions lists → account-level negative keywords. Read every entry. Look for single-word broad negatives, product categories the client has since started selling, and anything that overlaps with a current campaign's keywords. This takes twenty minutes and explains more zero-impression mysteries than any other check.

Shared lists and their attachments. The list contents matter less than which campaigns they are attached to. A list attached to eleven of twelve campaigns is usually a mistake in the twelfth, in one direction or the other.

Campaign-level negatives, campaign by campaign. Tedious. Do it for the top three campaigns by spend at minimum.

Each PMax campaign's negative keyword list separately, because nothing is shared and each one drifted on its own. Expect them to disagree with each other.

Then compare all of it against the last 90 days of search terms. The gap between what the lists block and what the account actually paid for is your real wasted spend number, and it is almost always larger on the PMax side, for the structural reason that no shared list reaches it. If you want the whole sweep in one sitting, the one-hour audit checklist puts these four checks in sequence with the rest.

One last thing the general guides never mention. Deleting an account-level negative does not retroactively re-admit anything. Your ads do not start showing for the term immediately in every campaign — they become eligible, and then they have to win auctions with no recent history on that query. On a mature campaign that is a few days of volatility. Budget for it rather than panicking on day two.