Most negative keyword advice assumes the problem is obvious rubbish. Somebody searches something unrelated, you block it, spend improves. That layer exists in B2B, it’s easy to clear, and it isn’t where the money goes.
The expensive waste is the query that’s entirely on-topic and entirely worthless commercially. Somebody searching your category, in your language, with genuine interest, who’ll never hold a budget for it. They click, they read, and where a form fill is the primary conversion a decent share of them convert. The click costs you twice, because a converted non-buyer teaches Smart Bidding to go and find more people exactly like them.
The four populations in B2B search that will never buy
B2B waste sorts into four groups. Three of them block cleanly before a single query arrives. The fourth doesn’t.
- Job seekers. Salary, jobs, career, vacancy, hiring, recruitment, internship, interview questions, cv, resume, certification, course. These get most expensive when the category shares a name with a job title, common in software sold to a named function. If you sell to revenue operations, everybody who wants to become one finds you first.
- Students and researchers. What is, definition, meaning, tutorial, how to, example, template, sample, free, pdf, dissertation, wikipedia. The word “free” needs a decision of its own rather than a default. In some businesses it signals a trial and it’s the best query of the month. In others it’s somebody who’ll never pay.
- Competitors, vendors and intermediaries. Competitor brand names, plus agency, consultant, reseller, partner program, white label, distributor, rfp, and anything implying the searcher wants to sell rather than buy. Conquesting is legitimate, so the list protecting a prospecting campaign mustn’t sit on the campaign bidding on competitor terms.
- The consumer sense of an ambiguous term. Every B2B category has one and you can’t guess yours from the armchair. Asset tracking also means a keyring that finds a lost bag. Monitoring also means a camera pointed at a cot. Only reading the search terms finds this group, and in the accounts I audit it’s usually the largest single line of waste in the plan.
The first three go in before launch. The fourth gets built from evidence, which is the argument for reading search terms inside a campaign’s first couple weeks rather than a tidy monthly report.
Where negative keyword lists should live in the account
Negatives sit at four levels, and the wrong level causes as many problems as no negatives at all.
- Account level is for universal disqualifiers that can never be right anywhere, with job seeker terms the clearest example. It’s the wrong home for anything that might suit one campaign.
- Shared lists applied selectively are the workhorse. A competitor list everywhere except the conquesting campaign. A research-intent list on bottom-funnel campaigns and not on the one meant to reach people early. A brand list on everything that isn’t the brand campaign. Name them after what they contain, because whoever inherits this account won’t know what “List 3” was for. Google’s documented account limits are 5,000 keywords per negative list, 20 lists per manager or child account and 10,000 negatives per campaign (with a much tighter cap of 1,000 for Display and Video).
- Campaign level is for sculpting between campaigns: brand queries out of non-brand, two campaigns kept from competing over one query, a product name that means something unfortunate in a single market.
- Ad group level, sparingly. Elaborate cross-negation suited exact-match-era Google Ads and now fights the behaviour Smart Bidding depends on. Thousands of ad group negatives is archaeology rather than strategy.
Negative match types behave differently from positive keywords
This is where the mistakes happen, because the words look identical and the behaviour isn’t.
Negatives don’t expand the way positive keywords do, and the documentation here is more interesting than the folklore. Google says negative keywords account for casing and misspellings on their own, then adds in the same breath that they won’t match close variants or other expansions. Read both lines together, because the gap between them is where money goes. Misspellings of a brand name look after themselves. Singulars and plurals don’t, so “salary” won’t cover “salaries”, and check that page again in six months (this behaviour has changed before).
Then the match types. Negative broad blocks a query only when the query contains every word in the negative, in any order, so “free trial” blocks anything containing both words, including plenty worth keeping. Negative phrase blocks queries containing that exact sequence, and it’s the right default most of the time: precise enough to be safe, broad enough to be useful. Negative exact blocks only the query as written. There’s no negative broad match modifier and there never was a negative equivalent of positive broad expansion, so anyone recommending a plus sign is working from a decade-old article.
One documented oddity explains the occasional query nobody can account for. Google notes that an ad can still show when someone searches a phrase longer than 16 words and the negative keyword falls after the sixteenth.
What to do about the search terms Google won’t show you
Google omits search terms that don’t have enough query activity to keep to its data privacy standards, and groups the low-volume ones into subthemes or an “other queries” bucket without ever exposing the queries themselves. A meaningful share of spend therefore sits against text nobody outside Google can read. Three responses:
- Measure the gap first. Sum the cost in the search terms export and compare it with campaign cost for the same window. Knowing you can see, say, three quarters of spend tells you what a query-level conclusion is worth, and it gives finance a number instead of an assumption.
- Pull longer windows. The threshold applies per report, so a ninety-day export surfaces terms a seven-day export suppresses. Review weekly on a rolling ninety days rather than reviewing last week, every week.
- Then shift weight from negatives onto structure. A query nobody can see is a query nobody can block, so controls that work without visibility matter more: match type discipline, tighter campaign themes, audience exclusions, and a conversion action that rewards qualified pipeline. A correct bidding objective filters better than any list you’ll ever build, because it operates on traffic you can’t see.
How to handle brand and competitor negatives
Brand negatives aren’t optional. Without them brand queries leak into non-brand campaigns, convert at a brand rate, and make acquisition look better than it is. Build a phrase-match list of the brand, its misspellings, the product names and the domain, then apply it to everything that isn’t the brand campaign. It’s the cheapest reporting fix in most accounts.
Competitors need a decision before they need a list. Are you conquesting or not? If the answer is yes, those terms belong in one named campaign with its own budget, because the economics differ and blending them into prospecting hides that from everybody reading the report. If it’s no, the list goes to account level and the decision gets made once instead of every quarter.
Negative keywords in Performance Max and broad match
Performance Max takes campaign-level negatives and account-level lists, with one limit to plan around rather than resent. Performance Max negative keywords apply to Search and Shopping inventory only, so they do nothing whatsoever to Display, YouTube or Gmail placements. The separate brand settings are the more important lever, and Google’s documentation notes that brand settings inherently limit traffic and that performance can fall when you apply them. That’s the trade, and you make it deliberately.
Broad match with Smart Bidding is an explicit bargain. You hand over query selection in exchange for signals nobody outside Google can access, which makes negatives the steering mechanism, and it’s why over-applying them defeats the arrangement. Negating broad match down to a narrow permitted set means you wanted phrase match.
There’s a second-order effect people miss. A negative doesn’t tell the algorithm that a query performed badly. It removes the query, so no learning happens and nobody can evaluate it again. Usually that’s what you want. Occasionally it means an assumption somebody made two years ago has been quietly protected from evidence ever since.
How often to review negative keywords in a B2B account
Treat this as maintenance rather than a setup task, and do it less often than most advice suggests.
New campaigns want a look every two or three days for the first couple weeks, when the cheapest findings appear and bad learning does the most damage. In steady state, weekly for the first quarter, then bi-weekly or monthly on a rolling ninety-day window, because reviewing seven days of data every week mostly produces noise. Once a quarter, review the lists themselves instead of the search terms: which list is applied where, what went in, and who put it there.
Everybody adds negatives. Almost nobody removes one. That asymmetry is what eventually kills the account.
A negative keyword list maintained too well will strangle the account
A list managed conscientiously for two years, by people doing exactly what the advice tells them, will suffocate the campaigns it was built to protect. Negatives only accumulate, because every review adds and none removes. Meanwhile the business moves. A new product arrives with a blocked word in its name, a new market treats a negated term as ordinary usage, or a term that meant a student in 2023 now means a self-serve buyer. The negative added for a campaign paused eighteen months ago still sits in a shared list applied to everything.
The symptom is easy to recognize. The account stops discovering anything. The search terms report looks the same month after month, the same forty queries and no new ones. Spend concentrates into a shrinking handful of terms, so average cost per click climbs while volume falls. Conversion volume declines with no improvement in cost per conversion, and that’s the tell which separates this from a bidding problem. Strangling an account doesn’t make its traffic cheaper, because what came out was the cheap end.
How to stop your negative keyword list killing the account
Record a reason and a date against every negative you add (a spreadsheet is enough). Without a reason column nobody ever feels able to remove anything, because the risk of deleting something important outweighs the invisible cost of keeping it.
Before you add a broad negative, check what it would have blocked. Filter ninety days of search terms for the phrase and read the list. Two minutes, and it prevents most of the damage anybody does with this tool. Prefer phrase over broad while you’re in there, because the precision costs almost nothing and the collateral damage from a careless negative broad has no ceiling.
Once a year, justify every line in the account-level list. Anything nobody can justify comes out, and nothing bad happens if a term reappears, because it’ll turn up at the next review. Keep the universal list short. Importing a four-thousand-term list from a blog post is the fastest route to the failure above, since it was built for somebody else’s business.
Negative keywords are a blunt instrument applied to a targeting problem, and they work best when they’re doing the least. If your list is doing a great deal of work, the real fix is usually somewhere else: the conversion action you optimize toward, the campaign structure, or the match types. That diagnosis is most of what an account audit produces, and it’s the first thing I look at in any Google Ads management engagement. Where the account is one channel among several, the question belongs in your B2B paid media strategy.