How Many Negative Keywords Are Too Many?

The number that matters is not the count

Google does not degrade, throttle or penalise an account for having a long negative keyword list. A campaign with 4,000 exact-match negatives can be perfectly healthy. A campaign with 30 broad negatives can be strangled.

So "how many is too many" is the wrong question, and the pages that answer it with a number — 100 per campaign, 500 per account, whatever — are guessing at a threshold that does not exist. The question worth asking is: how much of my traffic can each negative reach?

An exact negative like [plumber salary] blocks exactly one query. Its blast radius is one search term. A broad negative like hire blocks every query in the campaign that contains the word "hire" in any order, alongside anything else — including hire purchase excavator and plant hire quote near me, which are people with money. One line of text, thousands of dollars.

That is the whole article in two paragraphs. The rest is how to tell which of your negatives are the dangerous ones, and what to do about it.

The caps Google actually enforces

One cap comes up constantly in real accounts: a Performance Max campaign accepts up to 10,000 campaign-level negative keywords, and it does not accept shared negative keyword lists. No lists means no shortcuts. Every PMax campaign carries its own copy of every negative you want applied to it, so a large account with a dozen PMax campaigns is maintaining a dozen parallel lists by hand. That is where the 10,000 ceiling stops being theoretical — we go through the arithmetic in Negative Keywords in Performance Max and the 10,000 Cap.

Standard Shopping is easier. It takes negative keywords and shared negative keyword lists, the same way Search does. So does Search. If you are running Search plus Standard Shopping, a well-structured set of shared lists covers both without duplication.

For the per-campaign and per-list limits on Search, go and read Google's own account limits documentation rather than a blog post — including this one. Those figures have moved over the years, and quoting a number from 2019 as if it were current is how bad advice spreads. What is worth saying is that almost nobody hits them. In the accounts we audit, the lists that cause real damage are usually small: a couple of dozen broad negatives someone added years ago, sitting in a shared list attached to every campaign in the account.

One more limit that is not a count. Display, Video and App campaigns report no search terms at all — campaign_search_term_view covers Search, Shopping and Performance Max and nothing else. You can add negatives to a Display campaign, but you are working blind. There is no report to audit them against, no way to check what they blocked, and no way to know whether the list is helping. If you have inherited a Display negative list, treat it as unverifiable.

The failure mode: one broad negative in a shared list

Broad negatives are where accounts get hurt, because a broad negative blocks any query containing all of its words, in any order. The word you added to exclude one bad query keeps working forever, on queries you never imagined.

Some real examples of the shape this takes:

• A boiler installer adds broad repair to stop boiler repair cost. It also blocks boiler installation and repair quote — a customer asking for a quote.
• An e-commerce shop adds broad free to stop free running shoes. It also blocks free shipping running shoes size 9, which is a buyer checking delivery cost.
• A tool retailer adds broad diy to stop diy tutorials. It also blocks diy drill set and diy tool kit gift, which are products they sell.
• An equipment dealer adds broad hire because they sell rather than rent. It also blocks hire purchase digger finance, which is their highest-value enquiry.
• A software vendor adds broad free and later launches a free trial. Every query containing "free trial" is now unreachable, and nobody remembers why.

Multiply by a shared list applied to 40 campaigns and you have a single line of text quietly removing a slice of demand from the entire account.

Broad negatives are also the least predictable of the three match types. Broad, phrase and exact are your options, and the way a broad negative interacts with your positive keywords does not mirror the way broad positive matching works — the asymmetries are laid out in Negative Keyword Match Types Don't Work Like Positives. The practical consequence: a broad negative is the one type you cannot fully reason about from the text alone.

So when you audit a list, sort by match type first. Ten thousand exact negatives are a filing problem. Forty broad negatives are a revenue problem.

Three signs your negatives are choking traffic

Google gives you no metric for this. There is no "impressions lost to negative keywords" column, no share-of-voice figure that separates negatives from budget or rank. You have to infer it. Three signals are worth watching.

1. Negative keyword conflicts

Google Ads flags cases where one of your own negatives blocks one of your own active keywords. It shows up under Recommendations, and it is the only place the platform will tell you outright that you have blocked yourself. Check it monthly. It only catches conflicts with keywords you are bidding on, not conflicts with the long tail your keywords would otherwise have matched, so a clean report is not proof of anything — but a dirty one is proof of a problem.

2. The number of distinct search terms is falling

Pull unique search terms per week for a campaign over six months. If the count drops while budget, bids and keywords are unchanged, something is filtering. Negatives are the first candidate. This is a blunt instrument — seasonality and Google's own query matching changes move the number too — but a steady decline in query variety with flat spend is worth twenty minutes of investigation.

3. Campaigns that cannot spend their budget

A campaign sitting well under budget with impression share lost to rank near zero is not being outbid. It is not finding auctions to enter. On a broad-keyword campaign with a heavy inherited negative list, negatives are often the reason.

The reason lists rot is an information asymmetry, and it is worth naming clearly. A bad negative costs you money silently: the blocked query stops appearing in your search terms report, so there is no line item showing what you lost. A missing negative costs you money visibly: the wasted spend shows up in the report every week, in dollars. Feedback pushes you in one direction only. Every account drifts toward over-blocking unless someone deliberately pushes back. Google Ads Search Terms Report, and What It Hides covers the other blind spots in that report, which are considerable.

How to audit a list you inherited

This is a two-hour job and it is worth doing once a year on any account older than eighteen months.

Export everything. All campaign negatives, all shared lists, all PMax campaign-level negatives. One spreadsheet, with a column for which campaigns each item applies to.

Split by match type. Put every broad negative in its own tab. That tab is your work list. The exact negatives can wait, probably forever.

Flag single-word negatives. A one-word broad negative is the highest-risk object in a Google Ads account. Read each one and ask: is there any query containing this word that I would want to pay for? For "salary", "apprenticeship", "wikipedia" the answer is usually no. For "free", "cheap", "used", "repair", "hire", "jobs", "courses" the answer depends entirely on what you sell, and the person who added it four years ago was answering for a different business.

Cross-check against converting queries. Download 12 months of search terms with at least one conversion. Then check, by hand or with a formula, which of those converting terms would be blocked by each broad negative on your work list. Anything that blocks a historic converter goes immediately, no debate.

Then test the rest. Remove 15 to 20 suspects at a time, note the date, and wait two weeks. Compare spend, conversions and cost per conversion at campaign level, plus the new search terms that appear. Do it in batches so you can attribute the change.

The honest cost of that test: you will pay for junk traffic during it. Some of the queries you re-open really were worthless, and you will spend money confirming it before you re-block them properly. Budget for that. It is normally a small price compared with a broad negative that has been sitting on a converting query for two years — the sort of leak covered in How to Find Wasted Ad Spend in Google Ads, except running in the opposite direction.

What a healthy list looks like at scale

Healthy lists are long, boring and mostly exact.

Exact for individual junk queries. [plumber salary], [how much do plumbers earn uk], [plumbing courses near me]. Each one blocks one thing. You can add ten thousand of these and sleep soundly. This is why list length is a poor proxy for risk: the safest lists are the longest ones.

Phrase for patterns you have seen repeat. Once plumber salary, plumber salary uk and average plumber salary 2024 have all appeared in the report, phrase "plumber salary" catches the family and the next twelve variants you have not seen yet. Phrase is the workhorse. Note that negatives do not stretch to close variants the way positive keywords do, so plurals and common misspellings need their own lines.

Broad only for words that can never appear in a buying query in your account. That is a short list, and it is specific to your business. For most B2B software, salary qualifies. free does not, if there is a free trial anywhere in the funnel.

One purpose per shared list. Separate lists for jobs and careers, education and training, how-to and DIY, competitor brands, and clearly unrelated topics. Then you can apply the DIY list to the campaigns where DIY intent is worthless and leave it off the campaigns where DIY buyers are your customers. A single monolithic "Negatives" list applied everywhere removes that option, and it is the single most common structural mistake we see.

Date the entries. Google gives you no added-on date for negative keywords in the interface. If you maintain the list from a spreadsheet or through an API, keep your own date column. Six months later, "when did this appear and why" is the only question that matters and the platform will not answer it.

How automation changes the arithmetic

Nightly automation makes lists grow much faster than a human ever would. In a high-volume account, PPCleaner can add hundreds of negatives a month. That sounds alarming until you look at what they are: overwhelmingly exact and phrase negatives derived from search terms that actually appeared and actually cost money. Volume goes up; blast radius per entry stays small.

Three constraints matter more than the count.

Nothing that converted in the last 90 days gets blocked. A search term with a conversion in that window is off limits, full stop. Say the hard part: that is a 90-day window, not a lifetime one. A query that brought one sale 14 months ago and nothing since can be blocked. We think that is the right trade — a single conversion in 2023 is not evidence about 2025 — but it is a choice, and you should know it is being made on your behalf.

Every block is reversible for 90 days, in one click. That is what makes aggressive blocking safe rather than reckless. If a negative was wrong, you undo it and the traffic returns. After 90 days the undo record expires and the negative is just a negative like any other, which is why the audit habit above still matters.

PPCleaner adds negatives. It does not clean up yours. It never touches budgets, bids, ads or keywords, and it does not go through the broad negatives a previous agency added in 2021. If a single-word broad negative in a shared list is blocking your best queries, no amount of automated adding will find it. That is still a person with a spreadsheet, once a year.

And in Performance Max, the 10,000 ceiling is real. Because entries are mostly exact, a very large account accumulating for years can approach it. When you get close, the fix is not to stop blocking — it is to consolidate: replace clusters of exact negatives with the phrase negative that covers all of them, and delete anything blocking a query pattern that no longer exists. Ten exact entries collapsing into one phrase entry buys back nine slots and blocks more.

So: too many negative keywords is not 500, or 5,000. It is one broad negative too many, applied to one campaign too many, for one year too long.