Negative keywords are the cheapest growth lever in Google Ads

A small Phoenix HVAC operator was spending nine hundred and twenty dollars a month on the search term "free ac repair." The campaign was set up correctly. The bid was reasonable. The landing page was good. None of that mattered, because the people clicking on "free ac repair" were never going to pay for anything.

That account had been live for fourteen months when somebody finally pulled the search terms report. Nine hundred and twenty a month, every month, against a service that costs four hundred dollars to dispatch. The fix took ninety seconds: a campaign-level negative for "free." Spend fell, conversions held steady, and the campaign paid for the rest of its life.

Negatives are not glamorous. They do not show up in case studies. Nobody writes a thread about how they added the word "salary" to a list and saved a thousand dollars a month. So most accounts run for years without anyone really sweeping for them, and most of the dollars that should be cut never get cut.

This piece is the working version of how to do that sweep without breaking anything that is already working.

What a negative keyword actually is

A negative keyword is a word or phrase that tells Google not to show your ad when the search query contains it. They live at three levels: ad group, campaign, and shared library. They have three match types of their own, which behave a little differently from positive match types and which we will get into in a moment.

What they are not, despite what the name suggests, is the opposite of a positive keyword. A positive keyword tells Google "show me for searches that match this." A negative keyword tells Google "and skip the ones that include this." The two interact, but they do not cancel each other out. You can bid on the positive keyword plumber and add the negative keyword salary, and the ad will run on most plumbing searches and skip the ones containing the word salary. Both are doing useful work.

broadplumber against exact[plumber salary] as a negative is the cleanest possible block. Broad targeting on the positive side, exact match on the negative side. The ad runs on every plumbing-related search except the one specific phrase you wanted out.

Why this matters more than it should

There are a small number of words that drain spend in almost every account. Free. Cheap. DIY. How to. Salary. Jobs. Wholesale. Reddit. The exact list shifts by industry, but every account has its version. Until those words are blocked, every dollar spent on the campaign is a partial donation to whichever fraction of users typed those words.

The reason this is the cheapest growth lever is that it costs nothing to apply, costs nothing to maintain, and the savings compound forever. A new high-converting keyword has to be discovered, tested, and protected against bid creep. A new ad has to be written, queued, and rotated. A new negative is a one-line entry in a list that runs against every search forever.

That number does not mean every account is leaking thirty-eight percent. It means the variance is wide, the average is uncomfortably high, and the work to fix it is uncomfortably small.

The four kinds of waste that show up in almost every account

Negatives do real work in four directions. Most accounts have leakage in at least three of them.

Information seekers

Someone searches "how to fix a leaky pipe." They are not buying a plumber. They want a YouTube video. The ad still shows, the click still costs, and the bounce is immediate. These are the easiest negatives to find because they almost always start with "how to," "what is," or "why does."

DIY shoppers

Adjacent to information seekers but worse, because they sometimes click. "Replace garbage disposal myself," "ac compressor parts only," "fix my own roof." These people are explicitly excluding your service from their consideration, and they are paid for at the same rate as a real lead.

Wrong kind of buyer

Someone searches "wholesale roofing supply" looking for materials, and your ad for residential roofing services shows up. The intent is commercial when you serve residential, or vice versa. The query may be valid for somebody, but not for you.

Job seekers and the curious

"Plumber salary," "law firm associate hours," "what do dental hygienists make." These have nothing to do with your business and everything to do with how Google's broad match handles ambiguity. The single negative salary shuts most of these down across an entire account.

Where to look first

Most accounts hide most of their waste in the same place. Open the search terms report. Set the date range to the last thirty days. Sort by cost descending. The first fifty rows almost always contain enough leakage to justify the entire review.

The temptation is to look at every row. Resist it. The long tail of one-off searches with a click or two will not repay the time it takes to read them. The top fifty by spend will. After the first sweep, most accounts get diminishing returns past the top hundred and fifty rows in any given month.

The flagged rows are where the work is. Two patterns repeat. The first is a single tell-tale word: free, salary, diy. Block the word by itself instead of a longer phrase. One negative shuts down dozens of variants. The second is a category mismatch: someone wants wholesale, you sell retail. Use a phrase block here instead of a single word, because "wholesale" might be a term your business genuinely uses.

The single decision that matters per row

For every row that is not an obvious keep, the question is the same. Did the searcher likely intend to buy what I sell? If yes, leave it alone, even if it converted poorly today. If no, add it as a negative.

The biggest mistake new account owners make is over-blocking. They see a term that did not convert this month and they add it as a negative. Six months later the account is wheezing because every weak performer has been excluded and the remaining traffic is too narrow to find the next set of buyers. Negatives should answer "this person was never going to buy." Not "this query was a bad month."

Match types for negatives, in plain language

Negatives have three match types: broad, phrase, and exact. They behave a little differently from the positive equivalents, which trips up almost everyone the first time.

The trap with negative broad is that, unlike positive broad, it does not include synonyms or close variants. If you negative the word cheap, you do not block inexpensive. You have to block both. This is annoying, and it is the single biggest reason that "I added a negative for free, why is the leakage continuing" is the most common follow-up question on a first audit.

Where to put them: campaign, ad group, or shared list

Three levels, one rule of thumb each.

Campaign level is for blocks that apply to every ad group inside the campaign. Most negatives belong here. If you want to block "free" for the entire residential plumbing campaign, this is the place.

Ad group level is for blocks that apply to one ad group but not another inside the same campaign. Use it when the ad groups serve different intents that are easy to confuse. For example, a campaign might have an ad group for emergency repairs and an ad group for scheduled installations. The word "broken" might be relevant to the repairs ad group and irrelevant to the installations ad group.

Shared negative keyword lists apply to multiple campaigns at once. Use them for blocks that are universal across the account: out-of-service-area cities, perennial DIY terms, words that have nothing to do with the business in any campaign. The shared list is also the right place for an account-wide "evergreen" list that you build up over time.

The mistake to avoid: putting an account-wide block at the ad group level. We have audited accounts with the same ten DIY phrases blocked in fifty separate ad groups, instead of once in a shared list. Maintenance is impossible. Auditing is impossible. Just put account-wide blocks in a shared list and link the list to every campaign.

The first sweep, end to end

The first time you do a real sweep on an existing account, expect about an hour. Subsequent sweeps on the same account should take fifteen to twenty minutes.

The reason for waiting two weeks before re-sweeping is that ad serving has its own inertia. Match-type-related changes propagate over a few days. Quality Score recalculations take a little longer. Looking at the data the day after a big batch of changes is a good way to convince yourself you broke something when you did not.

What over-blocking looks like, and how to undo it

Two signals tell you that you over-blocked. The first is a sudden drop in impressions across a campaign that was steady before. The second is rising cost-per-click on the keywords that survived. Both can have other causes, but if they line up with a recent batch of negatives, look at the negatives first.

The way to back out is to remove the most aggressive blocks first. Single-word broad negatives are the highest-risk class. If you blocked the word cheap and impressions fell more than you expected, swap it for the phrase "cheap repair" instead. The phrase block will still catch the bargain hunters and let "cheaply" or "best cheap" through.

The other recovery move is checking what searches the negative is now blocking. Most ad platforms will not show you this directly, which is why journaling matters. Keep a brief note in a shared doc every time you add a non-obvious negative, with the date and the reason. Two months later, when you have forgotten, the note will save you an hour of guessing.

How a tool fits in (or does not)

Most of this work can be done in a spreadsheet. The reason a tool helps is not that the decision is hard, but that the decision repeats hundreds of times per sweep. Reading the term, judging the intent, choosing the match type, deciding the level: each is fast in isolation and grinding in aggregate.

A good tool reads the search terms report alongside a written description of the business, makes a first-pass call on each row, groups the blocks into themes, and hands the result back as a reviewable list with reasons attached. The human still makes every final call. The machine just shortens the per-row work from sixty seconds to ten.

We built Sensei Ads to do exactly this. The product takes a CSV and a one-paragraph business description and returns a categorized list of negatives with a written reason on every row. You accept, reject, or override. Then you copy the result into Google Ads Editor and you are done.

The reason to mention this here is not the upsell. It is that the choice between manual sweeps and tooled sweeps is mostly a question of how often you are going to do this. If you manage one account and sweep it once a quarter, a spreadsheet is fine. If you manage three or four and sweep them every two weeks, the tool earns the subscription back inside the first sweep.

What the rest of this section covers

Most of the pieces in the rest of this section start from a specific question that comes up in this work and answer it in more detail than this overview can.

The how-to-find piece walks through the report sweep at a slower pace, with a longer example. The match-types piece is for when broad-versus-phrase decisions get fiddly. The bulk-upload piece is the mechanics of getting hundreds of negatives into Google Ads Editor without making a mistake that takes an hour to undo. The shared-list piece is for the account-wide blocks. The Performance Max piece is for the campaign type that has its own awkward rules around negatives.

If you are starting from zero, read this pillar, then go to the how-to-find piece. The rest are situational.