A homeowner types "how to fix a leaking shower valve" into Google. Your plumbing ad shows. They click, read your service page for eleven seconds, and leave. You just paid $18 to help someone avoid hiring you.
That is the entire problem negative keywords solve. They are the list of words that, when they appear in a search, stop your ad from entering the auction. Nothing about your bidding, budget, or ad copy fixes this. Only exclusions do.
Most local service accounts have somewhere between zero and fifteen negative keywords in them. That is not a strategy — that's a rounding error.
What a negative keyword actually blocks (and what it doesn't)
Negative keywords behave differently from the regular kind, and the differences cost people money.
They do not match close variants. Google's documentation is explicit: if you exclude the negative broad keyword "flowers," your ad still can show for "red flower." Applied to your business: blocking "faucets" does not block "faucet." Blocking "job" does not block "jobs." You have to add singulars, plurals, and synonyms yourself. Misspellings and capitalization are handled for you; word forms are not.
Negative broad match is the default, and it works nothing like positive broad match. It blocks a search only when every word in your negative appears in it, in any order. Negative "running shoes" blocks "blue running shoes" and "shoes running" but not "blue tennis shoes."
Negative phrase match blocks searches containing your words in that order, with extra words allowed on either end. This is the workhorse for most local businesses.
Negative exact match blocks only that precise query with no extra words. It is almost useless as a primary tool — the same bad search reappears tomorrow with one extra word attached.
One more oddity worth knowing: if someone types a search longer than 16 words and your negative keyword lands after the 16th word, your ad can still show. Rare, but it explains the occasional impossible-looking search term in your report.
Practical rule: use single-word negative broad for whole concepts you never want (jobs, salary, DIY), and negative phrase for specific two- and three-word combinations you want to kill along with their variations.
The seven buckets of waste in local service accounts
Before you look at any data, you can predict most of the leak. These categories cost local advertisers money in almost every trade:
1. DIY and instructional. how, how to, DIY, tutorial, video, youtube, fix myself, step by step, diagram, instructions. A roofer bidding on "roof leak" will collect "how to patch a roof leak" clicks forever.
2. Jobs and training. job, jobs, hiring, careers, salary, apprentice, apprenticeship, school, course, certification, license, training, resume, union. HVAC and electrical accounts get hammered by this one.
3. Free, cheap, and price-shopping. free, cheap, cheapest, discount, coupon, low income, government grant, assistance program, pro bono. Be careful here — "free estimate" and "free consultation" are usually terms you want. Block "free" as a phrase in context ("free repair," "free service"), not as a single broad word.
4. Products and parts. parts, supply, supplies, wholesale, home depot, lowes, amazon, rental, rent, buy, for sale, used, kit, replacement filter. Someone shopping for a $34 capacitor is not booking a $400 service call.
5. Services you don't actually perform. This is the biggest one and the most business-specific. A residential plumber should exclude commercial, industrial, municipal, septic, well drilling, sewer excavation — whatever is genuinely outside scope. A family dentist should exclude oral surgery, orthodontist, braces, veterinary. A personal injury firm should exclude criminal, divorce, immigration, bankruptcy, estate, will, notary.
6. Geography you don't serve. Radius targeting isn't airtight, because people include place names in searches. Someone in your service radius searching "plumber Phoenix" when you serve Tucson is a wasted click. Block the named cities, suburbs, and neighboring states you won't drive to.
7. Research, reviews, and complaints. reviews, complaints, reddit, forum, scam, lawsuit against, salary, wikipedia, cost calculator, average cost. "Average cost" is arguable — some owners want it, most don't. Decide deliberately.
That list is a starting hypothesis. Your search terms report is the evidence.
Build from your own data, not a downloaded list
Open the search terms report (Insights and reports → Search terms, or from the Keywords page). Set the date range to the last 30–90 days. Sort by cost, descending.
Now read every term that spent money and produced no conversion. Not skim — read. For each one, decide: relevant, irrelevant, or unclear. The irrelevant ones become negatives; the unclear ones go on a watch list.
Two things to know about this report. First, Google withholds low-volume search terms for privacy reasons; queries that don't meet a volume threshold get bundled into "other search terms" rather than shown individually. So the report is a sample, not a census — which is exactly why the predictive buckets above matter. You cannot only react to what you can see.
Second, sort by clicks as well as cost. A term with nine clicks and no conversions at a low CPC is still a pattern worth blocking.
This is the part almost nobody does consistently, and it's why so many local accounts underperform for reasons that have nothing to do with bidding or creative. If you want the full build-and-maintain system rather than the summary — campaign structure, conversion tracking, the review cadence — that's what Launchpad is for.
Where to put your negatives
There are three levels, and using them correctly saves you a lot of duplicated work.
Account level. Found under Admin → Account settings → Negative keywords. These apply across eligible Search and Shopping inventory in your whole account, including Performance Max. The cap is 1,000 keywords. Use this only for terms that should never trigger any ad you run: jobs, salary, free porn-adjacent junk, cities you don't serve, DIY.
Shared negative keyword lists. You can build up to 20 lists of up to 5,000 keywords each and apply them to whichever campaigns you choose. This is the right home for themed lists: "Careers," "DIY," "Out of Area," "Commercial." Build once, apply to every campaign, update in one place.
Campaign and ad group level. Use these for separation. If you run a drain cleaning campaign and a water heater campaign, add each one's terms as negatives in the other so they stop stealing each other's searches. Same logic for a brand campaign versus a generic campaign.
Automated campaigns still respect negatives
If you're running Performance Max, negative keywords used to be nearly impossible to manage — the campaign-level cap was 100. In March 2025 Google raised it to 10,000 per campaign, matching Search. Important limitation: PMax negative keywords apply to Search and Shopping inventory only. They do not stop your ad from appearing on YouTube, Display, or Gmail — those need placement and content exclusions instead.
On the Search side, AI Max for Search came out of beta on April 15, 2026, and Google said campaigns using automatically created assets and the campaign-level broad match setting would be auto-upgraded to it. More AI-driven query matching means more surface area for irrelevant searches. Negative keywords remain one of the hard controls you keep, so get your exclusion lists in place before your campaigns get expanded matching, not after.
The four ways owners block their own customers
Over-blocking with broad single words. Excluding "repair" because of one bad search will nuke half your traffic. Ask: does any legitimate customer ever use this word?
Blocking a word that's also a local place name. "Free" is fine to block until you learn there's a Freeport in your service area. Check your negatives against local street and town names.
Blocking "emergency" or "near me" style modifiers because a couple of them didn't convert. Those are usually your best terms; the problem is the landing page or the phone answering, not the keyword.
Adding negatives and never reviewing them. Services change. A list built for the business you ran in 2023 will quietly suppress the services you added last year. Audit the list twice a year.
The 20-minute weekly routine
Once a week, on the same day: open the search terms report for the last 7 days, sort by cost, read everything above a click or two, add the junk as negatives (phrase match unless you're blocking a whole concept), and note anything that looks like a good search you're not already bidding on — that's your next keyword.
Twenty minutes. Every week. That single habit separates local accounts that get cheaper over time from the ones that stay expensive forever.
One related note: if the underlying problem is that your campaigns were never structured around distinct services in the first place, negatives will only patch it. That's the structural failure we covered in Why Most Local Service Businesses Fail at Google Ads — worth reading alongside this.