Negative keywords are one of the highest-leverage levers for lowering cost per lead, and most accounts underuse them. This covers the search terms report, how negative match types actually behave (which is not how most guides describe it) and where to put negatives so they stay maintainable.
Why negatives move cost per lead
Every click from an irrelevant query costs real money and produces nothing. Blocking those queries shrinks the denominator: fewer clicks, the same conversions, a better conversion rate.
The effects compound from there. Fewer irrelevant impressions raise expected click-through rate, which feeds Quality Score. Higher Quality Score lowers the cost per click Google requires for the same position. Lower CPC and higher conversion rate together push cost per lead down from both directions.
If you’re running broad match with Smart Bidding, this matters more. Google is exploring far more queries on your behalf, and the negative list is the only thing bounding that exploration.
What negatives are, and how they differ from regular keywords
Regular keywords invite traffic. Negatives forbid it: they tell Google not to show your ad when the search contains a given term.
Selling CRM software, adding free, jobs, and DIY as negatives keeps you out of “free CRM
software,” “CRM jobs near me,” and “DIY CRM setup.”
The bar for exclusion is fundamentally irrelevant to the business, not merely hasn’t converted yet. Those are different tests, and conflating them is how accounts quietly strangle their own discovery.
Match types: negatives do not behave like positives
Search campaigns support three negative match types: broad, phrase, and exact. Broad is the default.
Here is the part most articles get wrong. Positive keywords
match close variants automatically:
plurals, misspellings, and synonyms. Negatives do not. Add job as a negative and “jobs” still
triggers your ad. Block car and “cars” is still fair game.
You have to add every variant explicitly. This single misunderstanding causes more wasted spend than almost any other setup mistake, because it fails silently. The negative looks like it’s working, and the plural quietly keeps spending.
Negative broad match
Blocks a query only when all the words in your negative appear somewhere in it, in any order, with extra words allowed.
Negative broad cheap software blocks “cheap CRM software” and “software cheap plans,” because both
words are present. It does not block “cheap programs” or “discounted software,” since each is missing
one of the terms. It also does not block “cheap softwares,” because that plural is a close variant.
The risk is over-blocking. A generic word like tools as a broad negative can quietly kill
legitimate queries across a whole campaign. Use broad negatives sparingly, mainly for universal
blockers like profanity, where occasional over-blocking is acceptable.
Negative phrase match
Blocks any search containing your words in the same order, with words allowed before or after but not inserted between them.
Negative phrase "how to" blocks “how to build a CRM” and “learn how to use CRM.” It does not
block “CRM how-to guide,” because the order differs.
Close variants still aren’t included: "job" does not block “jobs.” Phrase negatives are the
right tool for patterns of intent: "how to", "definition", "tutorial".
Negative exact match
Blocks only when the search is exactly your term, with nothing before or after.
Negative exact [crm pricing] blocks “crm pricing” and nothing else. Not “crm pricing 2026,” not
“crm pricing plan,” not “crm pricings.”
Use exact as your default when clearing waste out of the search terms report. It’s the safest option: you block exactly what you confirmed was waste, without cutting off long-tail queries that might convert.
Account, campaign, and ad group level
Negatives apply at three tiers. If a term is blocked at account level, that wins regardless of what lower levels allow. Ad group negatives can block terms the campaign doesn’t exclude.
Shared lists can be attached at account and campaign level. Performance Max and Shopping respect account-level negatives and attached lists for search inventory, though behavior differs on non-search placements.
Account level is for terms that should never trigger any ad anywhere: profanity, free,
job-seeker terms, pure research terms, geographies you don’t serve. Find it under Admin →
Account Settings → Negative keywords.
Don’t put competitor or product-line terms here. Those belong lower down, where you can change them without affecting the whole account.
Campaign level handles strategy differences. Exclude jobs from lead-gen. Exclude cheap from
a premium line. If you run one campaign for “CRM software” and another for “marketing automation,”
cross-negatives keep queries routed to the campaign whose landing page actually answers them.
Ad group level prevents overlap inside a campaign. With separate enterprise and SMB ad groups,
enterprise goes negative in the SMB group and vice versa, so each query reaches the ad written
for it.
Shared lists are the maintainable unit
Shared lists live under Tools → Shared library → Negative keyword lists. One edit propagates to every campaign the list is attached to, which is the whole point. Campaign-only negatives mean hunting through campaigns every time you learn something.
Use a naming convention so the purpose is obvious a year later:
NEG - Universal - Brand Safety
NEG - B2B - Jobs & Education
NEG - Geo - Non-USGoogle’s account limits currently allow 5,000 negative keywords per shared list, 1,000 account-level negatives, and 20 lists per account. Worth knowing before you design a structure that can’t fit.
You don’t need a huge list on day one. Start with 50 to 100 terms and let real search data grow it.
The search terms report, and what it hides
Find it at Campaigns → Insights and reports → Search terms. Older accounts and plenty of guides still point at the Keywords view, which reaches the same report.
The critical caveat most guides skip: not every triggering query is visible. Google withholds low-volume queries for privacy reasons, so long-tail searches are frequently missing. Absence from the report is not proof a query never spent your money.
Practically, that means two things. Judge performance at the keyword and campaign level as well as query by query. And if a keyword’s cost per lead is climbing while its search terms look clean, hidden queries are a likely explanation.
Columns worth having: Search term, Match type, Campaign, Ad group, Clicks, Cost, Conversions, and Cost/conv.
Filters that surface waste fast
For a lead-gen account, filter the last 30 days:
- Conversions = 0
- Clicks ≥ 5
- Cost ≥ your target CPL
Then sort by Cost descending, so the most expensive non-converting queries surface first.
Also filter for terms containing free, job, salary, definition, how to, or geographies
outside your service area.
On newer campaigns, drop to Clicks ≥ 3 and Cost ≥ half your target CPL, but stay conservative. A term that hasn’t converted in 40 clicks is waste; a term that hasn’t converted in 4 is unproven.
Adding them from the report
Tick the boxes next to unwanted queries and click Add as negative keyword. The dialog lets you add to the ad group, the campaign, or a list. Use a list where you can.
The interface defaults to negative exact. Switch to phrase when you’re blocking a pattern rather than one query.
And because negatives skip close variants, add the variants yourself. When “internship” shows up burning budget, add “internships” in the same pass. Same for common misspellings.
A starting list for lead gen and B2B
A starting point. You’ll still prune it against your own search terms. Some of these are
legitimate for certain businesses (trial for SaaS, template for a design tool), so read
before you paste.
Universal low intent: free, cheap, discount, coupon, sample, trial, template, PDF, example,
definition, tutorial, "how to"
Job seekers: jobs, job, hiring, careers, career, salary, internship, internships, recruiter, recruiting, resume, CV
Education and training: course, courses, class, classes, college, university, training, certification, degree, school, bootcamp
DIY and research: diy, "do it yourself", guide, guides, blog, blogs, wiki, Reddit, Quora,
forum, forums, YouTube
Brand safety (profanity, slurs, adult terms) belongs in its own account-wide list.
The weekly cadence
- Pull the search terms report for the last 7 days, or 14 on lower-volume accounts.
- Filter for waste: Conversions = 0, Cost ≥ 50 to 100% of target CPL, Clicks ≥ 3 to 5.
- Add negatives for clear mismatches. Default to exact. File them in the right shared list.
- Look for recurring patterns (“template,” “sample,” “salary”) and add those as phrase negatives.
- Promote the winners. Top converting queries become their own keywords, in the ad group with the ad written for them.
- Log it: date, list, terms, reasoning. Three weeks later when volume drops, you’ll want to know what you blocked and why.
How negatives interact with broad match
Broad match with Smart Bidding explores widely, and that exploration is where the waste comes from. A solid negative list is what keeps the algorithm learning from genuine intent rather than noise.
A broad keyword crm software might pull “crm jobs,” “crm tutorial,” and “crm pricing.” If only
the last converts, block the first two and promote “crm pricing” into its own ad group with
matching copy.
One caution: when testing a new broad keyword, hold off on negatives for a week or two so data accumulates. Over-aggressive negatives strangle discovery, and you can’t tell a bad query from an unproven one on 3 clicks. If you suspect you’ve over-blocked, pause a suspect negative temporarily and watch what happens to CPL.
Common mistakes
- Assuming negatives cover close variants. They don’t.
jobwon’t block “jobs.” Add every variant. - Putting strategic terms at account level. A phrase negative
"software"applied account-wide kills every software query you have, including the converting ones. - Broad negatives on generic words.
toolsorplansas broad negatives block far more than you intended. - Never revisiting old lists. Business changes. A term that was irrelevant last year might describe this year’s product. Audit quarterly.
- Blanket competitor negatives. If you deliberately target comparison searches, blocking competitor names kills the strategy. Decide, don’t default.
Where this fits
Start conservative on day one, with account-level and shared lists covering the obvious junk, then let real search data drive weekly refinement for the first 60 to 90 days. Accounts running broad match with automated bidding need this managed more actively, because the exploration is wider.
A disciplined negative keyword practice is one of the most controllable ways to lower cost per lead. It needs no new creative, no new landing pages, and no more budget. It needs someone to look at the data every week.