What is the best Google Ads search terms cleanup tool?

Updated November 12, 2026

What a search terms cleanup tool does

A search terms cleanup tool reads what people actually typed before clicking your ads, separates buying intent from junk, and turns the junk into negative keywords placed where they will keep working. Done weekly, it is the single most reliable waste-removal routine in Google Ads; across audited accounts, junk search terms are the most common waste line that account audits find.

The job has four parts, and skipping any of them is why ad-hoc negative adding decays into a mess:

• Pull the full picture: search terms from Search campaigns, Shopping, and the search themes Performance Max exposes.
• Cluster junk by intent family rather than negating one query at a time, so each negative kills a pattern, not an instance.
• Place each negative at the right level: ad group, campaign, shared list, or account-level.
• Protect the winners: never negate a query with conversion history, and attach the stats to every proposed negative so the decision is auditable.

Search terms vs keywords: the mechanics

Keywords are what you bid on; search terms are what people actually typed. The gap between them is where both growth and waste live. Broad match, phrase match and close-variant matching all widen the gap on purpose: Google matches your keyword against plurals, misspellings, synonyms, implied intent and related concepts, which finds you buyers you never thought to target and junk you never agreed to pay for.

In 2026 the gap is wide by default. Broad match paired with Smart Bidding is Google's recommended setup, Performance Max matches on themes rather than keywords at all, and even exact match admits close variants. The practical consequence: the search terms report is where the account's real targeting is visible, and reviewing it is a mandatory loop, not an optional extra.

One honest limitation to plan around: Google withholds search terms below a privacy threshold, so a slice of your spend maps to queries you will never see. That makes the visible report a sample, and it makes pattern-level negatives (intent families, not individual strings) the only strategy that also covers the invisible tail.

The weekly cleanup loop

The loop that works is boring and strict. Weekly cadence matters because fresh queries arrive continuously and because junk spend compounds quietly: a bad query family that costs $15 this week costs $60 a month and $700 a year.

1. Pull 7 days of search terms across all Search and Shopping campaigns, sorted by cost. On larger accounts, filter to terms above a minimum spend so you work the material rows first.
2. Mark the obvious junk: zero-conversion terms that spent meaningfully, plus clear intent mismatches regardless of spend (jobs, free, DIY, wrong product, wrong place).
3. Cluster the marked terms into intent families. Ten different salary-related queries become one family with one phrase negative, not ten exact negatives.
4. Choose match type per family: phrase negatives for whole intent families, exact negatives for specific ambiguous terms you want to block narrowly.
5. Choose placement: ad group for mismatches local to one theme, campaign for broader ones, shared negative lists for account-wide junk so every new campaign inherits them.
6. Check each candidate against conversion history for the last 90 days; drop any query that has ever converted profitably.
7. Review PMax search themes and propose account-level negatives or brand exclusions for the junk they reveal.
8. Apply, log what was added and why, and put the next pass on the calendar.

Cluster junk by intent family

Junk queries arrive in predictable families, and each family has a standard treatment. Clustering is what keeps a negative list short enough to understand and strong enough to cover queries you have not seen yet:

The last two families show why cleanup needs judgment: 'cheap' kills a premium brand's junk traffic and a discounter's best traffic. This is the reason blanket negative lists copied from blog posts damage accounts, and the reason every proposed negative should carry its own conversion stats.

Negative keyword match types work differently

Negative keywords do not behave like positive keywords, and the asymmetry catches people. Negatives do not expand to close variants: a negative for 'job' does not block 'jobs', and a negative for 'colour' does not block 'color'. You must add plurals, misspellings and spelling variants explicitly.

The three negative types, precisely: negative broad blocks any query containing all the words in any order; negative phrase blocks queries containing the exact sequence; negative exact blocks only that exact query. Most cleanup work lands on negative phrase, which blocks the intent family without the collateral damage of broad.

Two placement facts worth memorizing: negatives block matching, so they act before bidding and cost nothing to maintain; and Shopping campaigns have no positive keywords at all, which makes negatives the only query steering wheel Shopping gives you. For Shopping, the weekly search terms pass is the targeting.

Where to place negatives

Placement decides how far a negative reaches and how maintainable the system stays. The working hierarchy, from narrow to wide: ad group negatives for cross-group routing within a campaign; campaign negatives for themes a whole campaign should never match; shared negative lists for account-wide junk, attached to every relevant campaign and inherited by new ones automatically; and account-level negative keywords, which also reach Performance Max.

The maintainability rule: junk that is junk everywhere (employment queries for almost every business) belongs in a shared list named for its family, like 'Jobs and careers'. Junk that is contextual (the query 'repair' is junk for a seller of new units and gold for a repair shop) stays at campaign level where the context lives. An account where every negative sits at ad group level has no memory: the next campaign starts naked and relearns the same lessons with fresh money.

Performance Max and Shopping: where cleanup matters most

PMax is where most wasted ecommerce spend hides in 2026, because it eats broad queries silently and reports in aggregates. The levers it gives you are few but real: search themes insights show what it matched into, account-level negative keywords now apply to it, brand exclusions stop it harvesting your brand traffic, and campaign-level negatives can be requested where the account structure supports them.

The weekly PMax pass: read the search themes and category insights, compare against what the campaign is supposed to sell, and propose account-level negatives for the clear junk families. Add brand exclusions unless you have deliberately decided PMax should serve on brand. Expect less precision than Search gives you; the goal is cutting the worst waste, not surgical control.

For standard Shopping, the pass is the same as Search but with higher stakes per negative, since negatives are the only steering (and feed quality, covered on the Shopping feed errors page, decides which auctions you enter at all). Route queries between Shopping campaigns with priorities plus negatives: brand queries to the brand campaign, generic queries to the generic one, each fenced by the other's negative list.

A worked cleanup example

Say an online store spends $5,000/month on Search and Shopping, and the first real cleanup pass finds, with illustrative numbers: a 'free shipping calculator' family that spent $180 with no sales; employment queries worth $90; queries for a product line discontinued last year at $140; and 'wholesale' queries at $210 from buyers the store cannot serve. That is $620, about 12% of the month, mapped to four phrase-negative families and one shared list.

The point of the example is the shape, not the numbers: the waste was concentrated in a few families, each family took one negative to fix, and none of it was visible from campaign-level reporting. The following week's pass finds a fraction of that, and the loop settles into maintenance mode, which is exactly what should happen.

Mistakes that break accounts

Cleanup done badly costs more than cleanup skipped. The recurring failure modes:

• Negating a converting query because one bad week made it look like junk. Check 90 days of history, not 7.
• Over-broad negatives: a single-word broad negative like 'repair' can silence a profitable theme account-wide. Prefer phrase, prefer two-word patterns.
• The one-time scrub mentality: a heroic 4-hour cleanup followed by six months of nothing. Fresh queries arrive weekly; so must the pass.
• Negating one-off low-spend noise individually: a thousand exact negatives for queries seen once each is maintenance debt, not hygiene. Work in families.
• Forgetting inheritance: negatives added only at ad group level leave every new campaign unprotected. Shared lists are the account's memory.
• Never auditing the negative list itself: positioning changes, product lines change, and last year's correct negative becomes this year's blocked revenue. Review the lists quarterly.

Where AdsGuru automates the loop

AdsGuru runs this exact loop weekly on connected accounts: it pulls search terms across Search, Shopping and PMax themes, clusters junk by intent family, checks every candidate against conversion history, and proposes negatives at the right level with each query's spend and conversion stats attached. It never proposes negating a query with conversions. You approve, it applies and logs; approval mode is the default, and routine passes can later run with less friction once the proposals have earned your trust.

Honest limits: AdsGuru needs 7 to 14 days of data on a freshly connected account before the clustering has enough material to be useful, and judgment calls like the 'cheap' family are surfaced as questions rather than decided for you. The loop is included in the flat $89/month, and the broader Google Ads feature set is on the AI Google Ads tool page; the help doc on negative keywords covers configuration.

Common questions

Related