How Close Variants Spend Your Google Ads Budget on Searches You Never Picked

Google matches your keywords to searches you never typed. Here is why it happens and how to stop the bleed today.

· QueryCut

The short version

  • Close variants cannot be turned off; your only controls are match type, negatives, and review cadence.
  • Derive your zero-conversion click threshold from your conversion rate: wait for a few multiples of (1 / CVR) clicks before cutting.
  • Cut a term when it hits ~3x target CPA AND has enough clicks for your CVR AND, if dead, has no recent clicks.
  • Negative keywords are literal: negative broad blocks all your words in any order but never expands to synonyms the way positive broad does.
  • Some spend sits in Google's hidden low-volume search terms bucket that you cannot audit, so lean on tighter match types too.

Close variants are Google's way of matching your keyword to searches that share its meaning, not just its wording, which means even an exact match keyword can trigger on queries you never added. You cannot turn close variants off. The only levers you control are the match type you choose, the negative keywords you add, and how fast you review the search terms report. Go to Campaigns > Insights and reports > Search terms right now, sort by cost, and you will see where your budget actually went.

What close variants actually do

Every match type in Google Ads now includes close variants, and that includes exact match. When you bid on the exact match keyword [running shoes], Google will also serve your ad on misspellings, singular/plural forms, stemmings, abbreviations, and searches it judges to have the same intent. That last category is the loose one. Google decides intent using its own language models, and it does not publish the logic or expose which signal triggered a given match.

This is the mechanism to internalize: your keyword is no longer a filter, it is a seed. Google uses it to estimate what you want to show up for, then runs its own query-understanding layer to decide whether an incoming search qualifies. The auction you enter is chosen by Google, not by you. That is why two accounts bidding on the same keyword can serve on completely different search terms depending on landing page, campaign history, and the broader intent signals Google attaches to each.

The practical consequence: broad match casts the widest net and will reach for related concepts, phrase match holds the meaning of your phrase but reorders and pads around it, and exact match stays closest but still drifts to same-intent queries. None of them are literal. If you assumed exact match means exact, that assumption is the first thing to drop.

Read the search terms report before you touch anything

The search terms report is the only place Google tells you what you actually paid for. Open Campaigns > Insights and reports > Search terms, set the date range to the last 30 to 90 days, and add columns for Cost, Conversions, and Clicks. Sort by Cost descending. The terms at the top are where close variants are spending your money, converting or not.

Be aware of a reporting limit that trips people up: Google does not show every search term. Terms that too few people searched are withheld for privacy, so a chunk of your spend sits in an unlabeled bucket you cannot audit term by term. Compare the total cost in the search terms report against the campaign total to see how much is hidden. If a large share is hidden, your negative keyword strategy can only do so much, and you should lean harder on tighter match types and campaign-level query themes.

For an account you inherited, this report is your first stop before changing bids or budgets. You are looking for two patterns: expensive terms with zero conversions, and terms that are clearly off-intent regardless of spend (a plumber seeing "plumbing jobs" or "plumbing salary"). The second pattern tells you the matching layer has misread the account's intent, which usually means the keywords are too broad or the negative list is thin.

The cut rule, and the math behind the click threshold

The decision you face on each wasteful term is: cut it now, or wait for more data. Cutting too early can kill a term that would have converted on click 40. Waiting too long burns budget on a genuine loser. The threshold that balances this depends on your own conversion rate, not on a universal number.

Here is how to derive it instead of memorizing one. If your normal conversion rate is 5 percent, you expect roughly 1 conversion every 20 clicks. A term that has taken 20 clicks with zero conversions is barely surprising: missing a single expected conversion is noise. By 60 to 100 clicks with still zero conversions, the term is running well below your account average, and the odds that it is simply unlucky get thin. So the real rule is: wait until clicks reach a few multiples of (1 / your conversion rate). At 5 percent that is 20, so wait for 60 to 100 clicks. At 1 percent conversion rate, one expected conversion takes 100 clicks, so you need several hundred clicks before a zero-conversion verdict means anything. At 10 percent, 30 to 50 clicks is enough.

Layer cost on top of clicks so you are not waiting forever on a cheap term. A workable combined rule: cut a term when it has spent roughly 3x your target CPA with zero conversions AND has accumulated enough clicks for your conversion rate (the multiple-of-1/CVR test above). Spend alone is not enough, because a high-CPC term can hit 3x CPA in 5 clicks, which tells you nothing. Clicks alone is not enough either, because a cheap term can rack up clicks without ever threatening your budget. Require both. And factor recency: if the last click on a dead term was over 30 days ago, cut it regardless, because old data will not improve and the term may have changed intent.

Negative keywords: what each match type really blocks

Negatives are how you fence off the variants you do not want. They work differently from positive keywords, and the biggest trap is assuming they mirror each other. They do not.

Negative exact match blocks only the exact search term, with no close variant expansion. A negative exact of [free] blocks the search "free" alone, but not "free shoes" and not a misspelling of free. This is the tightest and safest negative, but it means you have to enumerate the exact junk terms you saw in the report.

Negative phrase match blocks searches that contain your phrase in that order. A negative phrase of "free" blocks "free shoes" and "download free app" because the word free appears. Negative broad match blocks searches that contain all of your words, but, critically, it does not expand to synonyms or close variants the way positive broad match does. A negative broad of "cheap shoes" blocks searches containing both "cheap" and "shoes" in any order, but it will not block "inexpensive footwear" even though positive broad match would have matched that intent. Negatives are literal; positives are not. This asymmetry is why you still see off-intent synonym traffic after adding negatives, and why tightening match type sometimes does more than any negative list can.

Build a repeatable review instead of a one-time cleanup

A single negative keyword purge feels productive and decays fast, because Google keeps generating new variant matches as search behavior shifts. The durable fix is a cadence. Review the search terms report weekly for high-spend campaigns and every two to four weeks for the rest. Sort by cost, apply the combined cut rule, and add the clear off-intent terms as negatives at the right level: ad group for terms specific to one set of keywords, campaign or shared list for junk that applies everywhere ("free," "jobs," "salary," "reddit," "diy").

Remember the 10-word limit on a single negative keyword: Google ignores anything past the tenth word, so long-tail junk is better blocked by a shorter, high-signal fragment. Also watch for the negative that is too aggressive. A negative phrase of "job" to block "plumbing jobs" will also block "custom job quote," which could be a buyer. When a negative could plausibly appear in a converting query, use negative exact on the specific bad terms instead of a broad sweep.

Tools that mine the search terms report and pre-rank likely negatives (QueryCut among them) can compress the weekly review, but the judgment calls stay yours. The thresholds above are heuristics, not laws. You cannot see which signal Google used to match a term, and you cannot audit the hidden low-volume terms at all. Tighten match types where off-intent drift is persistent, keep the negative list current, and accept that some variant spend is the cost of staying in Google's auctions.

Questions people also ask

Can I disable close variants in Google Ads?

No. Google removed the ability to opt out of close variants years ago, and it now applies to every match type including exact. Your only levers are choosing a tighter match type, adding negative keywords, and reviewing the search terms report regularly to catch new variant matches as they appear.

Why does my exact match keyword still show for searches I never added?

Exact match includes same-meaning variants, not just the literal phrase. Google treats your keyword as an intent seed and uses its own query-understanding layer to decide which incoming searches qualify. That is why [running shoes] can serve on "shoes for running" or a close synonym you never entered. Use the search terms report to see exactly which variants are matching.

How many clicks before I trust a zero-conversion term?

It depends on your conversion rate. Take 1 divided by your normal conversion rate to get the clicks per expected conversion, then wait a few multiples of that. At 5 percent CVR that is 60 to 100 clicks; at 1 percent you need several hundred; at 10 percent, 30 to 50. Below that, a single missed conversion swings the read.

Should I use negative broad or negative phrase match?

Negative phrase blocks searches containing your words in that exact order. Negative broad blocks searches containing all your words in any order, which is wider but still literal, since it never expands to synonyms. Use negative exact for specific junk terms you want to block without risking adjacent converting queries, and reserve broad negatives for words that are always irrelevant like "free" or "jobs."

I just inherited this account. Where do I start?

Open Campaigns > Insights and reports > Search terms, set the range to 30 to 90 days, and sort by cost. Flag expensive zero-conversion terms and clearly off-intent terms first. Check how much total cost is hidden in the low-volume bucket, then decide whether to tighten match types before building out a negative keyword list.

More from the blog

Landing Page Experience: The Changes That Actually Move ItImpression Share Lost to Rank vs Budget: What Each One MeansHow Many Campaigns Is Too Many for the Budget You Have?Primary vs Secondary Conversion Actions: What Each One Actually Changes