Close Variants in Google Ads and Why Exact Isn't Exact

What close variants are today

A close variant is a search that Google lets your keyword match even though the words aren't the same. Exact match and phrase match both have them. Your [emergency plumber] keyword can show for "emergency plumbers", "emergancy plumber", "plumber emergency" and "24 hour plumber". The last one doesn't share a word with your keyword at all.

Google has widened this several times. It started with misspellings and plurals. Then came stemming ("design" and "designer"), abbreviations, accents, reordered words and dropped function words ("shoes for kids" and "kids shoes"). The biggest step was "same meaning" and "same intent". Now Google decides whether two searches mean the same thing, and if it thinks they do, your exact keyword is allowed to show.

That last step is why "exact" no longer works as a precision tool. Exact match used to be a promise about the words. Now it's Google's judgement about meaning. That judgement is usually right, but not always, and it can't know your business. It doesn't know you repair roofs but don't replace them. It doesn't know you sell a $300 service and not a $30 one.

To see how this sits alongside broad match, phrase match and the other reasons odd queries turn up, read why Google Ads shows irrelevant search terms. This article is narrower. It's about the variants your exact and phrase keywords let in, and what to do about each kind.

Most close variants are harmless, and blocking them costs you

Most close-variant matches in a healthy account are searches you would have bid on anyway. "running shoe" for [running shoes]. "accomodation paris" for [accommodation paris]. "car insurance" for [auto insurance]. "divorce attorney" for [divorce lawyer]. These are the same person with the same need, typing differently.

This is where the advice that ranks for this topic falls short. The usual fix is a script or a "close variant manager". It lists every search term that doesn't exactly equal its keyword and suggests negating it. Some advertisers have run that for years, and the "close variants are a ripoff" view on Reddit comes from the same instinct: any query I didn't type is a query I didn't want.

That approach has two problems.

It blocks good traffic. The person who types "emergancy plumber" at 2am has a burst pipe. They aren't a worse lead because they misspelled it. If you negate every typo and plural, you lose real customers, often at a lower CPC than the version you wrote. They may also never come back through another keyword, because there may not be one that matches them.

It treats every mismatch the same. "Running shoe" and "roof replacement" both get flagged because neither matches the keyword text. One costs you nothing and the other costs you money. A tool that compares strings can't tell them apart, because the difference isn't in the letters. It's in what the searcher wants.

So the job isn't "find close variants". The job is to find the small set of close variants where intent has shifted, and handle each one correctly. Sometimes that means a negative. Sometimes it means a keyword change. Overdoing negatives has a real cost, and the signs are covered in negative keywords hurting performance? six signs.

The close variants that change intent

The risky variants fall into a few families. Each one has a typical way of swapping your buyer for someone else.

Same meaning that isn't the same job

This is the most expensive family, because the queries look reasonable and the clicks cost the same.

Take a handyman business that does small roof repairs but no full replacements. It bids on [roof repair]. Google can treat "roof replacement" as close enough in meaning. To a searcher, repair and replacement are neighbours. To this business, a replacement lead is useless. They can't quote it, and the caller hangs up when they hear that.

Say the keyword costs about $15 a click and "roof replacement" brings in 60 clicks a month. That's $900 a month on leads the business can't serve. And it doesn't show up as a strange query. It sits in the search terms report looking like exactly the kind of thing a roofer would want.

The same thing happens with [tax accountant] matching "tax preparer". Some of those searchers want a seasonal storefront at a fraction of your fee. It happens with [boat rental] matching "yacht charter" too: a different price tier and a different customer. Whether those variants are waste depends on what you sell. That's exactly why Google can't make the call for you.

Stemming that turns a buyer into a learner

Stemming lets one word stand in for its family: train, trainer, training. Often that's harmless. Sometimes the noun names the person you hire and the other form names the skill you learn.

[dog trainer] is someone looking to pay a professional. "dog training" includes those people, plus owners looking for tips, videos and free guides. [kitchen designer] means hiring someone. "kitchen design" also brings in people browsing photos and looking for planning software. [personal trainer] and "personal training" split the same way, and "personal training" also pulls in people looking for certification courses.

These are mixed variants. Some of the searchers are buyers, so a blanket negative would cut real leads. The right move is usually a keyword change, covered below.

Singular and plural that name different products

Google treats singular and plural as the same word. That's fine until the singular is a different object.

[reading glasses] can match "reading glass". That's a magnifying glass, a different and much cheaper product. [contact lenses] and "contact lens" are the same thing, so no problem there. But [blinds] and "blind" can drift into searches about visual impairment. [scales] and "scale" can drift into anything from music to model building. These matches are rare, but when they happen the click is close to worthless. They're also easy to miss, because the query looks almost identical to your keyword when you skim the report.

Abbreviations with more than one meaning

Abbreviations save typing and add ambiguity. A physiotherapy clinic bidding on [physical therapy] may find "pt" queries in its report. "PT" also means personal trainer. An HVAC company bidding on [air conditioning repair] will be glad of "ac repair". But depending on the rest of the query, that can also bring in people with a broken car AC, which is a different trade.

Abbreviations are worth a look whenever the short form is used in more than one industry. The fix depends on how often the wrong meaning shows up. A few clicks can be ignored. A steady stream needs a negative.

Reordered words and dropped words

Google says it only reorders words or drops function words like "for", "to" and "in" when the meaning stays the same. Usually it does. But direction and relationship words carry meaning: "flights new york to miami" is not "flights miami to new york". "Software for accountants" (you want to sell to accountants) isn't the same as "accountants for software companies" (someone wants to hire one). When a reordering flips who is buying from whom, treat it as a different query, whatever the match type column says.

Negative or keyword change: how to decide

Once you've found a variant that shifts intent, you have three options. Leave it, block it, or move it. Three questions decide which.

1. Can this searcher ever buy from you? If the answer is no, add a negative. "roof replacement" for a business that doesn't replace roofs. "reading glass" for an optician. "personal training certification" for a gym selling sessions. Nothing about bids or ads will turn these people into customers.

2. Does it sometimes convert, just at a worse rate or a lower value? Then don't negate it. Make it its own exact keyword. Add [dog training] next to [dog trainer], in its own ad group if the message should be different. Google says that when a search term is identical to one of your exact keywords, that keyword gets priority. So "dog training" will now route to the new keyword, and you can set a lower bid, write ad copy that says "book a trainer", and send it to a landing page that qualifies the visitor. You've turned a hidden problem into a line item you control.

3. Is the searcher right but the destination wrong? Sometimes a close variant matches the right business but the wrong ad group. "Emergency plumber" traffic ends up under your [plumber] keyword with a generic ad, when you have an emergency ad group with a better page. That's a routing problem, not a waste problem. Add the exact keyword to the right ad group. If the variant keeps going to the wrong place, add a negative in the wrong ad group only, not across the campaign. A negative at the wrong level blocks a good customer everywhere. The trade-offs between levels are covered in account-level negative keyword lists vs campaign-level.

One more check before you block anything: look at conversions over a long enough window. A variant with 30 clicks and no conversions this month may have converted twice last quarter. Low-volume variants are noisy. If you aren't sure, the cheaper mistake is usually to move the variant to its own keyword rather than block it. Its own keyword keeps it visible and measurable. A negative makes it disappear.

Negatives don't get close variants

This is the part most people get wrong. Your keywords expand to close variants. Your negative keywords don't. Google documents this directly: a negative doesn't match plurals, misspellings or synonyms of itself.

So if you add "roof replacement" as a negative to stop it leaking in through [roof repair], here's what still gets through:

• "roof replacements" (plural)
• "roof replacment" (misspelling)
• "replace roof" (reordered and stemmed)
• "new roof installation" (same meaning)

Your positive keyword can still match all four through close variants. Your negative blocks none of them. The leak looks fixed in the report for a few days, then the same intent comes back under a slightly different spelling.

What to do about it:

Choose the match type on purpose. There are three negative match types: broad, phrase and exact. An exact negative blocks only that exact query, so it's precise and narrow. A phrase negative such as "roof replacement" blocks any query containing those words in that order, so it also catches "cost of roof replacement" and "roof replacement near me". A broad negative blocks queries that contain all its words in any order. That sounds tidy, but broad negatives behave unpredictably across match types. Test one before relying on it.

Write the variants yourself. For the roof example, a sensible set might be phrase negatives for "roof replacement", "roof replacements", "replace roof", "replacing roof" and "new roof". Check each against your converting search terms first. "new roof" might also block "new roof leak repair", which you want.

Go back after a week. The search terms report will show you the forms you missed. That's normal. Building a good negative list is a few rounds of adding, checking and adding again.

This asymmetry is the strongest argument for checking search terms on a schedule instead of once. The same intent keeps coming back in new spellings. How often should you check search terms in Google Ads covers how often to look, depending on your spend.

Finding intent-shifting variants in the report

The search terms report already labels close variants. Open it at campaign or ad group level and look at the "Match type" column. Searches that matched through close variants show as "Exact match (close variant)" or "Phrase match (close variant)".

A quick routine:

1. Set the date range to the last 30 to 90 days, so low-volume variants have time to show conversions.
2. Filter Match type to the two close variant values.
3. Sort by cost, highest first.
4. For each term with real spend, read it as if you were the searcher. Ask the three questions above: can they buy, do they sometimes buy, are they in the right place?
5. Sort the terms into three groups: leave, negate, move to its own keyword.

Expect to leave most terms alone. If you end up negating half the list, you're probably blocking typos and plurals, which is the script mistake again.

Two limits to know about. First, Google doesn't show every search term. Low-volume queries are left out of the report, so some of your close-variant spend will never appear by name. You can't negate what you can't see. That's a reason to fix the pattern with a phrase negative, not just the one query you happened to spot. Second, the match type column tells you how a term matched, not why Google thought it was close enough. The column's limits are covered in search term match source in Google Ads, and its limits.

Also look at where close-variant waste piles up. It tends to sit under your broadest-meaning exact keywords: short, generic ones like [plumber], [accountant] or [glasses], where "same meaning" has the most room. Long, specific keywords leave Google less room to drift. If one short exact keyword keeps leaking, make it longer, or add more specific exact keywords next to it. That can do more than any list of negatives.

Shopping and Performance Max: no keywords, same problem

Shopping and Performance Max have no keywords, so in the strict sense they have no close variants. Google matches searches to your products, feed and assets instead. But the same intent shift happens, often more widely. A Shopping campaign selling reading glasses can show for "magnifying glass" because the product titles share a word. Nothing in a keyword list stopped it, because there isn't one.

Search terms for Search, Shopping and Performance Max are all reported, in the API view called campaign_search_term_view, so you can review them the same way. Display, Video and App campaigns report no search terms, so there's nothing to review or negate there.

Negatives work differently on each:

• Standard Shopping accepts negative keywords and shared negative lists, just like Search. The catch is that there's no positive keyword to restructure, so the "move it to its own keyword" option doesn't exist. Your levers are negatives, product titles and campaign priority. More in Shopping campaign negative keywords work differently.
• Performance Max accepts up to 10,000 negative keywords at campaign level, but no shared negative lists. If you run several PMax campaigns, each one needs its own copy of your negatives. What the PMax report shows you, and what it leaves out, is covered in what the PMax search terms report actually shows.

The "negatives don't get close variants" rule matters even more here. In Search, you have a keyword to tighten. In Shopping and PMax, the negative is often your only tool, so writing out plurals and misspellings yourself isn't optional.

Where PPCleaner fits, and where it doesn't

PPCleaner reads the search terms your account paid for every night, across Search, Shopping and Performance Max. It judges whether each one could ever convert for your business and adds the ones that couldn't as negative keywords. That covers the first group above: "roof replacement" for a repair-only business, "reading glass" for an optician, the abbreviation that means another trade. It reads queries by intent, not by comparing strings, so it leaves "emergancy plumber" and "running shoe" alone. It never blocks a term that converted in the last 90 days, and every block can be undone in one click for 90 days. If a judgement turns out wrong, undoing a negative keyword takes seconds.

Here's what it won't do. PPCleaner never touches your keywords, bids, budgets or ads. So the second and third groups — the mixed variants that deserve their own exact keyword, and the right searchers landing in the wrong ad group — are still yours to handle. That's on purpose. Adding [dog training] as a keyword with a lower bid and its own landing page is a strategy decision about your business. An automated negative tool shouldn't make it for you. PPCleaner also can't see search terms that Google leaves out of the report, so the hidden low-volume share stays hidden from it, just as it does from you.

The split is simple. Let automation handle the clear "never" cases every night, because they keep coming back in new spellings and negatives don't catch variants on their own. Spend your own review time on the "sometimes" cases, where a keyword change gets back money a negative would throw away. If you're deciding how much to trust exact and phrase match versus broad in the first place, are broad match keywords worth it in 2026 covers that.