Why do the numbers in my search terms report not match my campaign totals?

Because Google withholds search terms that were not searched by enough different people, for user privacy. Those impressions and clicks still happened and you were still charged for them, but they never appear as a row. The gap between the sum of the visible rows and the campaign total is the hidden portion. It is normal, it varies a lot by account, and it means the report is a large sample rather than a complete ledger.

Do negative keywords match close variants like misspellings and plurals?

No. This is the detail that catches almost everyone. A positive exact keyword matches close variants, so [running shoes] can serve on “running shoe” or a misspelling of it. Negative keywords do not work that way: negative exact, negative phrase and negative broad all match only the words you actually typed. Singulars, plurals, misspellings, spacing variants and accented forms each have to be added as their own negative.

Should I put a negative on the campaign or in a shared negative keyword list?

Use a shared negative keyword list for exclusions that are true everywhere in the account, such as job-seeker language or free-and-DIY intent, because one edit then updates every campaign the list is attached to. Use campaign-level negatives when the exclusion is only right for that campaign, for example blocking a product line that another campaign is deliberately bidding on. Note that shared lists attach to campaigns, not to ad groups.

How many clicks should a search term get before I add it as a negative?

There is no universal number, but one click is never enough. A reasonable rule is to wait until the term has spent somewhere near your target cost per acquisition with no conversion, and to check a date range at least as long as your typical conversion lag before deciding. The exception is intent that can never convert regardless of volume, such as searches containing “jobs” or “salary” for a software product. Those you can block on sight.

Can I get a search terms report for Performance Max?

Not in the same form. Performance Max reports search themes and search categories rather than a full list of individual queries, so the level of detail is much lower than a Search campaign gives you. You can still apply negative keywords to Performance Max campaigns, and account-level negative keywords apply across the account, so the exclusion side of the workflow still works even though the diagnosis side is coarser.

How to read the Google Ads search terms report and turn it into negative keywords

Your keyword list is what you told Google you wanted to buy. The search terms report is what you actually bought. Those two things are never the same list, and the distance between them is where most wasted Google Ads spend lives. Everything useful about traffic quality starts with reading that report properly and ends with a negative keyword written at the right level, in the right match type, with the right words in it.

This is the whole workflow: where the report is, what it does not show you, how to read it without over-reacting, and how to turn a wasted row into a negative that blocks the family of searches you meant and leaves the ones you wanted alone.

Where the search terms report lives

In the current Google Ads interface the report sits under Campaigns, then Insights and reports, then Search terms. Older layouts put it under the Keywords section instead, and you can still reach the same data from a keyword's row menu, or by segmenting the Keywords view. The navigation has moved several times; the report has not changed.

It is worth being precise about what it contains, because the two reports next to each other get conflated constantly:

The keywords report lists the keywords you added, with the match types you chose. It is your side of the arrangement.
The search terms report lists the queries real people typed that triggered one of your ads, along with which keyword caught each one. It is Google's side.

Search campaigns give you the full version of this report. Shopping campaigns report search terms too, though without keywords to attribute them to. Performance Max does not: it reports search themes and categories, which are useful for direction but will not hand you a query to exclude.

Why the report is never the whole truth

Two things stop the search terms report being a complete record, and both change how you should read it.

Low-volume queries are withheld. Google only shows search terms that were searched by a significant number of users. Terms below that threshold are omitted for user privacy even though they served impressions, took clicks and cost you money. The practical consequence is that if you add up the cost of every row in the search terms report, it will come to less than the campaign total for the same period. That is not a bug and it is not something you can appeal. It does mean two things: the long tail you cannot see is real, and broad thematic negatives do work the hidden portion in a way that exact-match negatives on individual queries never will.

Close variants mean a keyword catches more than its text. Exact match has not meant literally-this-string for years. An exact keyword now matches close variants: misspellings, singular and plural forms, stemmings, abbreviations, accents, added or removed function words, reordering where the meaning survives, and paraphrases with the same intent. Phrase match covers queries that include the meaning of the keyword. Broad match goes furthest, drawing on the landing page, the other keywords in the ad group, the user's recent activity and the account's history, which is why broad match is both the biggest source of discovery and the biggest source of the rows you are about to exclude.

Put together: you cannot predict your traffic from your keyword list, so the report is not optional. It is the only feedback loop that exists.

Reading the report: which columns, which sorts

Start with the date range, because the default is usually too short to judge anything. For most accounts, thirty to ninety days gives enough volume for a single search term to say something. If the account has a long consideration cycle, use a range longer than your typical conversion lag or you will exclude terms that were going to convert next week.

Add these columns before you sort anything:

Cost, clicks and impressions, so you can tell expensive from merely noisy.
Conversions, conversion rate, cost per conversion and conversion value, so a term is judged on outcome rather than on whether you personally like the wording.
Match type, which tells you which keyword caught the term and how loosely, and is often the real answer to why a strange query appeared.
The added or excluded status, so you do not spend an afternoon proposing negatives that are already in place.

Then make three passes, in this order:

Sort by cost descending, filter to zero conversions. This is the shortlist of terms that are provably taking money and provably returning nothing. Almost every real win is in the first screen of this list.
Look for high impressions with a very low click-through rate. These are searches Google believes are relevant to you and users, seeing your ad, disagree. They cost little directly but they are a signal that a keyword is pulling in the wrong theme, and they drag the account-level quality signals with them.
Search the report for intent words that can never apply to you. Words like jobs, salary, career, free, DIY, template, course, tutorial, wikipedia, second hand, or the name of a product you do not sell. Search within the report rather than scrolling, because these are usually spread thinly across hundreds of low-cost rows that no sort will surface.

From a wasted search term to the right negative keyword

The instinct on finding a bad row is to exclude that exact query. Resist it. Excluding the query blocks one string out of a family of thousands, most of which are in the hidden portion of the report where you will never see them. It is a game you cannot win by playing it faster.

Instead, ask which word made the search worthless. Suppose you sell accounting software and the report shows accounting software jobs, accounting software salary uk and accounting software career path. The problem is not those three strings. The problem is that job-seekers and your customers use the same product noun, and the words that separate them are jobs, salary and career. Three thematic negatives retire the entire family, including everything below the visibility threshold. Three exact-match negatives on the full queries retire three rows.

Before adding a thematic negative, run it against your own keyword list. Does any keyword you are actively bidding on contain that word? Google's recommendations will flag a negative that completely blocks one of your keywords, but not one that merely narrows it, and narrowing is the more common and more expensive mistake. A retailer negating cheap because of one bad query can quietly kill an entire price-led ad group that was working.

Themes that are worth building deliberately, once, per account:

Employment — jobs, salary, hiring, vacancy, career, recruitment, internship.
Free and DIY — free, freeware, crack, torrent, template, spreadsheet, open source. Add these carefully if you run a free trial, since free is then a word your buyers also use.
Informational intent — what is, definition, meaning, wikipedia, how does, tutorial, course, certification.
Wrong product or wrong market — adjacent products you do not sell, sizes you do not stock, countries you do not ship to.
Competitor names — a genuine decision rather than a default. Bidding on competitors can be the best traffic in the account or the worst, and it should be one or the other on purpose.

On evidence: a single click is not a pattern. Wait until a term has spent something approaching your target cost per acquisition without converting, or until the intent is unambiguous regardless of volume. Job-seeker language does not need to prove itself over thirty clicks.

Negative match types, and the close-variant trap

Negatives use the same notation as positive keywords and completely different rules. No punctuation means broad, quotation marks mean phrase, square brackets mean exact.

Negative broad — running shoes — blocks a search only when it contains all of the words, in any order, with any other words present. A search for just shoes is not blocked, because it does not contain every word.
Negative phrase — "running shoes" — blocks searches containing those words in that order. Extra words either side are fine; best running shoes for flat feet is blocked, shoes for running is not, because the order changed.
Negative exact — [running shoes] — blocks only the search that is exactly those words in that order and nothing else. Any additional word and the ad can serve.

Now the detail that quietly undermines more negative keyword lists than anything else: negative keywords do not match close variants. Not negative exact, not negative phrase, and — the one people are most surprised by — not negative broad either. The close-variant machinery that makes your positive keywords generous is simply not applied to your exclusions. A negative broad of running shoes does not block running shoe, does not block runing shoes, and does not block a synonym or a related phrase that means the same thing. Only the exact words you typed count.

That has three practical consequences. Singular and plural forms both need adding. Common misspellings need adding, and the ones that matter are visible right there in the search terms report. And where a theme can be captured by one word rather than two, one word is stronger — a negative broad of jobs blocks every query containing that word, while accounting jobs blocks only the subset containing both, and neither will do anything about job singular unless you add it.

Two smaller rules worth knowing: negative keywords do not support search operators such as site:, and a negative that duplicates something already blocked at a higher level is harmless but adds to the pile you will have to audit later.

Where to put the negative: list, campaign or ad group

Google Ads gives you several places to attach a negative keyword, and choosing wrongly is how an account ends up with the same exclusion in forty campaigns and none of them quite in sync.

Shared negative keyword lists, built in the shared library and attached to campaigns. This is where universal exclusions belong. One edit propagates to every attached campaign, which is the entire point. Lists attach to campaigns only, not to ad groups.
Campaign-level negatives apply to every ad group in that one campaign. Use them when the exclusion is right here and wrong elsewhere — blocking a product line that a different campaign is deliberately targeting, for instance.
Ad-group-level negatives are the narrowest, and their main legitimate use is routing traffic between ad groups in the same campaign so that a query lands on the ad group with the right ad copy and landing page. Their main illegitimate use is as the default place to put everything, which produces exclusions that only work in one ad group while the query happily serves from the one next door.
Account-level negative keywords cover the whole account, including campaign types where the reporting is thin. Reserve these for terms that must never serve anywhere under any circumstances.

Google enforces caps on how many shared lists an account can hold and how many negatives each list and campaign can carry. They are generous enough that most accounts never notice, but check the current limits before designing a structure that assumes hundreds of lists.

A routine that does not strangle good traffic

Weekly for accounts spending seriously, monthly for smaller ones, and always after a significant change to keywords, budgets or landing pages, because those change what broad match reaches for.

Open the search terms report over a range long enough to be meaningful.
Sort by cost, filter to zero conversions, and read the top of the list.
For each bad row, identify the word that made it bad, not the whole string.
Check that word against your active keywords before excluding it.
Choose the match type on purpose, then add the singular, the plural and the obvious misspellings, because none of those come free.
Choose the level on purpose: shared list for universal, campaign for local, ad group only for routing.
Write down what you added and why, with a date.

That last step is not bookkeeping for its own sake. Negative keywords are the only part of a Google Ads account whose effect is invisible: there is no report of the searches a negative prevented, no line item for the customers it turned away. If traffic falls off a month later, the log of what was excluded and why is the only way to find the cause. Over-negation is a silent failure mode, and the accounts that suffer from it are usually the ones that were being diligent.

The report itself is straightforward. The discipline is in reading it often enough to catch waste early, thinking in themes rather than strings, remembering that your exclusions are far more literal than your keywords, and keeping a record you can reverse.