Google Ads Search Terms Report: What Google Hides, and Where It Puts It

Google withholds low-volume search terms from the report you read, then accounts for them in a second surface. Both statements are Google’s own, read signed out on 19 August 2026.

Search term and keyword are not the same thing

Google's definitions, word for word: "A search term is a word or set of words a person enters when searching on Google or one of our Search Network sites. A keyword is a word or set of words that Google Ads advertisers can add to a given ad group so that your ads are targeting the right audience."

One is what a person typed. The other is what you bought. The report exists because those two drift apart, and the drift is where the money goes.

What Google leaves out, in Google's words

The report shows terms "that a significant number of people have used, and that resulted in your ad being shown". The exclusion is stated once, plainly:

That sentence is on the help page, not on the report. Google does say it in more than one place, and the API documentation is blunter than the interface: "Clicks from these 'hidden' terms are included in the campaign total but are excluded from the search term reports." What no surface gives you is a number: not a count of withheld terms, not a share of clicks, not a percentage of spend.

The threshold has a history worth knowing, because it explains why an old audit no longer reproduces:

• September 2020. Google "updated the search terms report and Dynamic Search Ads search terms report to meet new privacy thresholds", which "ensure user anonymity by only reporting on terms that have seen sufficient search volume across all Google searches". Those last four words matter: the threshold is about how often a query is typed on Google, not about how often it appears in your account.
• September 2021. Google loosened it, and backdated the result: "Starting today, we're able to show you more queries that meet our privacy standards", with the data "shown in your reports for queries you received starting from February 1, 2021". The announcement is September, the data goes back to February.
• February 1, 2022. The older data went away: query data collected before September 2020 that does not meet the current thresholds "will be available in your reporting until February 1, 2022".

So the report you are reading today is the third version of this dataset in six years, and the middle version is the one most agency playbooks were written against.

The bucket exists. The money is not in it.

This is the part I got wrong when I first started looking, and it is the part most write-ups get wrong too, so it is worth being exact.

The withheld terms are not thrown away. Directly after the privacy sentence, the same help page says: "Search terms insights account for these low volume queries by grouping them into their relevant subthemes where applicable, or aggregating them as 'other queries', without exposing the queries themselves." The insights surface is explicit that it covers everything: "These groupings take all search terms into account, including those not exposed in the search terms report due to privacy reasons", and it names the catch-all: "Search terms that have lower spend or don't have an identifiable search category will be grouped together as 'other search terms'."

So there is a residual bucket. It is in a different report from the one you were reading, it shows categories rather than queries, and it exists.

Now the part that is true and that decides everything downstream. The metrics search terms insights reports for each category are clicks, impressions, click through rate, conversions, conversion rate, conversion value and search volume. Cost is not among them. You can see how many clicks arrived through terms you are not allowed to read. You cannot see what they cost.

That is why the two numbers still refuse to meet. The term list is short of the campaign by an amount that is documented in clicks and undocumented in money, and money is the column every report is reconciled on.

The search terms report itself behaves the same way in the API: search_term_view returns one row per reportable term and, in a thirty day pull, no row labelled other, unspecified or blank. The aggregate lives next door, in campaign_search_term_insight, where the catch-all category is an empty label rather than a name.

Two more things the same pull makes visible, both worth knowing before you read a term list as if it were literal:

• Match type is doing more work than the column suggests. The values that came back include NEAR_EXACT and NEAR_PHRASE, the close variant forms, alongside EXACT, PHRASE and BROAD. Exact and phrase are what they sound like: Google's own example table shows the term containing the keyword. Close variants add misspellings and plurals. Broad is the one that puts a term in front of a keyword with no words in common, and broad is what most accounts now run.
• Status says what you did, not what the term is worth. The values are added, excluded, both at once, or none. The column tells you whether you have already acted on a term. It does not tell you whether the term is representative of anything.

Thirty five Google reports, three that mention search terms

For the previous piece in this cluster I opened every Google Ads report Google publishes in its own public gallery. I went back through the same set with one question: how many of them show search terms at all, and does any of them mention the threshold.

Google publishes thirty five Google Ads reports there. Thirty four of them rendered for me on 19 August 2026; the thirty fifth is an older report whose link I could not resolve in this pass, so treat every count below as out of thirty four. Three mention search terms or queries anywhere in the rendered report. None mentions the threshold, the omission, or the word privacy.

One of the three is called Google Ads Search Query Performance. It is a whole report built on this dataset, published by Google, and it carries no note that the dataset is partial by design.

I am not claiming the note belongs on every chart. I am claiming that if the vendor who applies the threshold does not mention it on the report built from the thresholded data, no one downstream will invent the disclosure on their own.

And Google can be plainer than this when it wants to be. Its own page on placement reports describes the identical mechanic in one paragraph: reports "often show lower totals than the main Google Ads interface because they exclude placements with low activity to protect user privacy", the leftovers "are consolidated into the 'Total: other' category", and for verification you should "refer to standard campaign reports rather than granular placement reports". Threshold, residual bucket, do not reconcile: three sentences. The search terms documentation contains the same three facts spread across four pages.

Where to find it, and what the columns are worth

In Google Ads the report sits under Campaigns, then Insights and reports, then Search terms. You can open it for the whole account or for a selection, and the columns are the ordinary performance ones: impressions, clicks, cost, conversions, plus the keyword the term was matched to and its match type.

Three of those columns are worth more than the rest, and one is worth less than people think.

The term itself is the only column that carries information you cannot get anywhere else. Everything else in Google Ads is a number about something you chose; this is a number about something a stranger typed.

The matched keyword tells you which part of your account is buying this traffic. In an account with overlapping ad groups it is often the surprise: the term is fine, the ad group serving it is not.

Match type is the one to read before acting, for the close variant reason above.

Conversion rate on a single term is the one worth less than it looks. Most terms in a normal account carry a handful of clicks in a month, and a rate computed on four clicks is not a rate. Aggregate terms into themes before you judge them, or judge only the ones with real volume.

The n-gram habit, and why it survives the threshold

Because individual terms are thin and partially withheld, the durable technique is to stop reading them one at a time. Split every visible term into its words and pairs of words, then total cost and conversions by word rather than by term.

This is not a trick to recover missing rows. Nothing recovers those. It works because the words you care about, the ones signalling the wrong intent, appear across many small terms, and a word that shows up in thirty withheld searches will usually also show up in two visible ones. The threshold hides individual queries; it does not hide vocabulary.

A word-level view also survives the thing that makes term-level analysis fragile: the same intent arriving in twenty spellings. Two rows are usually enough to see a pattern that twenty rows would have spelled out.

The volume rule from the previous section still applies, one level up. Do not act on a word with four clicks behind it either. Set a floor before you start, thirty clicks or a month of spend you would notice, and treat everything under it as a hypothesis. A pivot table does this; so does any spreadsheet with a split formula. No tool is required and none of the paid ones do it better.

The nearest relative is auction insights, which is censored in a different way: it exists, it names competitor domains, and it cannot leave Google Ads. And whatever the term list does say, the conversion numbers beside it are still settling for weeks after you read them.

Google does this in more than one report

The search terms report is not a special case. Store visits have a privacy threshold. Modeled conversions have one, under a heading Google calls rigorous thresholds for reporting. Auction insights refuses to show a row below ten per cent impression share. Placement reports drop low activity placements and bucket them as Total: other.

What makes the search terms report worth a whole article is not that it is unique. It is that it is the report people build negative keyword decisions on, and the one where the disclosure is furthest from the data.

Performance Max does not have this report at all

Everything above is about Search and Dynamic Search Ads. Performance Max does not give you search terms. It gives search categories, grouped, without the keyword column, on its own insights surface. So in a modern account the question is not only how much of the term list is withheld, it is how much of the spend never had a term list in the first place.

If a meaningful share of your budget sits in Performance Max, read the search terms report as a report about part of the account, and check what fraction of spend it can speak for before you draw conclusions from it.

What this does to negative keywords

The standard advice is Google's own: when "a search term isn't relevant enough to the products or services you offer", you "add it as a negative keyword".

That advice is sound and it has a ceiling built into it. You can only exclude what you can see, and the terms below the line are the rare ones.

Resist the obvious next step, which I nearly took. It is tempting to say small accounts suffer most, and it does not follow: the threshold counts how often a query is typed across all of Google, not inside your account. Your exposure depends on how long tailed your traffic is. An account bidding exact match on head terms can see nearly everything it paid for. An account on broad match reaching for the tail can be missing a great deal, at any budget.

One more caveat, because the advice is older than the bidding. On target CPA or target ROAS with broad match, negatives are also training data: excluding every term that underperformed in a small sample narrows what the bidder is allowed to explore. Negate for irrelevance and brand safety. Leave the marginal ones alone unless you have volume behind the judgement.

The honest version of the claim is this. The search terms report is a list of what was popular enough to show you, and that is not the same as a list of what you paid for.

What my own template does with this

Nothing, and that is worth saying plainly rather than leaving you to discover it.

Google Ads Vitals reads a campaign level export: date, campaign, impressions, clicks, cost, conversions, conversion value. Five pages come out of that, covering cost per lead, ROAS and revenue by campaign against targets you type in. There is no search terms page in it, because search terms are not in that export.

So this article does not describe a feature I sell. If coverage, the ratio in the section above, turns out to be the number people want on a screen, that is a reason to add a terms page and put the ratio on it. Until then the honest position is that my dashboard has the same blind spot as the rest, arrived at from the other direction.

Sources and method

The API observation used the Google Ads API on 19 August 2026 against one ordinary advertiser account, and reports only structure: which resources answer, which enum values appear, and the absence of a residual row. No account figures appear in this article, here or anywhere else on this site.

The gallery count was made by opening each report Google publishes in its public gallery, signed out, on 19 August 2026, and reading the rendered text of each. A note on the address: the two domains currently point at each other. On the day I read them, lookerstudio.google.com returned a 301 to datastudio.google.com and datastudio.google.com returned a 302 back to lookerstudio.google.com/overview, so either name reaches the gallery and neither is safe to call the current one. Thirty four of the thirty five rendered; three of those mention search terms or queries, and none mentions the threshold.

• Google Ads Help, About the search terms report: the definitions of search term and keyword, the significant number of people wording, the privacy omission sentence, and the negative keyword advice.
• Google Ads Help, Improving the search terms report while maintaining user privacy: the September 2020 threshold, the February 2021 expansion, and the February 2022 end of pre-2020 historical data.
• Google Ads Help, About search terms insights: the statement that the groupings "take all search terms into account, including those not exposed in the search terms report due to privacy reasons", the "other search terms" catch-all, and the metric list that does not contain cost.
• Google Ads Help, About placement reports: the threshold, the "Total: other" bucket, and the instruction to verify totals against campaign reports rather than granular ones.
• Google Ads API documentation: "Clicks from these 'hidden' terms are included in the campaign total but are excluded from the search term reports."
• Google Ads API, live query, 19 August 2026: the search_term_view resource, the match type enum values including NEAR_EXACT and NEAR_PHRASE, and the absence of a residual row inside that resource.
• Google's public report gallery, read signed out on 19 August 2026: thirty five published Google Ads reports, thirty four of which rendered, including Google Ads Search Query Performance.

Questions people actually ask

WRITTEN BY
Olha · clinic data analyst

I build the reporting our managers open every morning at a multi-branch medical clinic — and package it so other practices don't have to start from scratch.

Published on 19 August 2026. One limit is worth stating outside the body text. The API observation shows that one account, on one pull, received no residual row and no other bucket; it is evidence about the shape of the response, not proof that Google never reports a remainder anywhere. The gallery count is a count of what the rendered reports say, read signed out against sample data, and a report can of course carry a caveat in a place a text read does not reach.

35ARTICLES WRITTEN