Google Ads Search Terms Report, and What It Hides

What the search terms report actually contains

It is a list of the queries that triggered your ads, joined to what each one cost you. It is not a list of everything people searched, and it is not a list of your keywords.

Each row is a search term paired with the ad group and the keyword that matched it, plus the match type of that match. The same term can appear on several rows if it matched in more than one ad group — "emergency plumber near me" can sit under a broad keyword in one ad group and a phrase keyword in another, splitting its cost across two lines. If you sort by cost and only read the top rows, you will underestimate terms that fragment this way.

The match type column is worth reading properly. It tells you how the term matched, not what the term is. An exact match keyword [running shoes] producing a row for "sneakers for running" is Google's close variant system doing its job. That is not a bug, and no amount of tightening match types will stop it. Only a negative will.

In the API, all of this lives in campaign_search_term_view. That view covers Search, Shopping and Performance Max. Display, Video and App campaigns return no rows at all — there is no search term to report, because nobody searched. If half your account budget sits in a YouTube campaign, the search terms report has nothing to say about it, and no tool built on this data can help you there.

One detail that trips people up: the string in the report is not always exactly what someone typed. Google normalises case and, per its own documentation, groups very similar search terms into a single row. So the row you are reading may represent a small family of near-identical queries rather than one literal search.

What Google hides, and why it is the part you want

Google only reports search terms that were "searched by a significant number of users". That is Google's own wording. It has never published the threshold, and it is not a threshold on your account's data — it is about how many people searched the term.

The restriction landed in September 2020 and was presented as a privacy measure. Google restored some historical data a few months later and has since said the report covers more terms than it did immediately after the change. It has never gone back to being complete, and it will not.

Here is the structural problem. The terms Google hides are, by definition, the rare ones. Rare is exactly where the junk lives. "emergency plumber" is searched by millions of people and will always appear in your report. "do plumbers fix washing machine hoses on a sunday" is searched by a handful, will not appear, and is precisely the sort of query a broad match keyword picks up and charges you $6 for.

So the report is biased towards the traffic you already understand. The head terms are visible and mostly fine. The long tail is invisible and mostly where the waste is. Any advice that stops at "review your search terms report weekly and add negatives" is describing a review of the healthy half of your traffic.

Measuring your own blind spot

Do this once per account. It takes five minutes and it changes how seriously you take everything else.

1. Pick a date range with real volume — 30 or 90 days.
2. Open the Campaigns view, filter to Search campaigns, segment by network, and note the cost on the Google search network row.
3. Open the search terms report for the same campaigns and the same range. Download it. Sum the Cost column.
4. Subtract.

What is left is spend you paid for and cannot attribute to any named query. In the accounts we audit, that gap usually falls somewhere between a tenth and a third of Search cost. Accounts running broad match on long-tail products sit at the high end. A tight exact-match brand campaign sits near zero. Yours might be outside that range in either direction — the point is to know your own number rather than assume it is small.

If the gap is large, that is the first place to look before you start line-by-line negative work. The same arithmetic is one of the checks in how to find wasted ad spend in Google Ads, alongside the other places money leaks without showing up as an obviously bad query.

Two caveats so you do not chase ghosts. Small mismatches of a percent or two come from rounding and from currency conversion in cross-currency accounts. And if you do not segment by network, search partner traffic will muddy the comparison.

Widening the date range does not bring hidden terms back

This is the most common workaround and it does not work. The reasoning sounds right: a term with one click a day is invisible daily, so pull 90 days and it will cross the line.

But the threshold is described as being about the number of users who searched the term, not the number of clicks you received. Pulling a longer range gives you more rows, because more distinct qualifying terms appear over a longer period. It does not promote a rare query into visibility because you accumulated clicks on it. If a phrase is rare across Google, it stays hidden at 7 days and at 365 days.

The same goes for downloading via the API instead of the UI. Tools that read campaign_search_term_view — ours included — see exactly what Google chooses to return. No tool has a back door. Anyone selling you "see the search terms Google hides" is either guessing from other signals or lying.

Performance Max and Shopping report differently

Both appear in campaign_search_term_view, but what you see in the interface is not the same thing, and what you can do about it is not the same either.

Performance Max. The interface gives you search categories under Insights: buckets of related queries with relative volume, not individual terms with individual costs. Categories are useful for orientation and useless for action, because you cannot add a negative keyword for a category. You can only negate strings. The per-term rows exist in the API view, which is where any serious analysis of a Performance Max campaign has to happen.

Performance Max accepts up to 10,000 negative keywords at campaign level, and it does not accept shared negative keyword lists. If you manage twenty Performance Max campaigns and you want the same 400 exclusions on all of them, that is 8,000 individual campaign-level entries, maintained separately, forever. The cap and the practical consequences are covered in negative keywords in Performance Max and the 10,000 cap.

One honest limit while you are here: a negative keyword only governs the search side of a Performance Max campaign. It does nothing to the display and video placements inside the same campaign. If your Performance Max spend is mostly going to placements rather than queries, negatives are the wrong lever.

Standard Shopping. Search terms are reported, and Shopping accepts both campaign-level negatives and shared negative keyword lists, exactly like Search. That makes Shopping the easiest place to run a disciplined negative programme.

The gap in Shopping is different. There are no keywords, so the keyword column carries no information you can act on. And the report will not tell you which product the query triggered. You can see that "cheap running shoes" cost you $240 and converted once. You cannot see from this report that $200 of it went to one overpriced SKU that was never going to win on that query. Product-level data lives in a separate view, and there is no join between the two. If you need query-by-product, you have to build it from bid segmentation or custom labels, not from this report.

Blocking terms you cannot see

You cannot add a negative for a string you have never been shown. You can add a negative that catches the family it belongs to. That is the whole workaround, and it is why negative keyword strategy has to be thematic rather than literal.

Treat the visible terms as a sample. If "plumber salary uk", "plumber apprenticeship wages" and "how much do plumbers earn" made it into your report, the hidden tail almost certainly contains fifty more phrasings of the same idea. A phrase negative on salary and another on how much do blocks all of them, seen and unseen, in one move.

The practical method is n-gram thinking. Take your visible search terms, break them into single words and two-word pairs, and total the cost against each fragment. You are looking for words that show up repeatedly in terms that never convert:

• jobs, salary, apprenticeship, career — job seekers, not customers
• free, diy, how to, yourself — people who intend to not pay you
• wikipedia, reddit, youtube, pdf, meaning — research, not intent
• used, second hand, rental if you only sell new
• competitor brand names you have decided not to bid on

Each of those, as a phrase negative, reaches into the part of the report you cannot read.

Now the cost of doing it. Broad, phrase and exact are your three negative match types, and broad negatives behave unpredictably across match types — the same broad negative does not always block what you expect once close variants are involved. Phrase is the safest instrument for thematic blocking. Exact is the safest for one-off surgical removals.

And thematic negatives overblock. If you sell software, a phrase negative on free kills "free trial project management software", which for many accounts is the single best-converting query in the account. If you sell flights, cheap is not junk, it is your market. Before you add a word, search your own report for it and look at what converted. Two minutes of checking prevents a month of silent lost revenue, and the loss is silent — a blocked query produces no impressions, no clicks and no line in any report telling you what you gave up.

That is the argument for reversibility. Any blocking system, manual or automated, should let you undo a decision without archaeology. PPCleaner keeps every block undoable for 90 days for exactly this reason, and it never blocks a term that converted in the last 90 days — the cheapest way to avoid the free trial mistake is to make it structurally impossible.

Reading the report without over-reacting

The report tempts you into two opposite errors: blocking on noise, and refusing to block anything without proof.

A term with 4 clicks, $11 spent and zero conversions is not evidence of anything. If your account converts at 5%, a term that never converts in 4 clicks is the most ordinary thing in the world. Blocking it is a coin flip dressed up as optimisation. Do this a hundred times and you have amputated a chunk of your long tail for no reason.

So ask two different questions of every term, and keep them separate.

Is it semantically wrong? Could a person searching this ever buy what I sell? "plumber salary" cannot. One click is enough. Block it and never think about it again. This judgement needs no data at all and it is the reason a human eye still beats pure statistics on the head terms.

Is it statistically dead? The term is plausible — someone searching it might buy — but it has now had 40 clicks and $180 with nothing to show. This judgement needs volume, and how much volume depends on your conversion rate and your average order value. A term needs to have had a real chance before you can say it failed.

Two things distort the second judgement. Conversion lag: if your typical path takes three weeks, a 14-day window will show zero conversions on terms that convert fine. Always look at a window longer than your lag. And seasonality: "christmas hamper delivery" looks dead in June.

The conversion action you count matters too. If your primary conversion is a form fill and half your revenue arrives by phone, the report is showing you a partial picture of value and you will block terms that were paying for themselves offline.

A routine that survives contact with a real workload

Weekly, twenty minutes, per account:

1. Search terms report, last 30 days, sorted by cost descending, filtered to conversions = 0. Read the top 40 rows. Apply the semantic test only — is this person a possible customer, yes or no. Block the noes as exact or phrase negatives.
2. Once a month, run the n-gram pass on 90 days of data and look for recurring junk words rather than junk terms. This is the step that reaches the hidden tail, and it is the step everyone skips.
3. Once a quarter, redo the cost-gap arithmetic. If your unattributable share is growing, your match types are loosening, usually because someone turned on broad match or a Performance Max campaign started eating search traffic.
4. Once a quarter, look at what you have blocked, not just what you might block. Negative lists rot. A phrase negative added two years ago for a product line you no longer sell is now blocking traffic for a product line you do.

That routine is entirely doable by hand on one account. It stops being doable somewhere around the tenth account, which is where a nightly automated pass earns its place — the arithmetic does not get harder, there is just more of it than a person will actually do on a Thursday afternoon.

Whichever way you do it, the ceiling is the same. The search terms report is the only view Google gives you of what your money bought, it is deliberately incomplete, and no tool sees more of it than the API returns. Work the visible terms as a sample of the invisible ones, and measure the gap so you know how big the sample is.