Everyone repeats the same statistic. Google hides about 40% of search terms. It comes from a study that is years old, and nobody has repeated it.

So we measured it against live accounts this year.

Across 11 advertiser accounts over 40 days, a median of 51.5% of search campaign spend could not be traced to any individual search term. Pooled across all of them, 49.9%.

The 40% figure understates the problem.

What we compared

Two numbers over the same 40 days, for the same campaigns.

The first is total spend on standard search campaigns. The second is the spend that appears attached to individual search terms in the search terms report.

Whatever is left over is money you cannot examine. There is no query attached to it, so there is nothing to judge, nothing to exclude, and nothing to learn from.

Performance Max, Shopping, Display and Demand Gen spend were all separated out. Performance Max is reported on its own further down, because it behaves differently.

The numbers

Hidden share of search spend	Accounts
Under 10%	1
10% to 25%	0
25% to 40%	2
40% to 55%	4
55% to 70%	2
Over 70%	2

Median 51.5%. Range 5.1% to 82.5%. Pooled 49.9%.

One account could trace nearly all of its spend to real queries. Two could trace less than a fifth.

The result holds when you remove the largest spender. Pooled drops to 45.8% and the median does not move. It is not one account carrying the number.

Size has nothing to do with it

We expected bigger accounts to see more, since more volume should clear Google’s reporting threshold more often.

There is no relationship. The correlation between account spend and hidden share is effectively zero. The 55% to 70% band alone spans accounts from tiny to five figures in 40 days.

One pattern did show up, and it is a hypothesis rather than a finding. The account with the best visibility ran 14 search campaigns. One of the worst ran 6. If the reporting threshold applies per campaign, more campaigns splitting the same spend might surface more terms. Eleven accounts is not enough to say that with any confidence.

Performance Max is worse

Performance Max search term reporting arrived for everyone, and it is a real improvement over the old category groupings. It is still the least transparent thing in the account.

Across the four accounts running it, roughly 76% of Performance Max spend could not be traced to an individual query. That is about 26 points worse than standard search in the same accounts, over the same days.

Treat that number as directional. Four accounts, and one of them accounts for most of the Performance Max spend in the sample. The direction is clear. The precision is not.

What this number is, precisely

This matters, so here is the exact claim.

We are not reading a figure Google publishes. Google shows an aggregated row for withheld terms, and we are not counting that row. We are subtracting what can be traced from what was spent.

So the gap is mostly the privacy threshold, and it may also absorb attribution lag or other edge cases where spend never lands on a query. It is the amount you cannot examine, which is not quite the same as the amount Google is deliberately withholding.

We prefer that framing because it is the one you can act on. Whatever the cause, that spend is not reviewable.

Two other limits worth stating. Forty days is a short window, bounded by when we started storing this data daily. And these are one agency’s accounts, weighted toward business to business and industrial advertisers, not a random sample of the internet.

Why this changes the job

Put this next to what we found last time. In the visible portion of the report, a median of 6.6% of spend went to queries we could prove were irrelevant, and 86% of that waste sat in terms with a single click.

Now add that roughly half the spend has no query attached at all.

Three things follow.

Your waste number is understated, and so is everyone else’s. Any figure derived from the search terms report describes half the account. Including ours.

Negative keywords cannot reach what you cannot see. You can only exclude a query you were shown. The rest keeps matching on terms nobody has read.

Account structure may be doing more than you think. If visibility really does vary with how spend is split across campaigns, then consolidating everything into one campaign has a cost nobody prices in.

None of this makes the search terms report less useful. It makes it the only window you have, which is an argument for opening it more often rather than less.

FAQ

How much of the Google Ads search terms report is hidden?

In our measurement across 11 live accounts over 40 days, a median of 51.5% of search campaign spend could not be traced to any individual search term. Pooled across accounts, 49.9%. The commonly cited 40% figure appears to understate it.

Why does Google hide search terms?

Google withholds queries that fall below a reporting threshold, which it describes as a privacy measure protecting people who search for something unusual enough to be identifying. The practical effect is that low volume queries, which is most of them, never appear individually.

Can I see search terms in Performance Max?

Yes. Performance Max search term reporting is available to all advertisers, with data going back to March 2023, and self-serve negative keywords came with it. Visibility is still worse than standard search. In our sample roughly 76% of Performance Max spend could not be traced to a specific query.

Does spending more reveal more search terms?

Not in our data. The correlation between account spend and hidden share was effectively zero. Large accounts and small accounts had similar visibility, and the best performing account in our sample was mid-size.

If half the data is hidden, is the search terms report still worth reviewing?

Yes. It is the only direct view of what you are actually paying for. The right conclusion is to review the visible half more often, not to treat the report as unreliable and skip it.

---

PPC Sherpa reads the visible half of your search terms report for you, every week, and tells you what it costs. You approve what gets excluded.

See plans

Every article about the search terms report tells you the same four things. Sort by cost. Filter for zero conversions. Add negatives. Review weekly.

Nobody tells you how much money is actually sitting there. So we measured it.

We ran automated audits across 12 live advertiser accounts over a seven week window, flagged every search term that was provably irrelevant, and added up what those terms cost. Here is what we found, including two things that contradict the standard advice.

The number

A median of 6.6% of search spend went to queries flagged as irrelevant.

The range was 0.14% to 19.15%.

That range matters more than the median. Here is how the 12 accounts distributed:

Wasted spend	Accounts
Under 1%	2
1% to 3%	2
3% to 5%	1
5% to 10%	5
10% to 15%	0
Over 15%	2

There is no typical account. There is a cluster around 5% to 10%, two accounts that are essentially clean, and two accounts hemorrhaging close to a fifth of their search budget.

The single largest advertiser in the set spent over $100,000 in seven weeks and wasted 0.14% of it. Two mid-size accounts spending a tenth as much wasted nearly 19%. Size does not predict waste. Attention does.

Pooled across the whole group, total waste came to 2.94% of total spend, because that one large well-managed account dominates the denominator. Both figures are true. The 6.6% median describes a typical account. The 2.94% describes the group as a whole.

How we defined irrelevant

This is where most waste estimates fall apart, so it is worth being specific.

A search term counted as wasted only if an audit produced an exact match negative keyword recommendation for that exact query, and only the money actually observed on that term during the window was counted.

Exact match matters. A broad or phrase negative covers a theme, and a theme includes queries nobody has examined individually. An exact match negative is a judgment about one specific search that was actually paid for.

Terms a human reviewed and rejected were excluded entirely. If someone looked at a flagged query and decided it belonged, it is not waste.

No projections. No annualized figures. No estimates of what might be hiding. Just observed cost on terms proven irrelevant, measured against spend from the same days.

This number is a floor

Three reasons the real figure is higher:

Google hides a large share of search terms. Queries below its reporting threshold never appear, so their spend cannot be examined by anyone, including us.

Performance Max data is newer and thinner. Query reporting only goes back to March 2023, high-volume terms get sampled, and low-volume ones are withheld like anywhere else.

Irrelevant is not the same as unprofitable. A search term can be perfectly relevant and still lose money on a weak landing page. That is a different problem and it is not in this number.

So 6.6% is what we could prove. Not what is there.

Finding one: the money is not in the big leaks

The obvious way to work a search terms report is to sort by cost and start at the top. Every guide recommends it. It feels productive because the first row is always something expensive and stupid.

In our data, 85.8% of all measured waste came from search terms with one click or fewer.

The single most expensive irrelevant term across all 12 accounts cost $169.91 and had one click. Nine of the ten most expensive terms were single clicks.

That changes what the work actually is. Sorting by cost surfaces a handful of one-off clicks on expensive queries, which is worth doing and takes five minutes. Then the report looks handled and you close it.

Meanwhile the real money is spread across thousands of rows that individually cost a few dollars and never rank near the top of anything. No single row justifies the time it takes to evaluate it. Collectively they are the entire problem.

This is why the search terms report gets reviewed and the waste persists anyway. The reviewing method is matched to the wrong shape of problem.

Finding two: there is no universal negative keyword list

The second piece of standard advice is to build a master negative list. Free, jobs, DIY, cheap, salary, resume. Apply it everywhere.

Across our 12 accounts there were 20,022 distinct irrelevant search terms.

99.3% of them appeared in exactly one account. Only 141 appeared in two. Two terms appeared in three.

And the handful that did overlap were not generic junk words. They were competitor brand names, shared between two advertisers who happened to be in the same industry.

A master list of generic negatives is still worth having. It just is not where the money is. The money is in the 99.3%, which by definition nobody can write down in advance, because it only exists in your account, in your vertical, generated by how Google chose to interpret your keywords this month.

That is the uncomfortable part. The work does not generalize. It has to be done per account, repeatedly, on terms nobody has seen before.

What this means if you manage accounts

Three practical conclusions from the data:

Sort by cost second, not first. Handle the expensive outliers, then work the long tail, because that is where 86% of the money is. If the tail is too large to work by hand, that is the real finding, not a reason to skip it.

Frequency beats thoroughness. Waste accumulates in accounts that are reviewed monthly, not accounts that are reviewed badly. A shallow weekly pass catches more than a deep quarterly one.

Assume your clean account is not clean. Two of our 12 were genuinely tidy. The other 10 all had money in the report. The accounts wasting the most were not neglected ones; they were accounts someone was actively managing.

What we are still measuring

A fair caveat: 12 accounts over seven weeks is a small sample. It is real production data rather than a survey, but it is not a study, and the money is concentrated in four of the twelve.

We will publish an updated number as the sample grows. If it moves, we will say so.

---

PPC Sherpa connects to your Google Ads account and finds these terms for you, including the long tail that is too large to work by hand. Every recommendation shows what the term actually cost, and you approve what gets applied.

See plans