Nearly half the organic clicks flowing into your site come from search terms Google refuses to show you. That's the headline finding from an Ahrefs study of 146,741 websites and roughly 9 billion clicks: 46.08% of all clicks in Google Search Console originate from "anonymized queries" that never appear in your keyword reports.
If you've ever stared at the gap between your Search Console chart totals and your query table rows, you already know something is off. This is why.
What Google Is Actually Hiding
Google defines anonymized queries as searches issued by fewer than "a few dozen users" over a two-to-three month period. The rationale is privacy: if a query is rare enough, showing it could theoretically identify the person who searched it. Searches containing personal or sensitive information get the same treatment.
That sounds reasonable in principle. A handful of obscure long-tail queries getting scrubbed shouldn't matter much, right?
The problem is scale. The long tail is enormous, and individually rare queries add up to a staggering volume of traffic. Google's own documentation used to describe these hidden queries as "very rare." After the Ahrefs study went public, Google quietly changed that language to "some" and added a note acknowledging that Search Console stores only "top data rows and not all data rows." That's a significant concession; "very rare" and "some" are doing very different amounts of work.
John Mueller's official position was that "nothing has changed in terms of functionality." Danny Sullivan acknowledged that "long tail terms that are very rare can add up." Both statements are technically true and completely miss the point. The functionality didn't need to change. It was already hiding 46% of your data.
Smaller Sites Get Hit Hardest
The Ahrefs numbers are averages. The reality for individual sites varies wildly, and not in your favor if you're running a smaller property.
- Small sites could see just 18.3% of their query data on average (range: 0–37%)
- Medium sites fared better at 39.7% (range: 21–61%)
- Large sites retained 76.3% visibility (range: 63–88%)
Read that again. Some small sites had zero percent query visibility. Every single click came from an anonymized query. Glenn Gabe reported seeing sites with over 80% of queries filtered.
This tracks. Small sites attract more long-tail, low-volume searches proportionally. Each individual query is less likely to cross Google's privacy threshold. But the practical effect is brutal: the sites with the least resources for SEO analysis are also the ones getting the least data to work with.
The Discrepancy You've Been Ignoring
Here's where it gets operationally messy. Search Console reports data at multiple levels, and they don't agree with each other.
John Mueller has explained the hierarchy: page-level and site-level reports include all clicks, including those from anonymized queries. Query-level tables exclude them. So when you look at a page showing 241 clicks but the query breakdown only accounts for 148, those missing 93 clicks aren't a bug. They're anonymized queries that Google stripped from the keyword view.
The page-level data captures 97–104% of total clicks across all site sizes. Query-level data? Far less. The chart at the top of your Performance report and the rows in the table below it are, in a meaningful sense, measuring different things.
Filtering for branded versus non-branded terms doesn't help either. Those filters operate on the visible query set, not the full one. You can't surface anonymized queries by slicing the data differently within Search Console's standard interface.
BigQuery: A Partial Fix
If you export Search Console data to BigQuery (available to any verified property owner), you get something the standard UI won't give you: anonymized queries appear as NULL values in the query field, but their click and impression metrics are preserved.
You still can't see what people actually searched. The terms remain hidden. But you can quantify how much traffic comes from those hidden searches, break it down by page, track it over time, and understand its share of your total performance.
That's useful. Knowing that 60% of a specific page's clicks come from anonymized queries tells you something important about that page's traffic profile, even without the keywords. It suggests the page is capturing a wide spread of long-tail variations, which is usually a signal of strong topical coverage.
The standard API and UI simply make this data vanish. BigQuery at least lets you see the shape of the hole.
What This Actually Means for Strategy
Our read: the obsession with individual keyword rankings has been on borrowed time for years, and this data makes the case concrete. When you literally cannot see half your keywords, building strategy around keyword-level tracking is building on sand.
The rational pivot looks like this:
Page-level performance becomes the unit of analysis. Mueller himself recommends relying on page-level or site-level data for the complete picture. Pages show nearly all your clicks. Keywords show roughly half.
Topic clusters over keyword lists. If you're covering a topic thoroughly enough, you'll capture long-tail queries you never specifically targeted. You won't see them in your reports, but you'll see their effect in page-level metrics.
Trend analysis over point-in-time snapshots. A 20% drop in page-level clicks means something regardless of whether you can identify which queries dried up. The directional signal is intact even when the keyword detail isn't.
None of this means keyword research is dead. You still need to understand search intent, identify content gaps, and structure your content around the terms people use. But the monitoring and measurement side of keyword work is operating with half the picture at best. Accepting that, rather than pretending the data is complete, is the first step toward better analysis.
The deeper question is whether Google's privacy threshold is set appropriately or whether it's overcorrecting. When "a few dozen users" over two to three months is the cutoff, a query searched 50 times across 90 days gets hidden. That's not exactly a privacy risk. But it is conveniently opaque for a company that sells the complete keyword data through Google Ads.