Gabriel Espinheira
Google Search Console will not show you most of the searches that bring in your best work. That is a deliberate design decision, not a broken setup.
Founders usually discover this about ten minutes before a call with whoever runs their marketing. A client mentioned last week that they found you by typing something oddly specific into Google, and you want to know how many others typed the same thing. You filter for it. Nothing. Forty-odd rows in the table, half of them your own company name, and not one of them looks like a sentence a real buyer would write.
Nothing is broken. You are reading a report that is accurate in its totals and censored in its detail, and nobody told you which half you were looking at.
TL;DR: Google Search Console omits any query not searched by more than a few dozen people over two to three months, so a small, specialised business loses its highest-intent terms from the query table. Click totals stay correct. The labels disappear. The whole record also expires after 16 months, with no way to recover it.
Why your Google Search Console query table is so short
Google's own documentation is unusually direct about this. In a 2022 Search Central post on performance data limits, it sets out the rule: "Anonymized queries are those that aren't issued by more than a few dozen users over a two-to-three month period." Those searches are stripped out of the table to protect the person who typed them. The same post adds that "while the actual anonymized queries are always omitted from the tables, they are included in chart totals, unless you filter by query."
Read that as an owner rather than an engineer and the asymmetry lands. A national retailer's head terms clear "a few dozen users" before lunch. But emergency dental appointment Rotterdam Sunday might be typed by nine people in a quarter, and every one of them is worth more to a practice than a thousand people searching dentist. The threshold is flat. Its effect is not.
So the rule quietly inverts. The more specialised your business, the more of your own search data Google withholds. The searches that bring you your best clients are the ones Google will never name for you.
"But the clicks are still counted"
That objection is right, and it deserves a straight answer before you throw the tool out of the window.
No traffic goes missing. Add up every row in your query table and you will get a smaller number than the chart sitting above it, because the anonymised searches are still inside the total. Google works the example in that 2022 post: four itemised queries account for 450 clicks while the site's real total is 550. Those 100 clicks happened. They arrived. They just have no names attached.
The same goes for the Pages tab, which is not filtered this way at all. If your question is "which pages earn clicks", Search Console answers it completely and you can stop reading here.
What is missing is intent. That is the only part of the report that changes a decision. "550 clicks last month" tells you nothing you can act on. Knowing that eleven people arrived after searching for a service you mention exactly once, in a sentence, on the fourth page of your site, tells you what to build next.
This is where founders draw the wrong conclusion. They filter for a service, see almost nothing, and decide there is no demand for it. Usually they are looking at a service that is working quietly, below the resolution of the instrument.
The interface narrows the view further. Google caps it plainly: "The maximum you can export through the Search Console user interface is 1,000 rows of data." The API will return up to 50,000 rows per day. Most people never leave the interface, so most people are reading the top of a list and treating it as the list.
Your search history is on a 16-month timer
Search Console keeps performance data for the last 16 months. That is Google's stated retention window, and it passes through to anything you connect to it.
Consider what that does to an ordinary question. You rebuilt the site two years ago. It cost real money and ran four months past the promised date. You now want to know whether it worked, so you open the date picker and drag it back. It stops. The months before the rebuild are gone. The comparison that would settle the argument no longer exists, and no support ticket will bring it back.
There is a fix, and it costs nothing but attention. Search Console can export your performance data daily into BigQuery, where, in Google's words, "tables are retained forever, by default." That is your archive.
The catch is the part people miss. The export cannot reach backwards. Asked directly whether the historical data could be pulled in, Google's John Mueller was blunt: "There's no historical export planned." It collects from the day you switch it on and not one day earlier. Every week you leave it off is a week that will simply not exist in your records.
Be honest about the trade. Setting it up takes about ten minutes, needs a Google Cloud project with billing attached, and you will probably query it twice a year. You are not buying an insight this afternoon. You are buying the ability to prove, in 2029, what the last three years of work actually did. If that is worth nothing to you, skip it, but skip it on purpose rather than by forgetting.
It is the first thing we switch on for any site we take over at SharpHaw, before a single page changes.
The new AI report counts impressions, not clicks
On 3 June 2026, Google launched generative AI performance reports in Search Console, giving sites a dedicated view of how they appear in AI Overviews and AI Mode. It reports impressions, pages, countries, devices and dates. Read that list again and notice what is absent: clicks.
You can now see that Google used your page to answer someone's question. You cannot see whether that sent you a single visitor.
The click behaviour underneath makes the omission harder to live with. Pew Research Center, studying 68,879 searches by 900 US adults in March 2025, found people clicked a result 8% of the time when an AI summary appeared, against 15% when one did not. They clicked a link inside the summary itself 1% of the time. It is a US sample measuring behaviour rather than your traffic, so treat it as direction, not as your number.
Put the two together and the practical advice is narrow. Watch the AI impressions. Do not rebuild your content around them. An impression you cannot connect to an enquiry is a number you cannot bank, and you have been sold enough of those already. Google says the reports are rolling out to a subset of sites and that it expects to add more metrics over time, so this may improve. Until it does, keep grading yourself the way that has always worked: tracked from click to client, not click to dashboard.
Do these four things this week
- Switch on the BigQuery bulk export. Today, not this quarter. It only records forward, so this is the one task on the list that gets more expensive every day you delay it.
- Stop reading the query table as a league table. Use the Pages tab to see which pages earn clicks, and compare a 28-day window against the previous one. Pages are complete. Queries are a sample.
- If BigQuery is too much, export the top 1,000 rows on the first of every month into a dated folder. It is a poor archive. It beats no archive.
- Ask whoever runs your search work to show you this report, not a rank tracker. Rank trackers query from one location on one device. Search Console reports what actually happened to real people. When the two disagree, the tracker is the one that is wrong.
Frequently asked questions
Why don't Search Console clicks match my Google Analytics sessions?
They count different events. Search Console counts clicks on a Google result. Analytics counts sessions that loaded your tracking code. Ad blockers, refused consent, abandoned loads and bot filtering all sit between the two. A gap of 10 to 20% is normal. A gap of 80% usually means the tracking is broken, not the traffic. We wrote about one common version of this in GA4.
Is average position the same as my ranking?
No. Google defines it as the average position of the topmost result from your site, averaged across every impression. A page sitting second for a rare phrase and eleventh for a common one does not "rank sixth" anywhere. Nobody has ever seen that position on a screen. Treat it as a trend line, never as a scoreboard.
Can I recover Search Console data older than 16 months?
No. Once it passes the retention window it is gone from the interface, the API and any connected report, and Google offers no way to request it back. The BigQuery export preserves everything from the day you enable it forward, which is why enabling it early is the whole game.
The point
Search Console is the most useful free thing Google gives a business owner, and it is not a report card. It is a partial, expiring record with rules worth knowing: totals you can trust, query labels you cannot lean on, and a 16-month clock running on all of it. Read it that way and it will show you where demand actually sits. Read it as a scoreboard and it will flatter you before quietly deleting the evidence.
Turn the export on this week. Then open your Pages tab and look at it honestly for ten minutes.
Want a second opinion on what your search data is actually saying? Book a 30-min call and bring your Search Console. You will leave with a read you can act on. If you would rather have the whole search and AI-visibility surface run for you every week, that is what our Content Engine does, tracked inside SharpOS so you can watch the work as it ships. Every price is on the Plans page.
Plan. Build. Iterate.
