What Search Console Will and Won't Tell You
Search Console is the only free source of your real query data, and it withholds more of it than most people realise. Here is what is missing, why, and what to do instead of guessing.
Google Search Console is the only place you can see the searches that actually reached your site. Everything else — volumes, difficulty scores, competitor estimates — is modelled. Search Console is measured.
Which makes its omissions worth understanding properly, because people build strategies on top of numbers they have misread.
The queries you cannot see
Open a page's query report and add up the clicks in the table. Then compare that to the page's total. The table will be short.
Google withholds queries that are rare enough to identify an individual searcher. It does not tell you which ones, how many, or where the threshold sits, and the threshold is not a fixed number of searches. The clicks still count toward your totals; the query text is simply gone.
For a site with broad, repetitive search demand this hides very little. For a niche B2B site, where almost every search is phrased slightly differently, it can hide the majority of your queries. The long tail is precisely the part that gets anonymised, which is a problem, because the long tail is usually where the buying intent lives.
What to do instead of guessing. You cannot recover the exact strings, but you can bracket them. Compare the page's total clicks with the sum of its named queries — the difference is the size of what you cannot see, and it is worth knowing whether that is five per cent or seventy. Then widen the date range: a query too rare to name in seven days is often named across three months, because the threshold applies to the period you asked for. That single change surfaces more real queries than any third-party tool will model for you.
Averages that are not averages
The position column is the most misread number in the product.
An average position of 8.4 does not mean you sit eighth. It usually means you rank third for a handful of searches and twenty-fifth for a long tail, and the mean landed in between — describing a position you have never actually held. Filtering to a single query and a single country is the only way to make the number mean what people assume it means.
Average CTR has the same problem in a more expensive way. It is total clicks over total impressions, so a page that ranks well for one term and appears far down the results for a hundred others will show a terrible CTR that no amount of title rewriting will fix. The impressions are the problem, not the title.
Impressions are not visibility
An impression means your result was in the response Google served. It does not mean anyone scrolled to it. A page can accumulate impressions for months from positions nobody reaches, and the graph will slope upward while nothing whatsoever is happening.
The useful test is whether impressions and clicks move together. Impressions rising with clicks flat is not "we are gaining visibility, clicks will follow." It is normally Google testing you across a wider set of queries where you are not competitive.
The window closes
Search Console holds roughly sixteen months. There is no archive behind it, and once a day ages out it is gone — including the months you would want to compare against after a redesign or a migration.
If you care about year-over-year, the time to start storing your own copy is before you need it. Any regular export will do. The point is that this is the one dataset nobody can re-derive for you later.
What it will tell you, that nothing else can
Set against all that, one thing makes Search Console worth more than any paid dataset: it knows which queries your specific site actually appeared for. No modelled tool knows that. They know what a keyword's volume looks like in aggregate; they cannot know that Google decided your page was relevant to it.
The highest-yield use of that is boring and reliably productive. Filter to queries where you sit somewhere on the second page. Those are searches Google already considers you a plausible answer for, and where a better-matched title, a section that answers the question directly, or an internal link from a stronger page can move you into a position that gets clicked. It is a finite, ranked list of pages you can edit this week, and it comes from data you already own.
The second-highest is comparing what you rank for to what you meant to rank for. Sort a page's queries by impressions and read them as a description of what Google thinks the page is about. When that description does not match your intent, you have found a content problem that no keyword tool would have flagged, because the keyword tool does not know what you were aiming at.
Where a tool helps
Nothing above needs a paid product. It needs the patience to filter properly and the discipline to distrust an average.
What a tool adds is joining that measured data to modelled data — attaching volumes and difficulty to your striking-distance queries, or checking whether the competitors ranking above you are the ones you thought. PandaCrawl reads your Search Console through your own Google connection and exposes it to an agent over MCP alongside keyword and SERP data, so those joins can be asked for in one sentence rather than assembled by hand across three tabs.
But the reading discipline comes first. A tool that joins misread averages to modelled volumes produces a confident chart built on a number that never meant what you thought it did.