Every month a client asks some version of the same question: “Search Console says our average position is 7.4, but I searched the term myself and we were nowhere near the top. Which one is lying?” Neither is. The two numbers measure different things, and the gap is where most reporting arguments start.
Average position is the most consistently misread metric in the Performance report. Understanding what each number counts is the difference between a report that drives decisions and one that just gets forwarded around.
What an impression actually counts
An impression is recorded when a link to your site appears in the current page of results, whether or not the person scrolled down far enough to see it. Google’s rule: if a user must click to see more results, an impression is counted for links in the current set, scrolled into view or not.
The exceptions matter. Inside an independently scrolling or expanding widget — a results carousel, an expanding FAQ result — the item typically has to be scrolled into view within that widget, or expanded by a click, before an impression registers. The same is true for infinite-scroll surfaces like mobile image search and Discover.
One more detail that saves arguments: scrolling away and back during a single query does not count as multiple impressions. If impressions doubled, twice as many searches surfaced you.
The practical takeaway is that impressions measure exposure to a results page, not attention. A page can gain thousands of impressions simply by starting to appear on page three for a broad term — not visibility in any sense a business owner would recognize, and it will drag CTR down while looking like growth on the chart.
Average position is the topmost position, averaged
Here is the mechanic that resolves most of the confusion. The position value is the topmost position occupied by a link to your property for a given query, averaged across all queries where you appeared.
Google’s own illustration makes it concrete. If one query returns your site at positions 2, 4, and 6, it counts as position 2 — the topmost. If another returns you at 3, 5, and 9, it counts as 3. The average is 2.5. Your site occupied six slots, and the report shows one number matching none of them.
Position is also counted by element, not by link. A results page is a stack of elements, numbered top to bottom on the primary side and then the secondary side. Each element occupies one position no matter how many links it holds, so a knowledge panel with eight links puts all eight at the same position. Ads do not occupy a position at all, and a featured snippet typically sits at position 1 because it links out to the source page.
This is why a position value of 11 is nearly uninterpretable in isolation: depending on the surface, it could mean the top of the desktop right-hand panel, the first result on page two, or the second row of image results — visible without scrolling at all. Google’s guidance is to watch position change over time, sudden movements in particular, rather than treat the absolute number as a rank.
Why the number never matches what you see when you search
When you search your own term from your own office, you are running one query, from one location, on one device, with your own search history attached. The report averages every impression across every location, device, and searcher in the period. Google addresses this directly: your specific search may differ from the average because of search history, location, and other variables.
For treatment centers and any provider serving a defined geography, this matters more than it does for a national brand. A facility might rank near the top in its own city, sit far down the page for the same term searched two hundred miles away, and never appear outside the state. Those impressions average into one property-level number describing no real searcher’s experience. The fix is not to distrust the data but to filter it: segment by country and device, and use the Pages dimension to isolate the location pages meant to perform in a specific market.
Chart totals and table totals disagree on purpose
The chart at the top of the report is aggregated by property. The table below it changes aggregation depending on the dimension selected: Queries, Countries, Devices, and Dates aggregate by property, while Pages and Search appearance aggregate by page.
So when two results from the same site appear for a single query, the chart counts one impression for the property while a page-grouped table counts one per page. The totals will not reconcile, and are not supposed to. If you have ever exported two views of the same date range and gotten different impression counts, this is almost always why.
Related: the newest data can be preliminary. It is still being collected, can change over the following hours, and Google marks it with a dotted line. Reading yesterday’s dotted-line dip as a ranking event is a good way to spend a week chasing nothing.
Your data is assigned to the canonical URL, not the URL that ranked
Search Console assigns click, impression, and position data to the canonical URL Google selected — the URL it treats as the best representative when several serve essentially the same content. Device variations, duplicate URLs, and in some cases alternate language versions all roll up into that one canonical.
For most sites this is a convenience. For sites with duplication problems it is a trap. If Google chose a canonical you did not intend — an old URL, a parameterized version, a near-duplicate city page — the data for the page you care about appears under a URL you are not looking at. The page looks dead while quietly earning impressions under another address. Auditing Google’s canonical selection is a prerequisite for trusting page-level data, and its documentation on specifying a canonical URL is worth reading before assuming the report is wrong.
Clicks, and the queries that never appear
A click is counted when it sends someone to a page outside Google Search, Discover, or News. A link that keeps the user inside Google does not count, and clicking a result, going back, and clicking it again counts once.
Query refinements are the notable exclusion. When a link inside search results runs a new Google query rather than sending the user to a site — switching from the web tab to the image tab, for instance — no clicks or impressions are logged for it, and all activity on the resulting page is attributed to the new query.
This is also why Search Console clicks never match sessions in your analytics platform. They count different events at different points in the chain, and the gap widens with redirects, slow pages, and privacy tooling. Reconciling them exactly is not a realistic goal; watching both trend lines is.
How to actually use the report
A few habits make the data worth the time.
Compare periods, not absolutes. Position averages are noisy by construction. A small improvement across a quarter on a stable query set is a signal; the same move week over week is not.
Filter before you interpret. Sitewide averages blend branded and non-branded terms, and branded queries almost always sit near position 1 — so a sitewide average improving can simply mean brand searches grew.
Read impressions and position together. Position getting worse while impressions climb usually means a page is ranking for a wider set of terms, not losing ground. That pattern is common when a page starts matching longer, more specific queries — often exactly what you wanted from a technical SEO program.
Separate the surfaces. Use the Search appearance dimension to see how much exposure comes from rich results versus plain links. As more queries resolve inside AI-generated answers, the relationship between impressions and clicks keeps shifting — the practical argument for treating answer engine optimization as a distinct workstream rather than a rebrand of the same checklist.
Remember what is missing. Paid placements do not appear here and do not occupy an organic position, so nothing in this report speaks to paid search performance. Blending the two in one dashboard without labeling sources is how reporting loses credibility.
The honest version
Search Console is the only free source of first-party data on how Google presents your site. It is also a set of averages over a system that personalizes nearly every result, which makes it excellent for detecting change and poor for answering “what position are we.” Anyone reporting a single average position as though it were a rank either has not read the documentation or is counting on the client not to.
Two pages are worth reading in full: Google’s explanations of impressions, position, and clicks and the Performance report itself. Google notes that the heuristics involved, including position numbering and carousel visibility rules, are subject to change — worth rechecking when a metric shifts without an obvious cause on your end.
