SERP

The ten blue links are still there on most queries, but they are rarely the whole page and often not the top of it. Above and among them sit generated summaries, featured snippets, People Also Ask, local packs, shopping units, image rows, video carousels and ads, in a combination decided per query.

Each of those does two things to an organic result. It pushes it down the page, and it takes some share of the searchers who would otherwise have had to click something. A result that is first among the blue links can easily be below the fold.

This is why position alone stopped describing a commercial outcome. A client told they rank first, who then searches and has to scroll to find themselves, will reasonably conclude they were misled, and the agency will spend the meeting recovering rather than reporting.

Two people searching the same words at the same moment can get materially different pages. Location changes local results and sometimes the whole composition. Device changes how many results fit above the fold and which formats appear. Prior activity and language settings change it again.

Rank tracking tools handle that by querying from a chosen location, on a chosen device, with no history, and reporting what comes back. That is a reasonable methodology and it is also a synthetic page that no actual person saw, which is the source of most arguments between an agency report and a client phone.

The useful framing for a client conversation is that a ranking is a distribution, not a number. The tool reports one sample of it. Search Console reports the average across the real impressions the site actually received, which is a different and better sample.

None of which makes rank tracking useless. It is a consistent instrument, so it measures change reliably even where it measures level imperfectly, and change is usually the question.

The value of a position depends entirely on what is above it. Ranking first on a query whose page opens with an AI summary, then People Also Ask, then a local pack, then four ads, puts the result a long way down a page that has already answered the question twice.

The same position on a query with a bare list of results is worth several times as much. Since both are reported as position one, the number on its own carries almost no information about the traffic it should produce.

The practical consequence for agency work is that page composition belongs in the keyword research, not only in the reporting. A term with high volume whose page is dominated by summaries and rich results may be worth less than a term with a fifth of the volume and a clean page.

It also changes what a win looks like. On a page with a featured snippet, taking the snippet is worth more than moving from three to two. On a page with a local pack, appearing in the pack is worth more than either.

Search Console reports the average position of the impressions your site received for a query, across every variation of the page it appeared on. It is the most honest ranking figure available, because it is measured from real appearances rather than from a query made by a tool.

It also moves for reasons unrelated to your site. If you begin appearing for a set of new, weaker queries, your average position gets worse while your visibility improves. If a page stops appearing for its worst queries, average position improves while nothing got better.

That makes it a poor headline metric and a good diagnostic one, and the fix is to read it alongside impressions rather than on its own. Impressions up and average position down is usually expansion. Impressions flat and average position down is usually a real loss.

Filtering matters as much as reading. The same report segmented by query, by page, by country and by device answers different questions, and the unsegmented number is an average over all of them at once.