Is Average Position in Search Console Accurate? Partly

By Ugur Saritepe · September 13, 2026

Is average position in Search Console accurate? The number is computed exactly as Google documents, and it still misleads. On four of our pages, 74 to 94% of impressions came from unnamed tail queries the pages never targeted, dragging every average away from how the pages really rank. That read cancelled a fix we had already sized at +146 clicks.

What Google actually averages

Search Console records the topmost position your page occupied each time it appeared in results, then averages that across every impression, an impression being one appearance in front of one searcher. Google's own documentation calls position "a complex metric that can be misleading if you don't understand the subtleties" and recommends watching change over time rather than absolute values. The subtlety that bit us is which queries get averaged: all of them, including every loosely related search your page surfaced for once at position 40. The metric is honest arithmetic over a mixture, and the mixture is the problem.

The morning it almost cost us a rewrite

game-scout.app, the games directory I run and write about openly, had a list page that looked broken by every summary number: best-flight-games, 117 clicks on 10,404 impressions in the month before 9 September 2026, a 1.2% click-through rate at headline position 9.2. An earlier read had already projected +146 clicks from retitling it into flying and plane vocabulary, batch three of a title-fixing cycle. This is what I typed when we went back in:

per-page query pull, named target queries split from the rest

page       head CTR   tail CTR   tail share of impressions
flight     4.51%      0.91%      94%
bus        4.85%      1.24%      91%
trucking   8.71%      1.50%      83%
samurai    2.71%      1.57%      74%

The averages were accurate; the summaries they produced were not. Each page's headline CTR blended two populations: a small head of queries the title actually targets, converting at 4.51 to 8.71%, and a long tail of adjacent searches converting under 1.6%. Nothing on those pages was broken. The blend just made healthy pages look sick, and it was about to make us operate on one.

The +146-click fix we withdrew

The retitle projection did not survive the decomposition. The flying and plane queries the new title would have targeted totalled about 355 of the page's 10,404 impressions, and they already converted at 2.6 to 5.6%. The projected +146 clicks assumed the whole page's impressions were reachable by a title change; the reachable share was about 3%. The batch was cancelled in the same session, and the titles stayed as they were. We have not touched them since.

The comparison that settled it: our cleaning-chore games list earns 146 clicks on 1,830 impressions at headline position 7.0, with 97% of its impressions on page 1 converting at 11.8%. The flight list earns fewer clicks from 5.7 times the impressions. Same site, same template, same month. The difference is not rank and not the title; flight sits on page 1 for over a hundred queries whose searchers mostly want something else, an intent mismatch no rewrite reaches. Re-checked on 13 September 2026 over the previous 28 days, the pattern holds: flight at 110 clicks / 10,054 impressions / 1.09% CTR, cleaning-chore at 140 clicks / 1,750 impressions / 8.0% CTR.

We had already classed 2,000 of the site's queries by what the answer demands and knew intent decides clicks at a given rank. This was the same lesson at the metric level: before acting on any blended number, find out which queries are in the blend.

Run the split on your own rows

The decomposition is one filtered pull per page and ten minutes of reading. It needs no tools beyond Search Console itself, though we run it by asking an assistant connected to our data:

1. Pull every query for one page (in Search Console: Performance, filter by page, Queries tab; over the API or MCP: one query_search_analytics call with a page filter).
2. Mark the head: the queries your title and H1 actually target, usually 3 to 10 phrasings. Everything else is tail.
3. Compute CTR for each group separately, and note what share of impressions the tail holds.

Good output looks like our flight page: a head converting at or above the rate typical for its position, with the tail holding most impressions at a fraction of that rate. That page needs no title work; its average is being manufactured by the tail, and the honest levers are a new page for the tail intent or acceptance. Bad output is the reverse: a head converting below what its position should earn. That is a real snippet problem on the queries you target, and the one case where the title and description deserve the rewrite. If the page shows impressions but almost no clicks at all, start with the impressions-but-no-clicks diagnosis first.

This post is part of our AI-assistant SEO series, written from the SEO work we do on our own sites. For a five-minute check of the basics on any site, the free triage report runs eight checks from a URL alone, no account needed.

FAQ

Is average position in Search Console accurate?

The number is computed correctly, but as a summary of how a page ranks it often misleads. Google averages the topmost position across every impression, including hundreds of loosely related queries the page never targeted. On four of our pages, measured 9 September 2026, 74 to 94% of impressions came from that unnamed tail, so the average described the tail, not the page.

Why is my CTR low when my position looks good?

Split the page's queries into the head queries the title targets and everything else before concluding anything. Our flight games list showed a 1.2% overall CTR at headline position 9.2, which looks broken. The head queries converted at 4.51%; the other 94% of impressions were tail queries converting at 0.91%. The page was fine, and the blended number was manufactured by the tail.

Should I rewrite my title when average position is fine but CTR is low?

Check whether the title's own queries already convert first. We projected 146 extra clicks from retitling a page, then found the queries the new title would target totalled about 355 of 10,404 impressions and already converted at 2.6 to 5.6%. The retitle was cancelled, because a title change acts on the head, and the head was already working.

In your AI assistant

Ask your AI about your own site

TrafficTriage plugs into Claude, Cursor, and ChatGPT. Your assistant runs the 8 checks, reads your Search Console data, and can even check your site every morning. Free account, no API key, nothing to install.

Sign up freeSee how it works