Google says average position no longer maps to a rank
John Mueller says the old position 1-10 is hard to map. Report impressions, clicks, and top-10 count per cluster, and date every move first.
Google's average position in Search Console describes roughly where a block of results sat on the page, averaged across every impression it recorded. It stopped being a rank a while ago. On 10 September, Google's John Mueller said it out loud on Reddit: the old "position 1 - 10" is hard to map, "or to make useful for site owners".
So an average position that slides from 8 to 24 can mean your page lost ground. It can also mean Google changed what sits above it. In the chart, both look the same. Here's how to report the metric so nobody on your team rewrites a page because the layout of the results changed.
What Mueller said, and the line everyone skipped
The thread asked how impressions get counted for generative AI results. Mueller pointed at the help doc, then added the part that matters for reporting: "Position for these is hard to do in a way that makes it useful, so we're currently tracking it like we do for many search features (as a block), & it's not separated out in the Gen-AI performance report." He closed by asking the forum what a useful position metric would even look like, as Search Engine Roundtable reported on 10 September.
Read that as the owner of the metric telling you it's a modelled estimate, tracked as a block, and open to redesign. If average position is the first row of your monthly SEO report, that's a problem worth fixing this week.
Why average position in Search Console moves when your rank doesn't
Google's own documentation has said this for years, in plainer language than most SEO posts manage. The Performance report help page calls position "an attempt to show approximately where on the page a given link was seen, relative to other results on the page", then averages that value across every impression, "because the position of the link will be different each time it is seen".
AI surfaces stretch that further. The same page states: "An AI Overview occupies a single position in search results, and all links in the AI Overview are assigned that same position." One AI Overview citing six sources hands all six the same number. And when Google starts showing an AI Overview on a query where it used to show ten links, every classic result underneath moves down a slot. Your average position falls. Your page never changed.
Google's advice on the same page is to watch the shape of the line: "a position number can mean different things in different situations, and so you should not make simple assumptions. We recommend that you monitor change in position over time, particularly sudden position changes, as well as absolute position."
Three numbers that hold up when the layout shifts
Swap the single average-position headline for three rows, each grouped by query cluster instead of by single keyword. A cluster of 20 queries absorbs the noise that one keyword amplifies.
- Impressions for the cluster. This is the demand Google has decided you're eligible for. It barely cares how the page is laid out, which makes it the steadiest week-to-week signal you have.
- Clicks for the cluster. The only number tied to a person arriving. If impressions hold and clicks fall, the layout above you took the traffic, and the fix is the title and the snippet rather than the body copy.
- Top-10 query count. How many queries in the cluster sit at an average position under 10, counted as a whole number. It moves when placement genuinely changes and stays put when one long-tail query wobbles from 60 to 80.
Average position still earns a place as the fourth row, read per cluster and read as a trend. The rule we use: it can open an investigation, and it can't close one. A cluster whose median position and top-10 count both drop in the same week is a real move. A cluster where only the average slid is usually a SERP feature arriving.
Date the move before you touch the page
When a number drops, the cheapest question is when. Pull the daily view for the cluster, find the first day of the fall, and check that date against the confirmed update log and the volatility trackers. The August 2026 spam update ran 18 to 20 August and finished in under three days, so a decline that starts on the 19th and holds is a different animal from one that starts on a Tuesday you deployed a template change.
Most shifts won't line up with anything Google announced, because most ranking updates now ship with no announcement at all. That's the case for keeping your own dated log of deploys, content changes, and volatility spikes. Six months of that log turns "rankings dropped" into "impressions held, clicks fell 30 percent, an AI Overview appeared on the head query on 19 August".
What to change in Monday's report
Three edits, about an hour of work. Group your tracked queries into clusters and give each one a tab. Put impressions, clicks, and top-10 count above average position in every view, so the eye lands on them first. Add a column for events, and write the date of every deploy and every confirmed update in it.
Then hold the line in the meeting. When someone asks why average position moved, the answer starts with whether impressions and clicks moved with it. Mueller is asking the industry what a better position metric looks like, which means the number you're staring at will keep changing shape. Build the report on the numbers that survive that.
If you want your Search Console reporting rebuilt around clusters and events instead of one fragile average, that's the kind of work we do. Tell us what you're tracking and we'll show you what we'd change.
