Average position in Google Search Console is the average place your site's best result held across the searches where it actually appeared — weighted by how many times it appeared, and calculated only over the rows sitting inside whatever filter you have applied. It is not "your ranking." Searches where you didn't show up at all aren't counted as a bad position; they're not counted at all.
That single detail explains almost every confusing thing this metric does. An average built only from appearances moves whenever the mix of appearances changes — which happens constantly, for reasons that have nothing to do with any page going up or down in Google.
What Search Console counts as one "position"
Every time your site appears in a set of results, Search Console records one impression and the position that result held in the list. A few rules govern that recording, and they're where most misreadings start.
It records the topmost result, once. If two of your pages appear for the same search, the property-level report logs the higher of the two and treats it as a single impression, not two. This is why a page you know ranks can look absent from a property-level view: another page of yours was above it.
Position is counted in the list, not on the screen. Position 1 means the first result in the ranked list — not the first thing the searcher's eye lands on. Ads, a map pack, a shopping row, a featured snippet block and a wall of "people also ask" can all sit between position 1 and the top of the visible page. Two sites can hold the same position and earn very different click-through rates, because one sits under a thin result page and the other under a crowded one.
No appearance means no row. A query you rank nowhere for doesn't enter the calculation as position 90 or position 100. It simply isn't there. Your average position describes the searches you are already visible in, and says nothing about the ones you aren't.
Some queries are withheld. Search Console omits rare queries that could identify an individual searcher. So the query-level rows you can see won't add up to the totals shown above them — that's expected rather than a bug, and it's one reason the query table and the site summary disagree.
Why the average moves when none of your rankings changed
Because the number is weighted by impressions and built only from appearances, three things shift it with no ranking movement at all.
New queries entering the mix. A page that starts picking up impressions for twenty new long-tail searches — most of them deep in the results, because that is where new visibility begins — adds twenty low positions to the average. Nothing you already ranked for moved. The average still falls.
The volume behind existing queries changing. Impressions weight the average. If seasonal demand doubles for the one query where you sit near the top, and everything else stays flat, your average improves purely because a good position now carries more weight in the calculation.
The filter you're looking through. Average position is computed over the current filter set, so it changes with the question you asked. Site-wide, one country, one page, one device and one query each produce a different number, and only the narrow ones mean anything specific. A site-wide average blends your brand-name searches — where you sit at the top permanently — with everything you are straining to rank for, and the result describes neither.
That is the fundamental limitation. It is an average over a set that keeps changing shape, so trends in it are only readable when the set is held still. Every useful reading of this metric starts with a filter.
Why a falling average position is often good news
This is the part that causes panic in monthly reports. A dropping average position, with impressions rising alongside it, usually describes a growing site rather than a declining one.
Growth in organic search typically arrives as new queries first, better positions second. A page gets indexed, starts surfacing well down the results for a scattering of related searches, and gradually climbs. Throughout that climb the site is collecting weak positions it did not have before, and every one of them pulls the average down while the underlying situation improves.
The inverse trap is just as common. An average position that improves while impressions fall often means you have lost visibility on the long tail — the weak positions that were dragging the average down have stopped appearing, leaving a tidier average over a smaller footprint. The number got better; the site got smaller.
This is why average position belongs beside impressions and clicks rather than on its own, and why it behaves as a leading indicator rather than a result. Positions and impressions move well before clicks and revenue confirm anything, and the gap between them is normal. Knowing the gap exists is what stops people abandoning work that is quietly succeeding — there is more on that lag in how long SEO takes to work.
Why it never matches your rank tracker
Two tools, two different measurements. Neither is wrong.
A rank tracker checks a fixed list of keywords, from a chosen location and device, on a schedule, whether or not anybody searched that term. It answers: where do I stand for the keywords I decided to care about?
Search Console reports what really happened — every search that produced an impression, from every location and device your searchers actually used, weighted by how often each occurred. It answers: across everything I was seen for, where was I seen?
So a tracker showing a strong position and Search Console showing a much weaker average for the same keyword can both be accurate. The tracker measured one location on one device; Search Console averaged in every search from every region and device, including the ones where you perform far worse. Local intent, personalisation and device differences all widen that gap.
Use the tracker for competitive keywords you have deliberately targeted. Use Search Console for what your visibility looks like in the wild. Comparing their headline numbers against each other is the one job neither is built for.
How to read average position so it tells you something
Narrow it until it means one thing, then watch the trend rather than the value.
- Filter to a single query and a single page. At that resolution the number describes something real: where this page sits for this search. That is the level at which a change is worth investigating.
- Split by device and by country before concluding anything. A mobile average sitting well below desktop is a genuine finding with a findable cause. A blended number hides it.
- Read it beside impressions and clicks, always. Position alone cannot tell you whether a move was a gain or a shrinkage. The three together can.
- Watch for instability, not only level. A query whose position saw-tooths week to week, with impressions split across two of your URLs, is a symptom worth chasing — that pattern is how keyword cannibalization shows up in the data.
- Compare like periods. Month against the same month, or a rolling window against the one before it. Comparing a quiet week to a busy one compares two different query mixes and tells you nothing about your pages.
And keep it out of the headline of any report. Average position is a diagnostic you filter your way into, not a scoreboard — the numbers that belong at the top of a report are the ones that connect to revenue, which is the case the SEO analytics and reporting guide makes in full.
FAQ
Does average position include the map pack, images or video results?
Those are reported under their own search appearance types, and the position recorded refers to placement within that element rather than the main list of results. Filtering by search appearance separates them; without that filter you are averaging different kinds of placement together.
Is a lower number always better?
Lower is better for a single query on a single page. Site-wide, no — a falling average alongside rising impressions usually means new visibility, and a rising average alongside falling impressions usually means lost visibility. Direction only carries meaning next to volume.
What counts as a good average position?
There is no target worth chasing, because the number depends entirely on the mix of queries underneath it. A site with broad long-tail coverage will show a worse average than a site visible for three branded terms while earning far more traffic. Judge positions query by query, and judge the site on clicks and conversions.
Why don't my query positions average out to my site's average position?
Two reasons: withheld rare queries never appear in the table, and the site figure is weighted by impressions rather than being a plain average of the rows. A straight average of the visible rows will not reconcile with it.
Does average position affect my rankings?
No. It is a report of what already happened, not an input Google uses. Nothing you do to the number itself changes anything; only changes to the underlying pages do.
Why did my average position change on a day I published nothing?
Because the query mix changed. Search demand shifts daily, competitors move, and result layouts change. Read the trend across weeks — single-day movement in this metric is almost always noise.
Treat average position as a question-answering tool rather than a grade. Filter to one page and one query, read it next to impressions and clicks, and it will tell you something true about where you stand. Leave it site-wide and it will tell you a story that changes every time your visibility grows. If you would rather have someone build reporting that separates real progress from statistical drift, start at WeSEO.
