The short answer
Average position is the topmost spot your site held in the results, averaged across every impression it earned. Because it is an average, it shifts whenever the mix of queries you appear for changes, so a site can gain rankings and watch the number get worse. Read it one query at a time, never as a sitewide score.
Average position is the number store owners check first in Search Console and the one they misread most often. It sits next to clicks, it gets a line on the chart, and it looks like a ranking. It is an average of rankings, and averages move for reasons that have nothing to do with the thing you think they measure.
The result is a familiar panic. The line goes the wrong way, someone assumes a Google update has hit, and an afternoon goes on checking pages that rank exactly where they did last week.
The examples here come from a store selling coffee equipment: grinders, espresso machines, descaling kits and the brewing guides that sit alongside them. The arithmetic is the same for any catalogue.
What is average position in Google Search Console?
Average position is the highest spot a link to your site occupied in Google’s results, averaged across the searches where you appeared. Google’s own definition of impressions, position and clicks puts it in two parts: for each search, Search Console records the topmost position your property held, then averages those figures.
Two details in that definition matter more than they look.
The first is “topmost”. If a search for “best burr grinder for espresso” shows your buying guide at position 4 and a grinder product page at position 9, that search contributes a 4 and nothing else. The 9 is invisible in the query view.
The second is what counts as a position. Google counts result elements from top to bottom, and a compound element such as a carousel or a knowledge panel takes a single position with every link inside it sharing that number. Position 3 is the third element on the page, which on a results page stacked with features can sit well below where a searcher stops reading. It is an ordinal, not a measure of how far down the screen you are.
Worth knowing
An impression is counted when your result appears on the page of results the searcher loaded, whether or not they scrolled down to it. That is Google’s documented behaviour for a standard results page. So a position 9 impression on a phone often means the searcher never saw you at all, and the average position figure has no way of telling you that.
How is average position calculated?
It is weighted by impressions, because every impression records its own position. A query searched a thousand times contributes a thousand positions to the average, and a query searched once contributes one. Google’s worked example uses two queries with one search each, which hides the weighting, but the logic holds as soon as the counts differ.
Here is what that does to a single page. Take a guide called “how to descale an espresso machine” on the coffee store. Over a month it earns 400 impressions at position 3 for its main query. That month the page’s average position is 3, and it is an honest figure.
The next month nothing about that ranking changes. Still position 3, still 400 impressions. But Google starts testing the page against thirty related searches it has never shown for before: “citric acid vs descaling solution”, “descale breville without solution”, “how often to descale a bean to cup machine”. It shows the page for those at around position 45, and together they add 2,000 impressions.
The page’s average position is now (400 × 3 + 2,000 × 45) ÷ 2,400, which comes out at 38.5.
The chart shows a page that fell from 3 to 38. The reality is a page that held its best ranking and picked up thirty new ones, which is usually an early sign Google is trusting it with a wider topic.
This is illustrative arithmetic rather than a real report, but it is a shape we see in our own data when a new article starts gathering long-tail queries.
Why does the property total differ from the page total?
Because Search Console calculates the number two different ways depending on which tab you are looking at. The Performance report documentation says data grouped by query, country, device or date is aggregated by property, and data grouped by page or search appearance is aggregated by page.
Aggregated by property, a search that showed two of your pages counts once, at the higher of the two positions. Aggregated by page, each page gets its own impression and its own position. Go back to the grinder search: the Queries tab records a 4. The Pages tab records a 4 for the buying guide and a 9 for the product page. Filter the report down to that product page and “best burr grinder for espresso” will show position 9, even though the site as a whole held position 4.
Neither is wrong. The property view answers how visible the site was for a search, and the page view answers how one URL performed. The same split exists in the Search Analytics API, where the aggregationType parameter takes byProperty or byPage, and property aggregation cannot be combined with a page filter. If an export and the on-screen chart disagree, this is the first thing to check.
Why did my average position drop when nothing lost rankings?
Because the site started appearing for more searches at lower positions, and those new impressions pulled the average down. This is the most common reason the line falls, and on a site that is growing it is close to inevitable.
Every new article enters the index somewhere deep. A fresh guide to dialling in a grinder usually starts showing well down the results for a spread of queries before it settles, if it settles at all. Each of those impressions is averaged in alongside the queries where you have held position 2 for a year. Publish ten guides in a month and the sitewide figure will very likely get worse, even if every one of them is on its way up.
Seasonality does the same. When “gifts for coffee lovers” surges before Christmas and the store ranks at position 18 for it, the weight of position 18 in the average rises with the demand, and the headline number sags every December for reasons that have nothing to do with the site.
Check before you worry. Filter to a handful of queries you care about and compare them period against period. If each one holds or improves while the total drops, the mix changed and the rankings did not.
Can average position improve while traffic falls?
Yes, and this is the more dangerous of the two misreadings because it looks like good news. When a site loses its low rankings, the impressions at those deep positions disappear from the average, and what is left is a smaller set of better positions.
Say the coffee store has old product pages for discontinued grinders still earning impressions at positions 50 to 70. Those pages are deleted, the impressions go, and the sitewide average improves by several places overnight. Nothing ranks higher than it did. The site is visible for fewer searches.
A shrinking site can post an improving average position for months, so read it next to impressions every time. Position improving while impressions fall is a site getting narrower, and whether that is fine depends on whether the lost impressions were ever going to earn a click.
What the data showed us
On our own store, Supplements Wise, the weighted average position moved from 17.9 in the first month of the window to 8.5 in the last, over the same six months that clicks grew from 184 in the first 28 days to 7,011 in the last 28. That looks like a clean story, and we still do not lean on the position figure. The site published a great deal of content in that window, which pushes the average both ways at once, and we cannot separate how much of that move is rankings and how much is mix. The clicks are the number we trust. The full timeline is in the case study.
What is a good average position in Google Search Console?
There is no good sitewide figure, and any article that gives you one is guessing. A sitewide average is a blend of your brand searches at position 1 and every deep long-tail impression you have ever earned, so it mostly measures how broad your visibility is, and broader is usually better even though it makes the number worse.
At the query level the question has a real answer. Position 1 to 3 is where the clicks are. Position 4 to 10 earns some. Beyond 10 a query earns very few clicks, but a query sitting between 5 and 20 with meaningful impressions is the most useful thing in the whole report, because Google already considers the page relevant. That band is the entire basis of our striking distance keyword method, and it works precisely because it reads position per query rather than as a site total.
For a page, a healthy average sits close to the position of its main query. An average far below the best query means the page shows for many searches it ranks poorly for: either an opportunity to cover them properly, or a sign the page is vague about what it is for.
How should you actually read average position?
Treat it as a per-query diagnostic and never as a headline metric. A few habits make it reliable.
- Look at one query at a time. Filter the report by an exact query, or export the query table and read each row. A single query’s position over time is close to a true ranking trend.
- Compare like with like. Hold the query set fixed across the two periods you are comparing. If the query list is different, the averages are not comparable.
- Split by device and country before drawing any conclusion. Mobile and desktop results differ, and a store selling in the UK and US will rank differently in each, so a blended figure can move because the traffic split moved.
- Always pair it with impressions. Position without impressions tells you nothing about whether the move matters.
- Let clicks settle arguments. If clicks to the pages you care about are rising, a sagging average position is almost certainly mix.
Key takeaway
The sitewide average position line on the chart is the least useful number in Search Console. The same metric filtered to a single query is one of the most useful. Everything else in this post is a way of getting from the first to the second.
What this costs to run
Reading position properly by hand is free and slow. Export the query table for two periods, match the rows, and compare each query against itself. For a store with a few hundred ranking queries it is an evening the first time and an hour after that.
The cost is doing it again every month, next to deciding what to publish and what to refresh. That loop is what QueryClimb exists to take off your hands. Search Insights tracks positions per query rather than as a blended average, so a new article’s long tail does not read as a drop. The content calendar turns the striking distance queries into a plan, the blog writer drafts against them, and internal linking points your existing pages at the new ones so they climb faster. The pricing is public.
If you have the hour each month, the manual method gets you the same answer. Our blog SEO system covers how the per-query view feeds the weekly publishing loop.
Frequently asked questions
What does average position mean in Google Search Console?
It is the highest position a link to your site held in Google’s results, averaged across every impression in the period you selected. Each search contributes only your topmost result, and searches with more impressions carry more weight in the average.
Why is my average position going down?
Usually because your site started appearing for more searches at lower positions, and those new impressions pull the average down. New content and seasonal demand both do this. Filter to the specific queries you care about and check whether each one actually dropped before assuming a ranking loss.
Is a higher or lower average position better?
A lower number is a better position, since position 1 is the top. A lower sitewide average is not automatically a better site, though, because losing deep rankings improves the average while shrinking your visibility. Judge it per query, alongside impressions.
Why is the average position different on the Pages tab and the Queries tab?
The Queries tab aggregates by property, counting only your topmost result for each search. The Pages tab aggregates by page, giving every URL its own position. When two of your pages appear for the same search, the query view shows the higher one and the page view shows both.
What is a good average position in Google Search Console?
There is no meaningful good number for a whole site. For a single query, positions 1 to 3 earn most of the clicks, and positions 5 to 20 with healthy impressions are where improvements to an existing page tend to pay off fastest.