Search Console Average Position vs a Rank Tracker: Why the Numbers Disagree

Average position in Google Search Console and the position in a rank tracker often disagree, and the gap causes more worry than it should. Search Console shows a keyword on page one, the rank report shows it lower on mobile or missing in another country, and someone asks which one is wrong. Usually neither is: Search Console averages what searchers saw, while a Google keyword rank check records one result for a fixed location, device and date.

Usually neither. The two numbers measure different things. Search Console averages what real searchers saw, wherever and whenever they searched. A rank check records where a keyword stands on one Google engine, on one device, on one date. Once you know how each is built, the disagreement becomes useful information instead of noise.

How Google defines average position in Search Console

Google’s help page on the Performance report defines average position as the average position of “the topmost result from your site”. In the chart it covers your whole site; in the table it is calculated for each query, page, country or other grouping you choose.

The companion page on impressions, position and clicks explains how that average is built:

• It is averaged over impressions. Each time your site appears in a search, the position of your topmost link is recorded, and Search Console averages those values.
• No impression, no position. A link must be seen for its position to be recorded. Google’s own example is a result on page 3: if the searcher never goes past page 1, no position is recorded for that search.
• It mixes places and devices. Unless you filter by country or device, the average combines every search in the period. The Performance report help notes that results are specific to the time, place, device and recent history of the person searching.
• Some blocks are counted in a particular way. An AI Overview or a carousel occupies a single position, and ads do not occupy a position at all.

So average position is a weighted summary of many different results pages. It is real data, but it is an average.

What a rank tracker measures instead

A rank check asks Google the same question under fixed conditions. In our reports that means a live check through a Google results API: one query per keyword per engine, at the depth you chose, with desktop and mobile as separate queries. An engine is one Google edition on one device, for example google.co.uk on mobile, and a city can be added as its own engine, checked from a point inside that city.

The result is one position per keyword and engine, on a known date, with no searcher history involved. When the site is not in the results examined, the report says so: “not in top 30” (or top 60). In our public sample report, 251 of the 400 keyword rows are reported that way. Search Console would have nothing to say about most of those searches, because a site that isn’t seen gets no impression.

Why the numbers disagree: a worked example

The figures below are an illustration, not client data. One keyword, 28 days of Search Console data:

Search Console’s average position is weighted by impressions: (600 × 3 + 300 × 5 + 100 × 14) ÷ 1,000 = (1,800 + 1,500 + 1,400) ÷ 1,000 = 4.7.

A rank check of the same keyword on one day of that period might return:

Both are correct. Search Console says the typical searcher, most of them on UK desktop, saw the page high on page one. The rank check says the page is 3rd on UK desktop, 6th on UK mobile that day, on page two in the US and not in the top 30 on US mobile. The US mobile weakness is invisible in Search Console, because nobody saw the page there.

Now watch the average move with no ranking change at all. Suppose UK desktop impressions double next month, from 600 to 1,200, because more people search in that season. Then (1,200 × 3 + 300 × 5 + 100 × 14) ÷ 1,600 = 6,500 ÷ 1,600 = 4.1. Average position “improves” from 4.7 to 4.1, and not a single position has changed.

Search Console averages what searchers saw. A rank check records where you stand. Read each one for the question it answers.

Six reasons the two numbers disagree

1. Places and devices are mixed. An unfiltered average blends every country and device where you appeared.
2. Impressions set the weights. Wherever most searches happen dominates the average.
3. Searches without an impression drop out. Weak rankings deep in the results often leave no trace in the average.
4. A period is not a moment. Search Console averages the days you select; a rank check records one moment.
5. Real searchers aren’t neutral. Their results depend on time, place, device and history. A rank check uses fixed conditions without a searcher’s history.
6. Positions are counted differently. Search Console gives an AI Overview or a carousel one position and ignores ads; any checker counts the page by its own rules, so small differences are normal.

Our guide on how to check keyword rankings accurately compares the ways of getting a ranking by hand and with tools, and where each one breaks.

When to use which

A rank check also gives you summary measures that Search Console doesn’t, such as a visibility score that counts non-ranking keywords as zero. Our article on the SEO visibility score explains how that number works.

How to use both together

The two sources are strongest side by side:

1. Filter Search Console to match your engines. Choose one country and one device in the Performance report before you compare it with the same engine in a rank report.
2. Build your keyword list from Search Console. Queries with steady impressions are the ones worth checking on fixed engines.
3. Check on a schedule. A check every month on the same engines and depth gives you comparable dates.
4. When the average moves, check the mix. If Search Console’s average position changes but the rank check is flat, look for a shift in countries, devices or queries before you look for a ranking problem.
5. Keep each number in its job. Report clicks and impressions from Search Console, and positions, SERP features and AI Overview citations from the check.

Our pay-per-check rank reports give you the second half of that routine: every keyword on four Google engines (your Google edition and google.com, each on desktop and mobile), with the check date on the cover and, on a repeat order, the change against your previous check. If you’d rather have someone read the report and turn it into actions every month, our SEO rank tracking service does that. Or order a single rank check and compare it with your Search Console data yourself.

FAQ

What is a good average position in Google Search Console?

Lower is better: 1 is the top position. But there is no universal good number, because a site-wide average mixes every query, country and device. Look at average position for individual queries, filtered to one country and device, and compare it with its own history.

Why did my average position improve when my rankings didn’t change?

Check the mix first. If more of your impressions come from searches where you rank well, the average improves even though no position moved. The worked example above shows a move from 4.7 to 4.1 caused only by more searches on UK desktop.

Why does Search Console show no position for some queries?

A position is only recorded when your link gets an impression. Google’s help gives the example of a page with 10 impressions on desktop and none on mobile: when you compare the two devices, the position value for mobile shows as a dash.

Is Search Console or a rank tracker more accurate?

Both are accurate for the question they answer. Search Console reports what real searchers saw, averaged. A rank check reports one position on a named engine, device and date. Use them together rather than choosing one.