Why Search Console Impressions Dropped in September 2025, and Why It Still Matters

This matters again now, a year later. Year-on-year comparisons are the standard way to separate seasonality from real decline, and for most of 2026 they compare against the inflated months before the change.

A site that lost nothing can look like it lost a third of its impressions. And a 16 month chart in Search Console still shows the September 2025 cliff, waiting to be misread.

What actually happened

Rank trackers and scrapers used to request 100 results per search. Every one of those requests counted as an impression for every site on the page, including sites ranking at position 60 or 90 that no person would ever scroll to.

• The parameter stopped working in the days around 12 September 2025.
• Deep impressions disappeared, mostly on desktop, because tools no longer loaded those results.
• One analysis of 319 sites found that close to nine in ten lost impressions in Search Console.
• Clicks did not change, because people had never been clicking those deep results in the first place.

The data after the change is closer to the truth. The data before it was inflated by machines, often for years.

Three metrics moved, one did not

Metric	What it did	Why
Impressions	Fell	Bot impressions at deep positions stopped being counted
Average position	Improved	The deep positions dragging the average down were removed
CTR	Rose	The same clicks divided by fewer impressions
Clicks	Unchanged	Real visitors were never affected

CTR is the one people forget. A report showing click-through rate improving since autumn 2025 may be showing arithmetic, not better titles.

A worked example

Take a page with a simple, made-up month of data.

• 200 impressions from real people, at an average position of 6, earning 20 clicks.
• 600 impressions from tracking tools, at an average position of 55, earning no clicks.

Before the change, Search Console reported 800 impressions, an average position of about 43, and a CTR of 2.5 percent. Nothing about that page was ranking 43rd for a real person.

After the change, the same page shows 200 impressions, position 6, and a CTR of 10 percent. Impressions fell by three quarters while clicks and real rankings did not move at all.

Every number changed except the one that pays the bills. That is the whole story of September 2025 in one page.

It landed in the middle of a spam update

The August 2025 spam update ran from 26 August to 22 September 2025. The parameter change fell almost exactly in its middle.

So many site owners saw impressions collapse mid-rollout and concluded the spam update had hit them. Most had not been touched by it at all. If your clicks held through September 2025 while impressions fell, the update was not your cause. Details of that rollout are in the August 2025 spam update.

Which comparisons are broken

• Any year-on-year comparison where the earlier period falls before mid-September 2025. Summer 2026 against summer 2025 is the common case.
• Any long trend line of impressions, position or CTR that crosses the date.
• Recovery baselines set before the change and judged after it.

The problem is closing on its own. From mid-September 2026, year-on-year comparisons put two post-change periods side by side again, and impressions become comparable once more.

How to compare safely

1. Use clicks for any comparison across the date. They were never affected.
2. Filter to queries ranking in the top ten if you need impressions. Real people see page one, so it was far less distorted.
3. Start impression trends from 13 September 2025 rather than earlier.
4. Annotate the date in every dashboard and report, so nobody rediscovers the cliff and panics.

Setting a baseline that survives changes like this is covered in how to set a recovery baseline.

How to explain it in a report

Managers and clients see the cliff before they read the notes. A short, plain explanation saves a lot of worried emails.

Something like: In mid-September 2025, Google stopped supporting a setting that rank tracking tools used to load 100 results at once. Those tools had been inflating our impression count. Impressions fell and average position rose as a result, but clicks are unchanged, so nothing was lost. From this point on, impressions reflect real searchers more closely.

Put that note beside any chart that crosses the date, and switch long-term comparisons to clicks until both periods sit after September 2025.

This is not your problem if

• Your clicks fell too. A real decline shows up in clicks, and needs its own diagnosis.
• Your drop started on a different date. Check daily data for the exact step.
• Only top ten positions fell. That is a ranking change, not a reporting one.

Measurement changes cause more misdiagnosis than most updates. When positions and clicks disagree, the reasons are laid out in rankings dropped but traffic is the same.

Other measurement changes worth knowing

September 2025 was not the only time the numbers moved. Since mid-2025, Search Console has counted AI Mode activity within its ordinary web search totals, so changes there can hide inside normal figures. Newer generative AI reporting is covered in no generative AI tab in Search Console.

Whenever a metric jumps with no change on your site, check for a reporting change first. It is the cheapest explanation to rule out.