A Google Search Console impressions drop looks alarming on a performance chart. The line falls off a ledge, the client asks what happened, and the instinct is to look for an algorithm update or a technical fault. For a large share of sites since mid-September 2025, the explanation is neither. The data changed, not the rankings.
The tell is in the other lines on the same chart. If clicks are flat and average position has improved while impressions have collapsed, you are looking at a measurement change. This post explains what changed, what two independent analyses found, and how to tell a reporting artefact from a real loss.
What the num=100 change actually was
For years, Google accepted a &num=100 parameter on results URLs. Add it and you got 100 results on a single page instead of ten. Rank trackers and scrapers relied on it, because one request returned the whole top 100 for a keyword.
Those automated requests were counted as impressions. If a tracker loaded 100 results and your page sat at position 74, Search Console could record an impression for you, even though no human ever scrolled that far. In mid-September 2025, Google stopped supporting the parameter. Trackers now have to page through results ten at a time, and the deep impressions they used to generate largely disappeared from the reports.
So the drop is real in the sense that the numbers are lower. It is not real in the sense that matters to revenue: fewer people did not see your pages. Arguably, the new figures are closer to what humans actually see than the old ones ever were.
What two analyses found
The most useful summary comes from Edge Marketing, an Australian agency, in a piece updated on 13 October 2025: its analysis of the num=100 removal in Australia. It reports its own campaign data alongside an earlier analysis by LOCOMOTIVE Agency.
| Measure | LOCOMOTIVE Agency (319 sites, Sept 2025) | Edge Marketing (100 Australian campaigns, Aug–Oct 2025) |
|---|---|---|
| Sites with impression declines | 87.7% | 83% |
| Sites losing unique ranking terms or queries | 77.6% | 70% |
| Average impression change | Not given here | −49.6% (weighted portfolio −54%) |
| Average position | Not given here | Improved by +1.1 |
| Clicks | Not given here | Remained stable |
| Hardest hit | Short- and mid-tail keywords | Not given here |
Two different samples, two different countries, the same shape. Most sites lost a large share of impressions and a chunk of their visible ranking queries, while the traffic that pays the bills did not move.
Why average position “improved”
The +1.1 improvement in average position is the part that confuses people most, and it is the clearest evidence that this is a reporting change. Average position is an average across every impression. When thousands of impressions at positions 30, 50 and 90 vanish, the average is pulled towards the positions that remain, which are the ones near the top.
Nothing moved up. The bottom of the distribution was removed. The same logic explains the lost “unique ranking terms”: queries where you only ever appeared on page five were only visible because a bot loaded page five. Without the bot, those queries never register an impression, so they drop out of the report.
LOCOMOTIVE’s finding that short- and mid-tail keywords took the biggest hit fits this too. Those are the terms rank trackers check most often, and the terms where most sites sit deep in the results rather than on page one.
How to diagnose a Search Console impressions drop
Before you blame an update, run through these checks in order. Each one takes minutes.
- Find the date. Set the performance report to compare the four weeks before mid-September 2025 with the four weeks after. If the drop lands in that window, num=100 is the first suspect.
- Look at clicks. If clicks are flat or within normal variation, your human visibility did not fall. This is the single most important check.
- Look at average position. An improvement alongside falling impressions points to deep impressions being removed, not to pages climbing.
- Filter by position. Compare queries where you ranked on page one with queries where you ranked deeper. If the loss sits almost entirely in the deep queries, that confirms it.
- Check the click-earning pages. Sort pages by clicks and compare the top twenty across both periods. If they held, the business held.
If clicks fell too, and the fall lines up with a different date, you have a real problem and should treat it as one. Our guide to the Google spam update covers how to check whether a confirmed algorithmic change is the cause, and reading a core update covers broader ranking shifts.
Report clicks, not impressions
Edge Marketing’s advice is to prioritise clicks over impressions, and it is right. Impressions were always a soft metric. They counted appearances nobody saw, and the num=100 change simply exposed how much of the number was machine traffic.
If you report to clients, annotate September 2025 on every chart and explain the change once, in writing. Then reset the baseline: compare post-change months with post-change months. A year-on-year impressions comparison that spans the change will look like a disaster for the next twelve months, and it will mean nothing.
The same applies to rank-tracking dashboards. The trackers relied on the parameter that was removed, so treat any sudden shift in tracked positions around that date with the same suspicion until you have checked it against Search Console clicks.
The limits of this explanation
The honest limitation is that num=100 explains the pattern, not every individual site. A site can lose bot impressions and real rankings in the same month, and the first can mask the second. The two analyses above describe averages across samples; neither can tell you what happened to yours.
That is why clicks matter more than any explanation. If clicks are steady, the drop is cosmetic. If clicks fell, keep investigating, and look at which pages and which queries lost them. You can also compare against sites that kept growing: in the database explorer, filter to your niche and look at the before-and-after traffic figures for recent confirmed updates. That is a method for spotting a sector-wide shift, not a diagnosis of your site.
Finally, do not try to win the old impressions back. They came from bots loading page ten, and no amount of optimisation will restore them. Spend the time on the queries where you already sit close to page one, because those are the impressions real searchers see. More on measurement sits in the data archive.
Frequently asked questions
Why did my Google Search Console impressions drop in September 2025?
Most likely because Google stopped supporting the num=100 parameter in mid-September 2025. Rank trackers used it to load 100 results at once, and those requests had been counted as impressions.
Does an impressions drop mean I lost rankings?
Not necessarily. If clicks are stable and average position improved, the drop is a reporting change. If clicks fell as well, investigate further.
Why did my average position improve when impressions fell?
Deep impressions at low positions were removed, so the average now reflects mainly the positions near the top. Your pages did not move up.
The takeaway A September 2025 impressions cliff with flat clicks is the num=100 change, not a penalty. Annotate the date, reset your baseline and report clicks.
