In mid-September 2025, many site owners opened Search Console and panicked. Desktop impressions had dropped sharply, sometimes by half or more, in just a few days. At the same time, average position suddenly looked much better. A lot of people assumed a Google algorithm update had hit them.
It had not. What changed was a small technical detail called the num=100 parameter. Google stopped supporting it, and that broke how rank-tracking tools collected data. Because those tools had been quietly creating impressions in Search Console for years, the impression numbers fell once they stopped.
Let me explain what num=100 did, when it stopped working, why your impressions dropped and your average position "improved," and how to report it honestly to a boss or a client. If you want to see what real updates happened around the same time, our Google algorithm updates timeline lists them all.
The num=100 removal at a glance
| Item | Details |
|---|---|
| What changed | Google stopped honoring the &num=100 URL parameter that showed 100 results on one page |
| When | Around 10–14 September 2025 |
| Type | Search results/technical change (not a ranking update) |
| Announced? | No blog post; Google confirmed when asked |
| Google's statement | "The use of this URL parameter is not something that we formally support." |
| Main effects | Lower desktop impressions in Search Console, better-looking average position, fewer keywords reported, disrupted rank trackers |
| Clicks | Largely unaffected, though some sites saw smaller drops |
| Who was affected | Almost every site using Search Console, and every SEO tool that scraped Google |
What the num=100 parameter actually did
Normally, a Google results page shows about ten organic results. For years, though, you could add &num=100 to the end of a Google search URL, and Google would show 100 results on one page instead.
Google removed the visible setting for this in 2018, according to Search Engine Land. But the URL parameter kept working. And that was very useful for one group in particular: rank-tracking tools.
A rank tracker checks where your site ranks for a keyword. With num=100, it could load one page and see the top 100 results in a single request. Without it, the tool has to load up to ten pages to see the same 100 positions. That is ten times the work for the same data.
Here is the part most people did not realize. When a tool loaded those 100 results, Google could count it as an impression for every site shown, even the site in position 87. Real people almost never scroll that far. Bots did, many times a day.
When num=100 stopped working
The parameter stopped working around 10–12 September 2025. Search Engine Land reported that it stopped "late Thursday night," and SEOs noticed within days as their tools started showing gaps. Zeo's analysis dates the desktop impression drops to "around September 10-12, 2025," while Found's write-up of 17 September 2025 put it at about 10–14 September.
Google did not announce it in a blog post. When asked, it gave Search Engine Land a short statement:
“The use of this URL parameter is not something that we formally support.”
— Google, as quoted by Search Engine Land, Google Search confirms it does not support the results per page parameter
Zeo also reported that Danny Sullivan said the change was intentional, aimed at simplifying search and curbing automated tools. Either way, the outcome was the same. One-page access to 100 results was gone.
Why Search Console impressions dropped
This is the bit that caused the panic. Once rank trackers could no longer load 100 results at once, they stopped "seeing" sites in deep positions. And since those bot views had been counted as impressions, Search Console impressions fell, especially on desktop, where most scraping tools run.
Zeo described it as "an artificial drop caused by bot-generated impressions being removed from the report." Found pointed to Brodie Clark and others, who suggested that "many impressions logged in GSC were from bots or scraping tools rather than human users."
Zeo also reported two wide-scale figures: 87.7% of sites saw total impressions decline, and 77.6% saw fewer unique keywords. Zeo does not name the original source for these numbers, so I treat them as an indication of scale rather than a precise measurement.
What about clicks? Zeo said clicks should be unaffected, because "clicks from human users continue as usual." Found was more careful: clicks "dropped in some cases, though to a much lesser extent than impressions." That matches what you would expect. Bots do not click, so most real traffic stayed the same.
Why average position suddenly "improved"
Average position in Search Console is an average of the positions where your site appeared, weighted by impressions. If lots of impressions at position 60, 80 or 95 disappear, the average naturally moves closer to the top.
Zeo calls this "a mathematical result of the metric change," and that is exactly right. Your pages did not climb. The bad positions simply stopped being counted.
Example (hypothetical): imagine a page with these monthly impressions before September 2025:
| Position range | Impressions before | Impressions after | Clicks (both periods) |
|---|---|---|---|
| Positions 1–10 (average 5) | 2,000 | 2,000 | 120 |
| Positions 11–20 (average 15) | 1,000 | 800 | 10 |
| Positions 21–100 (average 50) | 7,000 | 200 | 0 |
| Total | 10,000 | 3,000 | 130 |
Before: the weighted average position is (2,000×5 + 1,000×15 + 7,000×50) ÷ 10,000 = 37.5. CTR is 130 ÷ 10,000 = 1.3%.
After: (2,000×5 + 800×15 + 200×50) ÷ 3,000 = 10.7. CTR is 130 ÷ 3,000 = 4.3%.
Impressions fell 70%. Average position looks more than three times better. CTR more than tripled. And the business got exactly the same number of clicks. Nothing real changed. This is why you have to read these metrics together.
How rank trackers were affected
For SEO tools, this was a real problem, not just a reporting quirk. Search Engine Land reported that rank-checking tools were badly disrupted and that many notified their customers. Zeo listed the main knock-on effects:
- Data beyond the first few pages became harder to collect and may now be limited.
- Tracking 100 positions now means loading up to ten pages, which costs more.
- Some tools may track fewer keywords, or fewer positions per keyword, which reduces long-tail coverage.
Found also noticed anomalies in third-party trackers, especially beyond page one: missing data and odd jumps. If your rank tracker showed strange movement in mid-September 2025, check with the vendor before you assume anything happened to your rankings. Our comparison of SERP volatility trackers explains how these tools measure changes, and our guide to tracking keyword rankings accurately covers what to ignore.
Is it num=100 or a real ranking drop?
Real updates did happen around this time. The August 2025 spam update was still rolling out into September. So you need to tell the difference. This table helps:
| Signal | num=100 reporting change | Real ranking drop |
|---|---|---|
| Impressions | Drop sharply, mostly desktop | Drop on mobile and desktop |
| Clicks | Stable or slightly down | Down in line with impressions |
| Average position | Improves | Gets worse |
| CTR | Rises | Flat or falls |
| GA4 organic sessions | Stable | Down |
| Conversions from organic | Stable | Down |
For a full method of ruling out tracking issues, read our guide on whether a drop is an update or a tracking bug. And for other Search Console oddities during update periods, see Search Console data quirks to know when analyzing updates.
How to annotate and report the change
If you report SEO numbers to anyone, this change needs explaining. Otherwise, year-on-year impression comparisons will look terrible until late 2026. Here is how I would handle it:
- Add an annotation in Search Console and your reporting dashboard for 10–14 September 2025, labeled "num=100 removed; impressions and average position not comparable." Our guide to Search Console custom annotations shows how.
- Split desktop and mobile in your reports. Found recommends this, and it shows clearly where the drop sits.
- Lead with clicks, sessions and conversions, not impressions. Zeo recommends putting more weight on clicks, sessions, conversions and revenue.
- Stop comparing impressions and average position across the break. Compare after-September periods with each other instead.
- Cross-check with GA4. If organic landing-page traffic and conversions are stable, the drop is a reporting change.
- Use wider date ranges and smooth trends rather than reacting to daily figures.
- Ask your rank-tracking vendor how it adapted, and whether it now tracks fewer positions.
Zeo suggests adding a note to reports saying that "metrics may be skewed due to a change in Google's data collection method as of September 2025." I think that is a sensible line to copy.
If you are new to Search Console reports, our Learn SEO lessons on Google Search Console and SEO reporting explain the basics.
What num=100 taught us about impression data
The biggest lesson is uncomfortable. For years, a meaningful part of the impressions many sites saw was not people at all. It was bots. That means older impression trends, especially for keywords where you ranked on page three or beyond, were inflated.
In practice, that means three things for how you read older reports:
- Old impression peaks may be partly bots, especially for keywords where you ranked beyond page two.
- Old average positions were probably worse-looking than reality for real searchers, because bot views at deep positions dragged them down.
- Keyword counts were inflated by terms you only "ranked" for at positions no human ever saw.
The data after September 2025 is likely closer to what real searchers saw. That is good for decision-making, even if it hurts the charts. Clicks were always the more honest metric, and they still are.
This is not the first time a display or data change was mistaken for an update. In 2021, Google's title link update changed headlines without touching rankings. And just weeks before num=100 disappeared, Preferred Sources started changing what individual users see in Top Stories. For a full view of the year, see our 2025 algorithm updates review.
Questions people ask about the num=100 change
Did the num=100 removal hurt my rankings?
No. It changed how impressions were counted and how tools collect data. Your rankings did not change because of it.
Why did my average position get better?
Because low-ranking impressions that came from bots stopped being counted. The average moved up mathematically, not because your pages climbed.
Will impressions go back up?
Not to the old level, as long as Google keeps the parameter disabled. Treat September 2025 as a new baseline for impressions.
Why did my clicks drop a little too?
Found reported small click drops for some sites. These may be from other factors happening at the same time, such as real updates or AI Overviews. Check GA4 and look at specific pages.
Should I change rank trackers?
Not automatically. Ask your vendor how it adapted. Some now track fewer positions or charge more for deep tracking. Choose based on the keywords and depth you really need.
Reporting through the num=100 change with confidence
In brief: the impression drop of September 2025 was a reporting change, not a penalty. Annotate it, split desktop and mobile, lead with clicks and conversions, and stop comparing impressions across the break.
If you are not sure whether your drop is num=100 or something more serious, our SEO services team can review your data. If you suspect a real penalty, see our penalty recovery service, or simply get a quote.
Google updates around the same time
Before this update:
- Google's August 2025 Spam Update: A Nearly Month-Long Rollout (Aug 2025)
- Google Preferred Sources (August 2025): Readers Pick Publishers in Top Stories (Aug 2025)
After this update:
- Google's December 2025 Core Update Explained (Dec 2025)
- Google's February 2026 Discover Core Update: The First Discover-Only Core Update (Feb 2026)
Or browse every Google update on one timeline, newest first.
Sources for this num=100 removal guide
- Search Engine Land: Google Search confirms it does not support the results per page parameter
- Zeo: The impact of num=100 parameter removal on SEO reporting
- Found: September 2025 Google Search Console impressions drop; what's really happening? (17 September 2025)
- Search Engine Journal: Google algorithm history
