On September 12th, the SEO community discovered that Google had disabled the ability to use the parameter &num=100, designed to enable 100 results as the default display rather than 10.
The consequence of this change was the impact on various rank tracking providers, with many relying on this parameter in order to deliver ranking data in an efficient way, now having the potential to increase the cost of retrieving a single result by 10 times.
An unexpected result of this change was a significant change to Google Search Console data, with impressions declining dramatically for many of the properties that I have access to, also corroborated by the broader SEO community.
When digging deeper into the change at the query and country level, the data wasn’t really adding up, which made me think that this is either a data discrepancy or a change to reporting in GSC. But the timing of the change to the disabling of &num=100 and the GSC reporting anomalies lined up too closely.
While I joked that GSC was a customer of DataForSEO, others, such as Mark Barrera and Malte Landwehr highlighted the connection between the spike in impressions in recent times and scraping from SEO tools.
Has “The Great Decoupling” that many of us have taken as fact, instead been related to the excessive increase in scraping from SEO tools?
Looking at the GSC graph shown above, one of the early “decoupling” screenshots that was shared, which first kicked off the discussion around impressions increasing due to AI overviews, highlighted that the change first started around September of 2024, with other screenshots showing the increase at the beginning of 2025.
Google pushed back on many of the studies that highlighted that around the time of AI overviews ramping up as a SERP feature, this was when impressions first started to skyrocket, yet clicks weren’t following the same trend.
Instinctually, this made sense, and any pushback from Google felt disingenuous, considering that there were no promises being made of introducing an AI-related search appearance filter in GSC.
In retrospect, and seeing the impact that the disabling of &num=100 has had on SEO tools, the introduction of an AI-related filter in GSC may have only further reinforced what many suspected about the impact of AI overviews, but this conclusion may have still been inaccurate in some capacity.
When looking at the Semrush Sensor tool, it hasn’t been updated since September 10th (5 days ago), which is quite unusual for this tool. When checking my Semrush rank tracking projects, there are similar discrepancies where rankings have disappeared, and no data is showing.
The date when the drop in impressions first occurred in GSC was also on September 10th. While the datasets are distinctly different (one is from Google directly and the other is external), it doesn’t appear to be a coincidence that the changes occurred around the same day.
Theories to support the increase in bot impressions in GSC
When looking at the screenshot at the top of this post, which mirrors many other sites that I have access to, the decline in impressions in GSC is primarily from desktop, with mobile being impacted to a lesser extent.
For most SEO tool tracking projects, many will use desktop tracking as the default project setting. I don’t know exactly what the split would be among desktop and mobile, but I would expect that desktop would be much higher than mobile, with some tools only offering desktop tracking anyway.
The part of this situation that really blows my mind is the sheer number of impressions that were supposedly coming from bot impressions. In the example shown, the decline on desktop alone goes down by 200K+ daily impressions, which is honestly hard to believe that bot impressions could be this impactful.
Note: it is worthwhile highlighting that I did notice that not only impressions but also clicks in GSC (to a far lesser extent) were also impacted during this same timeframe since &num=100 was disabled. The Press Gazette recently put out an article detailing new research on the growth of AI bot traffic over time, likely having a connection to the change in clicks, where Google seems to be better at blocking or filtering out from their dataset.
The difficulty in comparing this data to earlier periods is that bot impressions will have always been consistent for sites in some capacity. But the ramping up of bot impressions has likely come about due to the increased competition within the AI Search space over this past year, which does line up closely with the original theories around The Great Decoupling.
Can’t help but wonder the connection between these two pic.twitter.com/1zwnOPqfXW
— SERP Alert ⚡️ (@SERPalerts) September 14, 2025
Many suspect that the disabling of &num=100 is an intentional move to reduce external tracking services abusing Google’s systems, with there even being the potential that the recent news that ChatGPT is using SerpApi is another reason for this.
Another theory is that Google wants to put the bad press to rest regarding the impact of AI overviews. However, the impact of this argument becomes difficult to measure if you only have access to 16 months of, likely heavily influenced, bot impressions.
Explanation of the measurement change in GSC
For an impression to be recorded within GSC for an organic listing, the listing itself needs to appear on the current page, and the link needs to be visible to users. The listing itself, if located on the current page, doesn’t actually need to be scrolled into view in order to register an impression.
For a normal search result for a typical Google user, the default ‘current page’ would be 10 organic listings. Because rank trackers tend to use 100 organic listings as the default view, a result that ranks in position #99 would then incur an impression within GSC, even though it is a bot impression.
Average position in Google’s AI Mode takes more of a sharp-left compared to the zig-zag of a traditional search results page.
This approach tends to be more simplified in mobile search results (where the sequence is clear), with there being rules put in place for desktop that… pic.twitter.com/XAX6DbcwHn
— Brodie Clark (@brodieseo) June 23, 2025
This is a topic that I’ve written about extensively in the past, where I’ve done experiments for how data is recorded within GSC for both AI Overviews and within AI Mode. With there often being reasons for a significant impression spike without clicks, our GSC data has unfortunately been polluted with a lot of bot impressions over the past year, which has reduced the reliability of the dataset.
Here is a list of some of the most popular SEO tool providers and how they have responded to now being unable to use &num=100 to efficiently deliver the top 100 results.
Semrush: an update has been published that encourages users to consider the value behind both the top 10 and top 20 results. When checking position tracking data, it seems like Semrush is continuing to track rankings beyond position 20 as normal, but the wording within the published update makes it sound like this will likely need to change.
Ahrefs: Tim Soulo has hinted at effectively the same change as Semrush in a Twitter post, which has been shared to gain feedback from users on the value that they would receive through having rank tracking data beyond position 20. It appears as though Ahrefs data hasn’t been disrupted in the interim, also, based on a review of their data.
SerpApi: a solution has now been discovered from a prominent API provider for being able to scrape 100 results from a single request rather than 10. This has been described as the Google Fast Light API.
Important GSC bot impression stats
Research from the SEO community following Google disabling &num=100 suggests that bots have been a lot more impactful than we otherwise thought. Here are some stats related to studies that have been completed following the changes:
- 87.7% of GSC properties experienced a drop in impressions following the change. – Tyler Gargula, LOCOMOTIVE (319 website dataset)
- Average position declined significantly for positions 21+ directly following the change. – Matthew Mellinger (100+ website dataset)
- The median impression loss shows a ~15% decline, with large websites having the highest concentration of the biggest impression drops. – Paul Grillet, ThotSEO (1334 website dataset)
- There was an aggregate drop of 25% in impressions for the week after the changes were made. – Serge Bezborodov, JetOctopus, Barry Schwartz (1,000 website dataset)
Note: many rank trackers are starting to readjust their approach, so impressions will start to rise again. The most important datasets assessing the change are the ones that directly follow the decline (around September 13th), which will likely be the most accurate depiction of impressions and average position in GSC, according to how the data is measured and collected.
What outcomes can we expect?
Google is yet to comment publicly on this situation, but I suspect that they will need to given how impacted their dataset has been. If they don’t roll back the &num=100 change, then it is likely that rank tracking costs may soar, or we will be left with a top 20 result set (instead of top 100) as the default.
In terms of GSC data, it will be interesting to see how it is altered over the coming days and whether this dataset can be used to make more sense of the impact of AI overviews. It is important to note that Google has always been upfront with their stance on scraping since its inception.
My expectation is that this is very much an evolving situation, so stay tuned, and I’ll be sure to update this post once we receive official confirmation from Google or if I discover anything new that is worthwhile highlighting. At this stage, this does appear to be a permanent change, especially considering Google has published an ad with the job title: Engineering Analyst, Anti-scraper, Search.
Do I think that AI overviews weren’t actually impactful in a negative way? Not necessarily, but I do think that the 1st party data that we were working with was unfortunately not particularly accurate, if ‘Impressions’ and ‘CTR’ over the past year were the primary metrics used to support these conclusions.
Update: a Google Spokesperson has now commented on the situation:
“The use of this URL parameter is not something that we formally support.” – Google Spokesperson
Here are my final thoughts on the impact of &num=100 being disabled:
SEO Update: Final update on the &num=100 situation.
We have now received a statement from Google related to the &num=100 parameter being disabled, with the response being somewhat unsurprising.
While @rustybrick was seeking feedback on both the rank tracking implications and… pic.twitter.com/kh38nsweRV
— Brodie Clark (@brodieseo) September 19, 2025