Why Your Average Position in Google Improved But Your Traffic Didn't

TL;DR

Google quietly removed the &num=100 parameter that SEO tools used to load 100 results per page. That change wiped out millions of bot-driven "impressions" for positions 20–100 in Google Search Console data. Result: average position suddenly looks better, impressions drop, and actual traffic often falls. Your site didn't suddenly rank higher - Google just stopped counting the robots.

The Mystery Spike

In September 2025, we started to notice a strange pattern in Google Search Console across dozens of sites we manage. For each site, the "average position" would shoot up - historically this was good news! Google is ranking your site higher.

However, there was often no corresponding increase in impressions or clicks. In some cases, they continued to fall, thanks to the AI-driven "crocodile effect".

So how could average position improve so dramatically without any real traffic lift?

After some digging (and excellent sleuthing by Dani Mora), the answer turned out to be a tiny bit of Google plumbing with a big knock-on effect.

The &num=100 Vanishes

For years, SEO pros and rank trackers relied on a simple trick: add &num=100 to the end of a Google search URL, and you'd get the top 100 results in one shot. It was a handy way to scrape large datasets quickly, tracking performance across many keywords.

In mid-September 2025, Google quietly retired that parameter. Now, search results are loaded continuously via infinite scroll and paginated API calls - no more "give me 100 at once."

In practice, many automated systems that depended on &num=100 suddenly stopped seeing anything beyond the first page.

Why Average Position Improved

Those bots were scraping content, but as they were generating pages of links, they were also being counted as searches by Google's own stats. Every time one of them "saw" your page at position 65 or 86, that impression was recorded, even though it wasn't a human.

When the bots stopped getting beyond page one, all those long-tail impressions vanished. That left a smaller, more human-generated dataset. This dataset is more heavily weighted towards your top positions.

So now:

• Higher average position - because only top-ranked, human-visible impressions remain.

• Fewer impressions - because bots stopped "seeing" you in positions 11–100.

The higher "average position" is deceptive. It's less that your site climbed the rankings, and more like a ladder that lost a few artificial rungs.

What Happens Next?

Some rank trackers will no doubt adapt to Google's new layout and resume polling multiple pages. But many of the older bots that once inflated impression counts will likely never return - scripts left running, teams moved on, dashboards unmaintained.

That's actually good news. We've lost a synthetically high number and gained a more accurate one.

It's not a satisfying answer when you're explaining to a client why their impressions KPI fell, but those impressions were never real users in the first place.

We've had similar conversations around analytics over the years: take over a site, fix the tracking, and suddenly engagement drops - not because users left, but because the data is now accurate where it wasn't before.

For now, we should continue to focus on clicks and conversions - the real humans. And if someone presents the "improved ranking" as proof of success, let them finish their coffee before breaking the news.

More on this over at Search Engine Land.