The ranking jump that wasn't: Google's September 2025 change
In the second week of September 2025, two unrelated DIWYM clients had the best fortnight of their Google reports. A Decatur family practice's average position went from 14.2 to 8.0. A Hartselle hometown pharmacy's went from 19.7 to 8.5. Those are the weekly averages for the week of September 1 and the week of September 15, 2025.
Neither business had changed anything on its website. No new pages, no new content, no technical changes. A Decatur equipment rental company's report showed the same pattern, going from 21.7 to 8.2 over the same two weeks. Three unrelated businesses, three different websites, one fortnight.
If you are an agency, that is a gift. A chart that goes from 14 to 8 in two weeks looks like a successful campaign. We could have put it in a report and taken a bow. We didn't, and this post explains why.
It is one chapter of our Search Everywhere series, built on a field study of three DIWYM clients. The wider findings are in our pillar post, what three North Alabama businesses taught us about search in the AI era.
What actually happened
Around September 10 to 14, 2025, Google stopped honoring a setting called num=100. Until then, a tool could ask Google to return up to 100 results on one page instead of the usual 10. Google did not announce the change. The industry dated it from the behavior it saw, and Search Engine Land and SEO Sherpa both covered it that month.
Why does that matter to a local business? Because a lot of what Google counts as an "impression" and a "position" came from automated tools, not customers. Rank tracking software, including the kind agencies use to check where you show up, was loading those deep result pages. Each load could count as an impression for sites sitting far down the list, and those deep positions dragged average position toward the bottom.
When the setting stopped working, those tools stopped loading deep results. The impressions they generated disappeared. Fewer low-position impressions meant the average position of the remaining ones looked better. That is our read of what happened, and it matches what the industry reported. Your site didn't move up. The count of where you were seen changed.
What the larger data showed
Search Engine Land analyzed 319 properties (websites in Google Search Console) after the change. Of those, 87.7% lost impressions and 77.6% lost ranking keywords. A change that hits nearly nine in ten sites at once, regardless of what each one did, is not a ranking event. It is a measurement event.
That is why we think the right label for what happened to our clients is a reporting artifact. Our three unrelated clients behaved the way the 319 properties did, and the timing lines up with the date the industry gave for the change.
Why we refuse to count it
We do not credit our work for the September 2025 position jump, and we do not present it as growth. There are three reasons.
It would be false. The practice and the pharmacy changed nothing that week. Taking credit for a change we didn't cause is claiming a result we don't have.
It would poison every later comparison. If we counted September 2025 as the baseline for "improvement," then anything we did afterward would be measured against a number that was never real.
You would eventually find out. If you ever hire someone else, or look at your own history in Search Console, the jump is right there. An agency that says "we did that" has handed you a reason not to trust the rest of the report.
So here is the rule we follow. We never compare average position across September 2025, and we never chart it as performance. Where our own studies cross that date, we use measures it didn't touch.
Which numbers it affected, and which it didn't
Clicks were not affected. A click is a real person tapping through to your website. Rank trackers don't click. Clicks measured before and after mid-September 2025 can be compared. For that reason, most of our seasonal analysis uses clicks.
Impressions and average position were affected. Both included the activity of tools that stopped loading deep results. Don't treat a before and after comparison that spans the change as meaningful.
There is an uncomfortable side effect, and we owe you the disclosure. In our study, the year-over-year impression comparisons use windows that start July 23 and end October 14, in 2025 and 2026. The 2025 half of that window includes days before the change. Our read is that the 2025 counts were probably inflated, which means the real growth is likely understated. We say so beside the numbers. You can see the comparison in seen more, clicked less.
What the three reports had in common
Look at the pattern across the three reports, because the pattern is the evidence. A Decatur family practice, a Hartselle pharmacy and a Decatur equipment rental company have different customers, different websites and different search terms. The family practice's site was a WordPress build, and the other two were built on Wix. The practice and the pharmacy had changed nothing that week, and the rental company's report showed the same step.
When three unrelated businesses move together in the same two weeks, the likeliest explanation is something outside all three. That is the logic we apply to any result, ours or anyone else's: ask what else could explain it before you ask who gets the credit.
It also shows why a single business's chart can mislead you. Looked at alone, a drop from 14.2 to 8.0 is easy to believe. You can build a story around it, and it will feel true. Only when you hold it against other sites does it stop being a story and become a question.
How to spot it in your own Search Console
Check today, in about five minutes.
Open Google Search Console and go to Performance, then Search results.
Set the date range to the full 16 months, which is as far back as Search Console goes. Mid-September 2025 is still visible for now.
Turn on Total impressions and Average position.
Look at the chart around September 10 to 14, 2025. Many sites show impressions stepping down, and average position stepping up (a lower number), in the same few days.
Hover over the days before and after. If clicks, in contrast, stayed level, you are looking at a counting change.
If your chart shows the step, congratulations on having ordinary data. Now check any report you received for fall 2025. If it describes that jump as an improvement, ask what was changed on your site that week.
What to ask your marketing provider
Does your report show average position across September 2025? If so, how is it handled?
Which metrics did you use to measure progress? Clicks, calls and direction requests are not touched by the change. Impressions and position need a caveat before and after mid-September 2025.
What did you change, and when? A real result has a cause you can name. Our practice example is the opposite: nothing changed, and the line moved anyway.
For the wider story of how Google's results have changed over the last two years, see the AI search timeline every local business owner should know. For how to read the rest of your report, see how to read your marketing report in 2026.
What this means for you
Search data is not as solid as the charts make it look. A reporting change can look like a win, and a win can hide behind a reporting change. The honest response is to measure what a real customer did (clicked, called, asked for directions) and to say plainly when a number moved for reasons that have nothing to do with your business.
That is also how we work with our own clients, through our local SEO and AI enhancements: we report the measures that are not affected, we say when a number can't be compared, and we don't take credit for a change we didn't make.
Frequently asked questions
What was Google's num=100 change?
Around September 10 to 14, 2025, Google stopped honoring the num=100 setting, which let tools request up to 100 results on one page instead of 10. Google did not announce it; Search Engine Land and SEO Sherpa dated it from what they observed. Tools that loaded deep result pages stopped doing so, and our read is that this changed the impression and position counts in Search Console for many sites, including our clients'.
Did my site's ranking actually improve in September 2025?
Probably not because of anything you did. A Decatur family practice's average position moved from 14.2 to 8.0, and a Hartselle pharmacy's from 19.7 to 8.5, between the weeks of September 1 and September 15, 2025, with no changes on either site. Our read is that the counting changed, not the ranking. Check whether your clicks moved too, because clicks are not affected by the change.
How many websites were affected?
Search Engine Land analyzed 319 properties in September 2025. Of these, 87.7% lost impressions and 77.6% lost ranking keywords. That is a large share of sites changing at the same time, whatever each one had done, which is why we treat it as a measurement change and not as a ranking change for any single business. Your own Search Console chart will likely show the same step in mid-September 2025.
Why won't you count it as a win?
Because we didn't cause it. Two unrelated clients, with no changes on their sites, showed the same jump in the same fortnight, and a third client's report showed it too. Counting it as growth would be false, and it would make every later comparison unreliable. We never compare average position across September 2025, and we say so in our reports so that you can trust the rest of the numbers.
Which metrics can I still trust across September 2025?
Clicks. A click is a person tapping through to your site, and automated tools don't click, so clicks before and after mid-September 2025 can be compared. Business Profile calls and direction requests come from Google's profile data, not from Search Console's results counts, so this change did not touch them. Impressions and average position were affected, so treat any comparison that crosses mid-September 2025 with real caution.
How do I check my own Search Console for the change?
Open Performance, set the range to the full 16 months, and turn on impressions and average position. Look closely at September 10 to 14, 2025. Many sites show impressions stepping down and average position improving in those days while clicks stay level. If yours does, the apparent improvement is a counting artifact, and any report that called it a win deserves a polite question about what was actually changed.
Keep reading
What three North Alabama businesses taught us about search in the AI era: the pillar post for this series.
Seen more, clicked less: how we compare impressions and clicks, and why windows matter.
The AI search timeline every local business owner should know: the dates behind the changes.
How to read your marketing report in 2026: nine metrics and five questions to ask.
Zero-click search: why businesses get fewer visits even when they rank #1: another reason rankings and visits don't line up.
If a report you've received shows a big jump in the fall of 2025, we will help you work out what is real. Call (256) 274-1289 or message us for a free visibility check.