It is annual review season, which means a lot of business owners are about to open Google Search Console, click the year-over-year comparison, and get a nasty surprise.
Impressions down, sometimes by half. Average position suspiciously better. Clicks flat.
Before you panic, or worse, before you make a budget decision based on that chart, you should know that the comparison itself is broken. Not your site. The comparison.
There is a seam in your data
On September 11, 2025, Google quietly switched off a setting called num=100. Anything in your Search Console history from before that date was measured one way. Everything after it was measured another way.
So any year-over-year comparison that spans September 2025 is not comparing your performance in two periods. It is comparing two different measurement systems and calling the difference a trend.
Right now, in October 2026, that seam sits squarely in the middle of every twelve month comparison you can run.
What num=100 was, and why it mattered
For years you could add a parameter to a Google search URL and get 100 results on one page instead of ten. Almost nobody did this by hand, but every SEO tool on the market did it constantly. If a rank tracker needs to know where your site sits for "roof repair Bellingham," the cheapest way to find out is to pull 100 results in one request and look for you in the list.
Here is the part that affected your data. Search Console was logging those automated requests as impressions.
Every time a tool pulled 100 results for a keyword, every site sitting in positions 11 through 100 was credited with an impression. Nobody looked at any of them. Multiply that across every SEO tool, every tracked keyword, every day, for years.
When Google removed the parameter, that entire layer of phantom impressions vanished overnight.
Which explains the weird pattern
| Metric | What it did | Why |
|---|---|---|
| Impressions | Dropped sharply, often on desktop especially | Automated impressions stopped being counted |
| Average position | Improved, meaning the number got lower | The impressions removed were mostly deep ones, position 40 or 70 or 90, which had been dragging the average down |
| Clicks | Barely moved | Software never clicked anything. Real people carried on as normal |
| Click-through rate | Went up | The same clicks divided by far fewer impressions |
If you pull up a sixteen month view and see impressions falling off a cliff while average position climbs in a near mirror image, both starting on the same day in September 2025, you are looking at this. It is one of the cleanest signatures in analytics.
So what should you compare instead?
Compare clicks, not impressions. Clicks were never inflated by this, because automated tools do not click. A year-over-year click comparison across the seam is honest. An impressions comparison is not.
Compare conversions above all. Calls, form submissions, booked jobs. These come from your own business rather than from Google's counting methodology, and they survive every platform change.
If you need an impressions trend, start it after September 2025. Compare this October to last November through this September. Shorter window, but at least everything in it was measured the same way.
Treat the average position improvement as an artifact, not an achievement. Your rankings did not suddenly get better in September 2025. The maths changed. If anyone has shown you that improvement as a result of their work, that is worth a conversation.
The bigger pattern worth knowing
This was not a one-off. Google has been steadily making its search results harder to collect at scale, and it has kept going. In August 2026 it started masking where search result links point, which is the same campaign aimed at the same target.
Each of these changes creates another seam. Which means the habit worth building is not memorising this particular one. It is getting comfortable asking a different question when a chart looks dramatic: did my performance change, or did the measurement change?
Most of the time, lately, it is the measurement.
One genuinely good piece of news
Your current data is better than your old data.
Fewer impressions, but the ones you have now reflect actual humans seeing your listing. Your click-through rate is a real number instead of one diluted by machines. That is an upgrade, even though it arrived looking like a loss.
The only thing it cost you is the ability to compare cleanly across one date in September 2025. Which is annoying for exactly one more year, and then it is behind us.
Looking at a chart that does not make sense? I am in the Sycamore Building in Fairhaven, and I am happy to look at your Search Console with you and tell you plainly which part is signal and which part is an artifact of how Google counts. Get in touch.
