How I Took a Page From 2K to 20K Daily Impressions in 6 Weeks (And Why the Clicks Didn’t Follow Yet)

A large share of this visibility now arrives through Google’s AI features. Here is what that means for clicks, and what it doesn’t.

One page went from roughly 2,000 to around 20,000 daily impressions in about six weeks. Total impressions reached 969K, but clicks stayed at 631, a 0.1% CTR. A big reason for the gap is that much of the new visibility came through Google’s AI features, where the answer is often shown inside the result. The impression counts. The click often does not.

969K

Total impressions (main Search report)

131K

Impressions from Generative AI features

631

Total clicks / 0.1% CTR

One honest note before anything else. Google’s Generative AI features report currently shows impressions only. No clicks, no CTR, no queries, no position. So I can show the correlation and explain the likely mechanism, but I cannot prove that every missing click was taken by an AI answer. Anyone who claims they can prove that with this report has not read the report.

What Actually Happened

This is a real page I worked on. To keep the client’s details private, I’m focusing on the pattern rather than the site name, because the pattern is the useful part.

For weeks the page sat flat. Somewhere around 2,000 daily impressions, sometimes less. Then in early June the line changed direction and kept climbing. By August it was regularly touching 20,000 impressions a day, with the occasional spike higher.

That is a large move for a single page. But the interesting part is not the spike. It is the second graph. When I opened the Generative AI features report inside Search Console, 131K of the total impressions came specifically from AI surfaces. That line started from near-zero on June 3, which is the day Google launched that report.

So two things were true at once. The page was more visible than ever, and a meaningful share of that visibility was coming from AI-generated results, not only the classic blue links.

Reading the Two Reports Side by Side

It helps to separate what each report is telling you, because they measure different things and people mix them up constantly.

The main report told me the page was earning far more visibility and almost no extra clicks. The AI report told me where a large slice of that visibility was coming from. Neither report alone explains the story. Together they point to something specific.

Impressions are not the outcome. They are a signal that Google is choosing to show the page. The real question is what happens next, and whether that visibility turns into anything a business can use.

Why AI Impressions Can Rise While Clicks Stay Flat

Here is the mechanism as I read it. Treat this as interpretation, not a confirmed Google statement.

When a page appears inside an AI Overview or AI Mode answer, Google records an impression. But the AI result often answers the user’s question directly on the page. If the user gets what they came for, there is no reason to click through. The visibility happened. The visit did not.

That is not a failure of the page. In many cases it is the opposite. The page was clear enough, structured enough, and trustworthy enough to be pulled into the answer in the first place. That is exactly what you want a page to do in AI search. It just doesn’t show up in your click count the way it used to.

There is another factor worth naming. High impressions at low CTR can also mean the page is showing for broad or loosely related queries where intent is weak. Without query data in the AI report, I can’t fully separate these two causes. Both can be true at the same time.

What made this page retrievable in the first place

The work behind the jump was not a single trick. It was the ordinary, unglamorous stuff that makes a page easy for a system to understand and reuse:

• Entity clarity. The page made it obvious what it was about, who it was for, and how it related to the wider topic.
• Clean structure. Clear headings, direct answers near the top, and information laid out so a machine can extract it without guessing.
• Factual, specific content. Concrete details rather than vague filler, which gives an answer engine something worth citing.
• The page lined up with the rest of the site and the business’s information, so the signals reinforced each other instead of competing.

None of that guarantees AI visibility. But it makes a page a far better candidate for it. In my experience, retrievability usually comes down to whether a system can understand and trust the page quickly, not whether it contains a clever keyword.

What This Data Can and Cannot Prove

I want to be careful here, because this is where most AI-visibility content overreaches.

So the honest summary is this. The correlation is strong and the mechanism is reasonable. But the current report doesn’t give me the query, click, or position data I’d need to prove causation cleanly. I’d rather tell you that than dress up a correlation as a certainty.

What This Changes About Measuring SEO

For years, the simple mental model was impressions lead to clicks lead to conversions. That chain is loosening. A page can now be highly visible, genuinely useful, and still send fewer clicks than its visibility suggests, because the answer is delivered before the click.

That doesn’t make impressions meaningless. It makes them incomplete. If you judge this page on clicks alone, you’d call it a disappointment. If you judge it on visibility and citation inside AI answers, it’s doing something valuable. Both readings are partly right, which is exactly why you need more than one number.

Here is how I’d adjust the way you measure a page like this:

1. Track AI impressions separately. Use the Generative AI features report as its own line, not blended into total impressions.
2. Watch the click line underneath the spike. A rising impression graph with a flat click line is a signal to investigate intent, not a win to celebrate.
3. Tie SEO back to business outcomes where you can. Calls, forms, bookings and enquiries still tell you whether visibility is doing real work.
4. Accept partial answers. Until the AI report matures, some of this stays interpretation. Report it that way.

Common Mistakes People Make With This Data

• Celebrating the impression graph and ignoring the flat click line. The graph going up and to the right feels like success. It isn’t, on its own.
• Blending AI impressions into total impressions. That hides the exact thing you want to understand.
• Claiming proof the report can’t support. Impressions-only data cannot prove AI answers stole your clicks. It can only suggest it.
• Assuming low CTR always means AI. Broad or loosely matched queries can produce the same pattern. Both can be happening together.
• Treating AI visibility as worthless because it didn’t send traffic. Being the source an answer engine pulls from is its own kind of result, even when the click doesn’t happen.

The Practical Takeaway

Visibility and clicks are drifting apart. This page is a clear example. It earned far more visibility, a large share of it through AI surfaces, and the clicks did not follow at the same rate.

If you’re seeing a similar pattern, don’t panic about the click count and don’t oversell the impression count. Open the Generative AI features report, look at it on its own, and ask a more useful question than how many impressions did I get. Ask whether the page is being shown for the right reasons, whether it’s being cited in answers, and whether any of that is turning into calls, enquiries, or sales.

The definition of a result is changing. Being the answer, not just the link, is becoming part of what SEO delivers. We just need to measure it honestly while the tools catch up.

Read More: 568K Generative AI Impressions in 28 Days: What This Tells Us About AEO and the Next Phase of Search