Heard of "AI Overview"? If you are at home in SEO land or you have been at the Search Central in Zürich last December, you probably have. It is what everybody wanted to know about.

The AI Overview is in SEORank.tech.

The SEO World wants the "AI Overview"

There are (new) companies that sell you tools to analyze how AI sees your content. Sites like Peec AI, Scrunch, Surfer AI Tracker, and Profound are all trying to solve a real problem. How does AI see your content? They do it by reverse engineering. They construct prompts they think people are asking, fire them at ChatGPT or Perplexity or Google's AI, and observe whether your site shows up. It's educated poking. Sophisticated poking, sure – but poking. The very thing these tools measure shifts every time you look at it. They're trying to put a ruler on smoke.

We didn't build something like that. We built something simpler and we think more honest. We don't guess what people are searching. We know, because they already searched it, and Google Search Console recorded it, we just show you the data.

What are AI Queries?

AI-assisted search behavior can be identified by a distinctive pattern of conversational search queries, question-shaped, long queries.

My favourite is #6 "what does sic mean in an article", because I had been digging into that some time ago. I doubt, that we all typed the exact same long query into Google independently, about 217 times.

The most plausible explanation: AI agents are generating these phrases as follow-up suggestions, related searches, rephrased searches, or embedded links and many different users are clicking through with the identical string. Maybe Google aggregates slight variations of the same (different punctuation, minor word order shifts, typos) and decided they are semantically equivalent.

The query isn't coming from the human; it's coming from the AI that served them first, and Google is catching the ripple.

This observation alone is worth a dedicated research and post, the mechanics of how AI systems funnel users toward specific search strings, and what that means for how impressions accumulate on long-tail queries, is a pretty young area.

Whatever the exact mechanism, the impression count tells you that this particular phrasing has traction. There's a meaningful cluster of people arriving at that query through some shared upstream path. That's a signal worth acting on.

What SEORank's AI Overview actually does (and doesn't do)

Our AI Overview on SEORank.tech filters the GSC data for queries with initially 5+ words (you can adjust that). You can filter that down by searching within those queries, and you can see the impression counts aggregated over time. No black-box scoring, no AI-detection claims.

Just the data, and a philosophy: follow the big trends, not the tiny glitches. Long queries are a proxy for conversational intent, and impressions tell you where you already have traction before you even think about optimizing.

The AI Overview is not a tool to reverse engineer the exact prompts that are driving traffic to your site. It's a tool to identify the long-tail queries that are already driving impressions, and to understand the trends in how users are interacting with your content through AI-assisted search behavior. It's about leveraging the data you already have and to make informed decisions about your SEO strategy in the context of AI-driven search.

Derive Actions from the AI Overview

By filtering for "docker" I saw that there are a lot of queries for "docker without docker desktop", which is a topic I had written about and that seems to be of more interest than I had realized.

That's a specific, actionable signal: this page covers that topic well enough that Google (or an AI Overview) surfaces it for that kind of query. For your content you can now ask: is there a dedicated page for this? Is the explanation clear and concise enough to be quotable by an AI? Is there a way to make it more comprehensive, or to add a FAQ section that directly addresses this query? You can also look at the trend over time: is the interest in this topic growing, stable, or declining? That can inform your decision on how much effort to invest in optimizing for this query.

I added a filter for looking at the data in the "Queries" page in seorank I see that the interest used to be very high beginning of 2025, it is slightly picking up again. Also this page tells me that there are 1% of queries and just 1 page being found for it. I decide to drop this topic, there are so many things this blog is about docker is just one of them.

Optimize for Usefulness, not for citations

What these long queries reveal is intent. The intent of people who are already comfortable talking to AI systems and expect search to work the same way.

The best response to that isn't to stuff all your pages with FAQs and hope. It's to make your content genuinely precise, well-structured, and trustworthy. That's what gets you cited.

The tools that help you understand your readers are more durable than the tools that try to game an algorithm.

Finally

We built this tool because we saw tha lack of clarity and simplicity of looking at GSC data. You get tools telling you about made-up metrics like "visibility" etc. All these data just build on ground truths, which we see GSC data to be one of the best sources, so we try to make that more useful.

Ping me if you have questions, feedback, suggestions. Also if you have ideas for seorank or want to get a free trial for six months, just reach out via email or mastodon, see links below.