The screenshot era of AI visibility is ending

For the last year, proving AI visibility has mostly meant taking screenshots.

You type your brand into ChatGPT. It mentions you. You screenshot it. You put it in a deck. Everyone nods. The client feels seen.

Then you type the same thing the next day and you are gone. Or a competitor is there instead. Or the answer is subtly wrong about what you actually do.

That is not measurement. That is a mood board.

And it is ending, because two things landed in the same week that, between them, change what serious AI visibility work looks like.

The first is that Google is starting to show you the data.

Search Console is rolling out dedicated performance reports for generative AI. Visibility across Google’s AI surfaces, including AI Overviews and AI Mode, broken out separately while still counting in your overall totals. It is a limited rollout to a subset of sites for now, and it is impressions-only at this stage, not a full citation-quality report. But the direction is the thing. This is first-party data. Not an agent screenshot. Not a third-party estimate. Google telling you whether you showed up in its AI answers.

That matters more than it sounds.

It means “are we visible in AI” stops being a vibe you argue about and starts being a number you can check. The same way SEO stopped being astrology the day Search Console showed you impressions and clicks. The grown-ups in the room have been waiting for this. It drags AEO out of the séance and onto a dashboard.

So far, so good for anyone who likes evidence.

But here is the second signal, and it is the one that should make a lot of people put their “rank number one and you’ll get cited” pitch back in the drawer.

Ahrefs re-ran its analysis of where Google’s AI Overview citations actually come from. They looked at 863,000 keyword searches and four million AI Overview URLs. The finding: only about 38% of the URLs cited in AI Overviews rank in the top 10 for that same query. An earlier version of this study put it closer to 76%. So the link between “I rank for this term” and “I get cited in the AI answer for this term” has roughly halved.

Read that again, because it quietly demolishes a popular assumption.

You can be the number one organic result and not be in the AI answer. You can rank nowhere near the top 10 and get cited anyway.

The reason is query fan-out. When the AI builds an answer, it does not just look at the page for your exact search. It explodes the question into a cluster of related sub-questions, runs those, and pulls sources from across the whole spread. So the thing that gets you cited is not your rank on one keyword. It is whether you show up, clearly and credibly, across the family of questions around the topic.

And then there is the detail that should make every marketer sit up. The single most-cited domain in that AI Overview data was YouTube. Video accounted for 5.6% of all citations, and 18.2% of the citations that do not rank in the top 100 for the query at all. A platform most B2B businesses treat as an afterthought is quietly feeding the machine.

Put the two signals together and you get the actual state of play.

AI visibility is becoming measurable. And the old method of getting it is wrong.

This is the bit I want to land, because it is where most of the noise lives right now.

There is a whole cottage industry forming that wants to sell you AI visibility as either a mystery or a trick. The mystery crowd says it is unknowable, dark-arts, trust-us-we-watch-the-prompts. The trick crowd says it is simple, just add this file or hit this number and the AI will love you.

Both are wrong, and the data above is why.

It is not a mystery, because Google is literally building you a report. It is not a trick, because ranking, the most obvious lever, only explains a minority of citations now.

What is left is less exciting and far more useful. It is operational.

If you want to be in AI answers, you have to be legible and credible across the cluster of questions a buyer might ask, in the places the machine actually pulls from. That means clear pages that answer real questions, not one keyword stuffed forty times. It means proof a machine can extract, not proof buried in a PDF or a testimonial graphic. It means showing up in third-party sources and comparison content, because the AI cross-checks. It means treating video as a real surface, with titles, descriptions and transcripts that actually say what you do, not three minutes of ambient brand music. And it means watching the first-party data now that it exists, instead of arguing about screenshots.

None of that is glamorous. It is the same lesson as last week, and the week before. Plumbing first. Evidence second. Magic never.

For anyone selling marketing services, this reframes the offer, and it reframes it in a way that is honest.

The weak version of AEO is “we’ll get you mentioned in ChatGPT,” sold on a screenshot and a prayer. That was always going to age badly, and now there is data showing why.

The strong version is closer to a visibility audit and a maintenance loop. Where do you currently appear when AI systems answer the questions your buyers ask? When you appear, are you described correctly? Which sources does the AI lean on, and are you in them? Which competitors keep showing up that should not? What is missing, buried, or too vague to be extracted? And now, crucially, what does the first-party Google data say when you can actually see it?

That is a brief I can stand behind. It does not promise to control the model. It promises to make the business as legible and well-evidenced as possible, then measure what happens, mark what is inferred, and keep tuning.

For business owners, the takeaway is simpler than the mechanics.

Stop asking your agency for the screenshot. Start asking three questions. Can you measure where we show up in AI answers, now that Google is starting to report it? Do you understand that ranking alone will not get us cited, and do you have a plan for the wider question cluster? And are we treating proof, third-party mentions and video as part of this, or just rewriting the same blog?

If the answer is a confident speech about prompts and a screenshot, be careful.

The screenshot era was always a placeholder. It was what we did while there was nothing better to point at.

Now there is something better to point at. A real report on one side, real data on how citations actually work on the other.

Which means the excuse for hand-waving is gone.

Measure it. Understand the mechanism. Fix the evidence. Stop performing visibility and start earning it.

The dashboard is arriving. Best to have done the work before it shows everyone what you look like.

---

Jason Sibley is the founder of Cleo, a post-agency marketing and AI company. JasonVsTheNoise is where he writes about what is actually happening with AI, marketing, and how businesses should be thinking about both.