AI search demands a new audience playbook

For publishers, one of the observations that’s often cited about AI search is that the people who click through are more intentional than those who come from traditional search. In other words, sure, AI might be nuking your referral traffic, but at least the people coming from there are more likely to engage, and potentially become loyal readers.

And that’s true—the stats show it. But it’s an oversimplification of a more interesting reality. It turns out that the audience in AI search isn’t just a blob of traffic that you need to work extra hard to get the attention of.

People who ask AI portals for information about something can have wildly different intentions, and those intentions can change as they do more research, often in the same AI conversation. Yes, clicking through and engaging is still the prize, but how to get there requires understanding the journey before that happens.

What the funnel looks like now

One of the better attempts I’ve seen at mapping the different parts of the “blob” is this study from Scrunch, an AI search analytics company. While its report is oriented toward brands selling products rather than to publishers (it focuses on AI search behavior around GLP-1 medications like Ozempic), I believe the insights about how different types of information seekers use AI search translate very well.

The study classifies AI users into different buckets based on intent. There are knowledge seekers, who are curious but not yet committed; evaluators, who are comparing options; access seekers, who are ready to buy; and then a couple of other categories (side effect navigators and regimen planners—remember this is about medications), which I think can be combined into “post-decision” users.

Importantly, each category has a different likelihood to convert and a different tendency to move into other categories. As you might expect, those deliberately seeking to buy or transact are very valuable for conversion, but it’s a relatively small group (just 9% of the overall “conversation”). Because evaluators are a much larger group (20%), they may offer more aggregate opportunity than the smaller, higher-intent access seekers.

The audience inside the audience

It’s one category and one snapshot, so the exact buckets are not universal. However, I think the taxonomy translates well into publisher audiences.

1. Orientation readers: These are readers who may be new to a topic or are just seeing the latest information. They want to understand the basics about a story or topic, or simply get caught up on the latest news about it.
2. Evaluation readers: These readers want to go beyond the surface. They’re deliberately asking for more analysis and different perspectives on a particular topic. Remember, they’re still doing all this in an AI service.
3. Action readers: These readers have a clear picture of something they want to do and are seeking guidance on taking that action, or a place to do so.
4. Support readers: This group has already taken action and wants some kind of ongoing support with their area of interest.

In the case of publishers, conversion isn’t necessarily just readers buying third-party products (which would be relevant only to sites that do affiliate marketing) but engaging in a deeper, deliberate way: subscribing to premium content, signing up for a newsletter, downloading an app, buying an event ticket, and more.

Each of these groups is seeking a different kind of information, and publishers need to respond with different kinds of content to reach them. This isn’t entirely new—it’s similar to a content model the BBC and others adopted more than a decade ago—but in an AI world, it matters whether large language models (LLMs) can cite the content with confidence.

For example, an orientation reader may be wondering whether the electric-car market is shrinking or growing, whereas an evaluator may be comparing the coverage in The Wall Street Journal with competitors or publications deep in the niche, like Electrek.

An action reader might go straight to “Which EV newsletter should I subscribe to?” or “What’s the best site to follow for EV policy and pricing?” This helps explain why niche and B2B publications often punch above their weight in AI search.

Why the citation still matters

But the key point is this: Concentrating only on the people who click through from AI search is too narrow a focus. Readers go from one state to another within AI search. So even if you don’t capture readers in every search where you’re cited, the fact that it’s your narrative guiding them, with attribution, can begin their orientation toward your publication.

The publishers that understand this will need to offer content across different types of readers to stay present across their journey. Being present in more answers will of course increase the chance of scoring a visitor, but it will also keep your brand visible as the reader continues exploring, even if that journey is largely taking place in an AI platform.

If the Scrunch report helps map the journey inside AI, another example helps explain what publishers should think about once a reader does arrive.

On3 is a college sports outlet and network of more than 70 team fan sites. Like many publishers, it’s faced declines in traffic from search and social, so it chose to focus on revenue per session as a key metric. It uses AI recommendations and first-party data to keep readers moving among articles, forums, video, email, and commerce.

Most publishers may not have the resources to offer all of that, but the underlying point still applies: In an AI world, the click matters less as a finish line than as the start of a carefully managed next step.

This of course assumes the content is AI-friendly in the first place. Writing informational passages in a straightforward way, repeating common questions about the topic and answering them directly (there’s a reason you’re seeing FAQ sections everywhere), and practicing good technical maintenance are all important to get right.

Where visibility becomes value

So, yes, while AI kills traffic, it also functions as an audience qualifier, with the people who click through being the most likely to engage with your publication.

What isn’t obvious is that as a publisher you also have an active role in that qualification. You can influence what audiences see in those summaries, but it depends on being present, understanding the different types of readers, and offering them the right mix of valuable content around your core topics, structured so the bots can parse it easily.

In AI search, the publishers that win will be the ones shaping reader intent before they ever win the click.

View the full article