Whatever You Call It, AI Search Traffic Engages More Than Any Other Channel

There is a loud debate happening on LinkedIn right now about whether GEO is a real discipline or just SEO with a new name.

Google recently published a guide titled "Optimizing your website for generative AI features on Google Search," and the punchline in their own words is that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO." A wave of SEO professionals took a victory lap. Andrii Shum at SeoProfy posted that the document "officially ends the GEO-vs-SEO debate." Rahul Marthak at sneo.ai wrote that "Google just officially said stop doing this and half the SEO world wasn't listening." Lily Ray, Founder of Algorythmic and VP of SEO & AI Search at Amsive, summarized her take to her audience: "SEO is still the foundation for AI search."

On the other side, voices in publications like Wired and Singularity Digital continue to argue that AI search is its own discipline, with its own tactics, and dismissing it as "just SEO" misses the point.

Cyrus Shepard, longtime SEO voice and former Head of SEO at Moz, is reasoning from a different angle entirely. In a recent LinkedIn post he asked whether AI citations are "worth pursuing" and answered yes, despite noting that AI citation CTRs are "significantly lower than regular search results." His framing is not about whether the tactics belong to SEO or to a new discipline. It is about whether the work pays off either way.

I am not going to take a side on the tactics question. I do not have a particularly strong view on whether llms.txt does anything, whether content chunking matters, or whether buying mentions to influence AI visibility is worth it. Google says no. Other people say yes. The tactical experiments will play out in their own time.

But there is a related, more concrete question I can answer with our own data: regardless of which optimization camp is right, are AI search visitors actually worth the effort?

I run Clickport, a privacy-first web analytics product. Our customers track their websites with a cookie-free script. We measure every channel the same way, and the data sits in our own database. We pulled 180 days of aggregate engagement metrics from across the Clickport platform and looked at how AI search traffic behaves compared to every other channel.

The answer is unambiguous. Whatever you call this category of traffic, and whatever you do to acquire it, the visitors it sends you are more engaged than the visitors from any other channel. By a lot. And more consistently than any other channel. The data does not settle the optimization-tactics debate. It does answer the value-of-traffic question.

What we measured, and how

The numbers below are aggregate engagement metrics computed on the Clickport platform over the trailing 180 days. The engagement score is a 0-100 composite of scroll depth and time on page, with bot traffic filtered out at ingestion. Every figure is reported as a percentage, mean, or standard deviation. The tracker is cookie-free, so every session is independent and there is no visitor identity to slice. There are no raw session counts, no per-site numbers, and nothing here traces back to any individual site or customer.

The headline: AI Search is in a tier of its own

Here is the cross-channel engagement score on the Clickport platform over the trailing 180 days.

The mean engagement score for AI Search on the Clickport platform is 60.8 out of 100. The next-best channel, Organic Search, comes in at 44.3. That is a 37% gap. Direct traffic (people who type your URL or click from a bookmark) sits at 23.9. Organic Social, despite the social-buttons-on-every-page energy of modern marketing, lands at 35.9. Referral traffic from other websites comes in at 43.6.

In our scoring scale, AI Search is the only channel that lands solidly in the green band. Organic Search and Referral are middle-of-the-pack orange. Organic Social and Direct are bottom-quartile.

I want to be careful here. The mix of sites on the Clickport platform is not a representative sample of the entire web. A different mix might shift the absolute numbers up or down. But the relative ranking of channels is what most people care about for actionable purposes, and that ranking is unambiguous in the data.

And the engagement is consistent, which no other channel is

The mean is one half of the story. The other half is the standard deviation: how much engagement varies between sites.

Standard deviation matters here because it tells you whether a channel is reliably good or just sometimes good. A channel with mean 50 and std dev 5 is the same on every site you measure. A channel with mean 50 and std dev 20 is the same on average but actually varies from 30 to 70 depending on which site you look at. The second case means you cannot really predict what a new site will see.

In our data:

• AI Search has the lowest cross-site standard deviation of any channel: 7.5. Wherever AI Search has enough traffic to register, engagement stays in the green band (above 50). Even at the low end of the range, AI Search clears that threshold.
• Organic Search has more than double the variance: std dev 16.2. On some sites, Organic Search engagement is excellent. On others, it is mediocre or worse. There is no consistent floor.
• Every other channel sits in between.

This is the part I find most interesting. Organic Search engagement on a given site is a coin flip. It depends heavily on what queries happen to find that site, what keywords the search engine matched, and whether the landing page was the right answer to the search. Sometimes you get the perfect query-to-page match and Organic Search is fantastic. Sometimes you get a flood of off-topic queries and your Organic Search visitors bounce.

AI Search does not behave that way. Across every site in our dataset with enough AI Search volume to measure, AI Search visitors arrive engaged. The variance is small. The floor is high.

Why I think this is happening

I want to be careful to call this a hypothesis, not a proof. With the dataset I have, I cannot conclusively isolate the mechanism. But the mechanism that fits the data best is this: LLMs filter for relevance before the click in a way that traditional search engines do not.

When somebody types a query into Google, the search engine matches the query against its index using a combination of keyword signals, link signals, freshness, authority, and a hundred other things. The output is a ranked list of pages that are probably relevant. The user then has to click and judge whether the page actually answers their question. Sometimes the page does. Sometimes it does not. The click happens before the relevance check.

When somebody asks ChatGPT, Perplexity, Gemini, Copilot, or Claude a question, the LLM does something different. It reads (or summarizes from its index of) candidate pages, reasons about which ones actually answer the user's question, and writes an answer that cites the pages. The user reads the answer first. Then, if they click through to one of the cited pages, they are clicking because they want to verify, to read more, or to take an action. The relevance check has already happened. The click is a deliberate next step, not a guess.

That is the prewarmed effect. AI search visitors arrive past the awareness stage. They have already had their basic question answered. They click because they want something specific, and the page they land on is one the LLM has already determined is genuinely relevant. So they engage. And the engagement does not depend on how well your specific site's keywords match a specific query, because the LLM was matching intent, not keywords.

That is also why the variance is so much smaller for AI Search than for Organic Search. Organic Search is a keyword-matching system, and keyword matching is noisy. AI search is closer to a relevance-reasoning system, and relevance reasoning is more uniform.

I do not think this should be controversial. It is the same observation other people have made from different angles. Search Engine Land reported that LLM traffic converts at consistently higher rates. Microsoft's Bing Webmaster blog acknowledged that AI search changes how conversions get measured. Adobe measured lower bounce rates. What I think our data adds to that picture is the engagement-score lens and the cross-site consistency finding. The conversion-rate story has been told. The consistency-of-engagement story has not, as far as I can tell.

Every major AI engine looks similar

One question I had before pulling this data was whether the effect was a ChatGPT effect specifically, or an LLM effect generally. The answer in our data is unambiguous: every major AI engine delivers similar engagement.

Every major engine delivers traffic that engages above the 50-point threshold. ChatGPT dominates volume, which is unsurprising given its market share. But Perplexity, Gemini, and Copilot are not far behind on engagement, and Copilot actually outperforms ChatGPT on a per-session basis (though with a much smaller sample, so treat that ranking as directional).

The takeaway is not "optimize for the highest-engagement engine." The takeaway is that whichever LLM the visitor came from, they will engage. So if your strategy is to get cited in any of them, you do not have to overthink which one.

About 29% of AI Search sessions arrive with the referrer stripped, which usually means the visitor was inside an LLM's mobile or desktop app and the app does not pass the source URL through. We classify these as AI Search based on UTM parameters or other tracker hints, but we cannot tell which specific app sent them. The engagement on these sessions is slightly lower (consistent with mobile app sessions across all channels), but still above the engagement floor of every non-AI channel.

A note on volume: small, but growing fast

The most important caveat to all of this is that AI Search is still a small share of total traffic. On the Clickport platform, it averages around 1.5% of all sessions. Compared to Organic Search (around 30% on most sites) and Direct (often the largest single channel), AI Search is a sliver.

But the trend is steep. Over the trailing three months, AI Search share has roughly quadrupled. The starting point was a fraction of a percent. The current state is a small but real fraction. The trajectory points toward AI Search becoming a meaningful channel within the next year or two, not five years out.

This is consistent with what some of the largest publishers are already telling their teams. Rand Fishkin highlighted in a recent LinkedIn post that Condé Nast CEO Roger Lynch has directed all of the company's brands to operate as if search traffic to their properties will be zero. In Lynch's own words from a TBPN interview: "Every year, our search traffic was down more than we had forecast. So last year I told our teams, 'Assume there's no search.' You have to have your businesses planned as if search is zero. We don't expect it to be zero, we expect it to be a single-digit percentage of our traffic." If a publisher the size of Condé Nast is planning around a single-digit percentage of search traffic as the new baseline, then a small but consistently high-engagement channel like AI Search is exactly the kind of channel worth taking seriously now rather than later.

The combination is unusual. A channel that is currently small but high-quality, with rapid growth, is the kind of thing that rewards early investment. By the time AI Search is 10% of traffic on most sites, the early movers will already have established AI-citable content. The cost of catching up later is usually higher than the cost of getting in front of the trend now.

This is, incidentally, where the GEO-vs-SEO debate becomes practically relevant. Whatever tactics turn out to work for getting cited in AI search results, whichever camp is correct about the methods, the visitors that arrive when those tactics succeed are objectively the most engaged stream any site is currently receiving. So the value of the optimization work is higher than the volume share would suggest.

What "engaged" actually looks like in practice

Mean engagement of 60 is a number. What does it actually look like in terms of visitor behavior?

The two visitor types are not just slightly different. They are different in kind. An AI Search visitor reads an article. A Direct visitor lands and leaves.

The mechanism explains both. AI Search visitors arrived because an LLM told them this specific page would answer their question, so they read it. Direct visitors arrived because they typed a URL or clicked a bookmark, which is often just opening their banking site or a tool they use repeatedly. The destination usually does not need to be read; it needs to be used.

Desktop is doing most of the work

A descriptive note worth including, even though it does not change the headline. AI Search traffic is more heavily weighted toward desktop than Organic Search is. On Clickport, roughly 86% of AI Search sessions come from a desktop browser. For Organic Search, it is closer to 75%. The remaining share is mobile and tablet.

This makes sense intuitively. People use ChatGPT, Perplexity, Gemini, and Copilot heavily on desktop computers during research-mode sessions. They are at a laptop, working on something, asking questions. They click through to read in depth. Mobile AI search exists (and is growing), but the desktop-research use case still dominates.

For marketers, this has a practical implication. If you are optimizing AI-citable content, you can lean a bit more heavily into desktop-friendly formats: longer articles, tables, code blocks, structured data. The visitors are mostly arriving on screens where those things render well.

What about conversions?

I want to address this explicitly because the question will come up. Other studies (including some I linked earlier) have reported that AI search traffic converts at three to five times the rate of organic search. We see similar directional signals in our data, but I am not going to lead this piece with a multiplier number. Here is why.

Conversion rates depend on how each site's owner defines a conversion. A site that tracks form submissions has different conversion semantics than a site that tracks newsletter signups, which has different semantics than a site that tracks button clicks. When the definitions vary that much, aggregate conversion rates become hard to interpret cleanly. Some sites do not define conversions at all, which means their sessions contribute to a denominator without ever contributing to a numerator. The math produces a number, but the number is misleading.

What I can say honestly is that where conversions are tracked on the Clickport platform and AI Search has enough traffic to measure, AI Search converts at a higher rate than Organic Search. The lift varies by site, but the direction is consistent. It is not 5x in every case. It is sometimes 2x, sometimes more. The engagement story is cleaner because engagement is uniformly tracked. The conversion story is real but messier.

If you want the conversion-multiplier story, other studies have published it. I would rather show you the metric we measure cleanly, which is engagement, and let you draw your own inference about conversions.

What this means for your site

Three practical takeaways, in rough order of importance.

Track AI Search as its own channel, not bucketed with Organic Search. Most analytics tools (including older Google Analytics setups) lump AI search referrals into Organic Search or Referral. This obscures the quality difference. If you cannot see AI Search separately, you cannot make decisions about it. Clickport classifies AI Search as its own channel by default. Other privacy-friendly analytics tools have started doing the same. GA4 can be configured to surface AI sources but requires manual work. We covered the attribution mechanics in detail in How to Track AI Search Traffic.

Even at a small volume share, AI Search is worth optimizing for. A channel that delivers two times the per-visitor engagement of your next-best channel is worth more than its volume share would suggest. A 1.5% AI Search share with 60+ engagement is worth more than a 5% share of a 25-engagement channel. The math is in your favor.

The optimization-tactics debate is unresolved, but the value-of-traffic question is not. Whether GEO is its own discipline or just SEO, whether you should write llms.txt files or not, whether content chunking matters, the visitors that succeed in arriving at your site from an LLM citation are the most engaged stream you currently have access to. Whatever you do that leads to more AI citations, it pays off more per visitor than every other channel.

A short note on what this piece is not

To be clear about what this article is and is not:

• It is not an argument for either camp in the GEO-vs-SEO debate. I do not have a view on whether llms.txt, content chunking, or AI-targeted formatting work as tactics. People who care about those tactics will figure it out empirically over the next year or two.
• It is a data observation. Visitors who arrive from AI search engines, however they arrive, behave differently from every other channel. That observation is independent of which tactics produced the visitor.
• It is compatible with both camps. If your view is that AI search optimization is just good SEO, this data fits. Solid SEO that earns LLM citations delivers higher-engagement traffic than any other channel. If your view is that AI search is its own discipline, this data also fits. Visitors that respond to GEO-specific tactics engage at a fundamentally different level than visitors from any other channel.

Both camps can use this data without endorsing the other. That is intentional. The numbers do not pick a side.

How to see this on your own site

If you want to know what AI Search engagement looks like on your own traffic, the simplest path is to install an analytics tool that classifies AI Search as a distinct channel and surfaces engagement scoring per channel. Clickport does this by default. If you are already using Clickport, the Sources panel breaks AI Search out separately, and the Engagement column tells you how it compares to your other channels on the same site.

The setup is a script tag and a five-minute install. There is no plugin to install, no PHP, no database overhead, no cookie banner. Within a few days you will have your own version of the table above, for your own site.

You may find that AI Search is already a meaningful share of your visitors and you just have not been measuring it. You may find it is still a sliver. Either way, looking at the engagement column next to it is usually the moment where the channel stops feeling abstract and starts feeling like a thing you actively want more of.

That is what happened to me when I first saw it in our own data. The headline number was striking. The cross-site consistency was the part that made me write this piece.