AI search influences more conversions than its traffic suggests

Across our client base, the conversions AI search touches are growing faster than sites' own traffic — and neither device, client type, nor journey position explains why.

To no one’s surprise, the use of AI in search and discovery continues to grow – and we have the numbers. Across the Knotch client base, direct AI search traffic, in other words, page views that came from someone clicking on a citation in an LLM, is growing roughly 3% per month. Yet the number of conversions that can be attributed to visitors who arrived on sites from a direct AI citation is growing approximately 7% per month. That’s more than twice as fast. No other marketing channel shows a larger gap. These numbers suggest that marketers are significantly underestimating AI search's influence when they evaluate it through traffic alone.

You might think this is a case of flat traffic inflating the gap – that conversions are outpacing views simply because site traffic is languishing. That’s not happening. Both views and conversions within this sample set grew through 2025, peaking in November before easing back as the year drew to a close, and most clients saw their own AI search traffic grow, too. The conversions that AI search touches are growing even faster than a traffic base that's simultaneously growing on its own — not outpacing a stagnant field.

It's not a device story

AI search's traffic skews heavily to desktop use, taking an 81% share compared to 19% mobile. That's in contrast to sitewide traffic, which is  46% desktop and 54% mobile. It’s worth pausing on this, because it raises an obvious question: is the device split behind the effect, rather than the channel itself? It isn't. If differences in device use explained it, the gap should show up far more on one than the other. It doesn't — conversion growth outpacing view growth holds on both desktop (+4.30 percentage points per month) and mobile (+2.90), at similar strength. Whatever is driving this, it isn't someone’s choice of screen.

This held up against everything we threw at it

We looked for any other potential explanation for this discrepancy in growth. We tested whether the pattern was concentrated among B2C clients — it wasn’t. We tested whether it was explained by where AI search tends to sit in the conversion journey, whether it's typically an early, middle, or late touch. It turns out that AI search’s mix is mid-pack compared to other channels, not distinctively early or late, though its share of last-touch conversions has ticked up slightly over the period. None of it explains the effect away. Most single-channel performance claims don't survive this level of scrutiny, but this one does.

It connects to something bigger

Part of why this is hard to explain with device or journey data alone is because some of it may be happening outside what referral tracking can see at all. Knotch Labs' related research on indirect LLM influence found that direct AI referral traffic is tiny — but in a survey of more than 20,000 respondents, 35% said they'd used an AI tool before visiting, most without ever clicking a tracked link. That lines up with what we're seeing here: The conversions we can measure may indeed be the visible, trackable edge of a much larger AI-influenced audience that mostly never shows up in referral data. Audiences using AI are often embarking upon major segments of their journeys before landing on your site, even if they ultimately click a citation. We can't confirm that mechanism with this dataset directly, but the two findings point in the same direction.

What this means for marketers

Keep investing in AI search visibility. But don't judge its value by traffic or last-click reporting alone; instead, track the audience journey. Knotch traces every touch to each asset across a visitor’s path, so you can see where AI search shows up along the way, not only where it happens to close.

But even that fuller view may not catch everything – so don’t assume that it does. If AI's influence extends well beyond what referral tracking can see at all, as our related research suggests, the real impact is likely much bigger than what any attribution model shows.

‍

Published on July 17, 2026