How LLM traffic converts compared with traditional search

You have probably seen the claim by now. Traffic from ChatGPT and Perplexity converts several times better than organic search, because the assistant has already done the comparison work and the person arrives ready to act.

It is a good story, and it may well turn out to be true. The problem is that the research does not currently agree with itself, and the disagreement is large enough that anyone quoting a single figure at you is quoting selectively.

What the studies actually found

Semrush estimates that an AI search visitor can be worth around 4.4 times more than an organic search visitor. Seer Interactive's data tells an even stronger story: 15.9% of ChatGPT referrals and 10.5% of Perplexity referrals converted, compared with just 1.76% from Google organic search. On those numbers, AI referral traffic is the best-converting channel most companies have never deliberately built for.

Then Amsive ran a paired statistical test across 54 websites over six months and found something different. Organic converted at 4.60 percent and LLM referrals at 4.87 percent. The median site-level difference was 0.09 percentage points, and the p-value came back at 0.794. In plain terms, the gap was indistinguishable from random variation.

An academic study went further still. Comparing more than 50,000 ChatGPT referral transactions against 164 million traditional ones across roughly a thousand ecommerce sites, it found ChatGPT visitors converted less often and spent less per visit than every major channel except paid social.

Three credible datasets. Three incompatible conclusions.

Why they disagree

The variation is not sloppiness. It comes from four things, and each one matters when you are deciding what to believe about your own account.

• Sector mix. Amsive found LLM traffic converting better for B2B sites, at 2.17 percent against 1.16 percent for organic, while B2C was almost identical. Financial services and travel showed higher LLM conversion, plausibly because conversational search suits complex, considered purchases.
• What counts as a conversion. A newsletter signup and a demo request are both conversions. They are not the same event, and datasets that mix them will not agree.
• Sample size at the tail. Many sites in these studies received a trickle of LLM sessions. A handful of conversions on a small base produces a spectacular percentage and tells you almost nothing.
• Attribution setup. Very few analytics installations separate AI referrers cleanly from generic referral traffic. Studies that rely on default configurations are measuring something fuzzier than they think.

SALT.agency analysed roughly 672,000 LLM referral sessions and found organic outperforming LLM referrals on engagement in most sectors, with health as a notable exception. The pattern across all of this research is not that one channel wins. It is that the answer depends heavily on who is asking.

The number nobody leads with: scale

According to the Amsive dataset, LLMs drove just 0.24% of total sessions but contributed to 0.42% of conversions. In fact for the early 90% of the websites analyzed, LLM traffic accounted for less than 0.6% of sessions. BrightEdge put AI search at under 1 percent of referral traffic in the fourth quarter of 2025, while growing 527 percent year on year.

Both halves of that matter. A 15 percent conversion rate on 0.3 percent of your sessions is a rounding error this quarter. The same channel growing at several hundred percent a year is a serious acquisition source inside two years. Treating it as either irrelevant now or decisive now gets you the wrong answer.

AI referral traffic is small, growing quickly, and converts somewhere between slightly worse and considerably better than organic depending on your sector, your offer and how you count. Anyone more confident than that is selling something.

The measurement problem underneath all of it

There is a deeper issue that none of the conversion studies fully solve. Most AI-influenced discovery never produces a referral click at all.

Zero-click queries rose from 56 percent to 69 percent following the rollout of AI Overviews. Someone reads an answer that names three providers, does not click any of them, and searches your brand name two days later. That session lands in your analytics as direct or branded organic. The AI answer that created it is invisible.

Researchers have started calling this the dark funnel, and it points in one direction: your AI-influenced pipeline is almost certainly larger than your dashboard shows. As of late 2025, only around 16 percent of brands were systematically measuring AI search performance at all, so most companies do not even have the partial view.

What we tell clients to do about it

1. Stop arguing about the benchmark and measure your own. Your conversion rate for AI referral traffic is available in your analytics this week. It beats any figure you can borrow from a study of somebody else's sites.
2. Separate AI referrers from generic referral traffic with a custom channel grouping. Until you do this, every subsequent number is guesswork.
3. Add a self-reported attribution question to your primary form. One optional field asking how someone found you catches a meaningful share of the discovery that analytics cannot see. It is the cheapest fix on this list by a wide margin.
4. Track the leading indicator separately. Citation and prompt coverage move before sessions do, so they tell you whether the work is landing months before the traffic proves it.
5. Report the branded search line alongside AI referrals. If AI visibility is working, branded search volume usually moves with it, and that correlation is the closest thing to evidence you will get for the dark funnel.
6. Value the channel as an option rather than a line item. At current volumes the revenue case is thin for most businesses. The case for building the capability now, while citation positions are cheap to earn, is much stronger.

The question worth asking

It is not whether AI traffic converts better in general. That question has at least three defensible answers and the argument will run for another two years.

The useful question is whether it converts better for you, at what volume, and whether that volume is rising. All three answers are sitting in your own analytics, and none of them require you to pick a side in somebody else's study.

Get your AI visibility score

We will run your brand against a tracked prompt set and show you where you are named, where you are missing, and who is being named instead.