AI sends less traffic, but it's worth far more

Where the conversion gap comes from

Seer Interactive tracked a single B2B client over six months (Oct 2024–Apr 2025) and found a striking gap: ChatGPT referrals converted at 15.9%, Perplexity at 10.5%, and Google organic at 1.76%. That's not a rounding error — it's a structural difference in visitor intent. People who arrive from an LLM have already had a research conversation; they're further down the funnel before they land. Adobe's retail data points the same direction from a very different angle: by March 2026, AI-referred visitors to US retail sites converted 42% better and drove 37% more revenue per visit than non-AI traffic — a complete reversal from March 2025, when AI traffic had converted 38% worse.

ChatGPT vs Perplexity vs Google organic

The Seer data lets us compare AI sources directly. Perplexity sits between ChatGPT and Google organic — consistent with the idea that more guided, citation-heavy answers produce higher-intent visitors. Engagement depth reinforces the story: ChatGPT visitors averaged 2.3 pages per session versus roughly 1.2 for Google organic (Seer), and Conductor/Knotch data shows AI-referred sessions run roughly 2.3× longer than traditional search sessions.

Adobe adds another dimension beyond conversion rate: time on site. AI-referred shoppers spent 48% more time on site and visited 13% more pages per session than non-AI visitors in March 2026.

Still just 1% of all traffic

Conductor's benchmark across 13,770 domains and 3.3 billion sessions puts AI referral traffic at 1.08% of all traffic, growing at roughly 1% month-on-month. ChatGPT accounts for 87.4% of that AI slice. The distribution is uneven: IT and technology sites see the highest share at 2.80%; Communication Services the lowest at 0.25%. The quality story is real, but the volume story is not — at least not yet. Frame the ROI as quality per visit, not as a replacement for Google organic at scale.

Higher conversion is real in some datasets. Not all.

The high conversion figures above deserve scrutiny, and two large-scale studies offer a necessary counterweight.

Amsive ran a paired analysis across 54 websites and found that LLM traffic did not convert significantly differently from other channels overall — a p-value of 0.794, which is about as far from statistical significance as you can get. A study published in Marketing Science looked at 973 e-commerce sites, comparing roughly 50,000 ChatGPT referral transactions against 164 million traditional-channel transactions. The result: ChatGPT referrals converted lower and produced lower revenue per session than most major channels.

The honest conclusion is that higher-intent conversion from AI referrals is real in some contexts — particularly single-client B2B and US retail — but not universal. Vertical, product type, brand familiarity, and the nature of the LLM answer all matter. The Seer result is a single client. The Adobe data is US retail. Neither generalises automatically to B2B SaaS, services, or markets outside the US. The Amsive and Marketing Science studies are larger, more varied samples, and their null or negative results deserve equal weight.

Visibility in LLMs is a traffic qualifier, not just a vanity metric

Even if you weight the sceptical studies heavily, the direction-of-travel is clear: AI referral volume is growing fast (393% YoY in retail), the visitors who arrive are engaged, and in some contexts they convert far better. The practical implication is not to ignore Google and chase ChatGPT — it's to make sure the content LLMs cite about your brand is accurate, current, and positioned for the visitor intent that AI answer-generation attracts. Someone arriving from a ChatGPT recommendation has already been pre-sold; your job is to not lose them.

Baseline Labs tracks which LLMs mention your brand, in what context, and how sentiment shifts over time. If AI traffic is starting to show up in your analytics, knowing what the model said about you before the click matters.

Track your AI traffic value