If you have read anything about AI search traffic in the last year, you have seen this: ChatGPT traffic converts at 15.9%, against 1.76% for Google organic. It gets quoted constantly, usually with no source, sometimes attributed to whoever quoted it last.
The number is real. It comes from a Seer Interactive case study published on 3 June 2025 by Nick Haigler and Garman Chan.
What is almost never quoted is the sample it came from.
What the study actually measured
| Source | Seer Interactive, published 3 June 2025 |
| Sample | A single Seer client |
| Period | 1 October 2024 to 30 April 2025 |
| AI traffic | around 11,000 sessions |
| Google organic | around 14 million sessions |
| AI-attributed conversions | 1,370 |
| Conversion definition | not specified in the case study |
Seer were straightforward about this being one client’s data. The distortion happened downstream, as the figure was restated by people who dropped the sample size, the date range and the single-client caveat, until it read like an industry benchmark.
Three things follow from that table.
The denominators are not comparable. Roughly 11,000 sessions against 14 million. Seer put the AI side at 0.07% of organic traffic. Rates computed on small denominators move violently, and a rate is not more trustworthy because the percentage is dramatic.
The conversion event is undefined. This is the biggest gap. A 15.9% rate means something entirely different if the conversion is a newsletter signup rather than a purchase. Without knowing which, the comparison to a 1.76% organic rate is not like for like, because the organic figure is drawn from an enormous and much more mixed pool of intent.
The data ends in April 2025, which makes it more than fifteen months old as of August 2026. Seven months is the length of the measurement window, not the age of the finding, and those get conflated constantly. AI referral behaviour has not been stable since.
What it probably does show
None of that makes the finding worthless. The direction is plausible and it matches what a lot of operators report anecdotally: traffic arriving from AI answers tends to convert better than traffic arriving from a search result.
The likely mechanism is unglamorous. It is selection, not quality.
Someone who arrives from an AI answer has usually already asked the comparison question, the pricing question and the “is this right for me” question, and got answers before they ever clicked. The click happens near the end of the decision rather than the start. Search traffic, by contrast, includes everyone at every stage, including people who will never buy anything.
So you would expect fewer visits, later in the funnel, converting at a higher rate. That is a shift in where the click sits in the journey, not evidence that AI traffic is inherently better.
It also implies something uncomfortable: the sessions you lose to AI answers and the sessions you gain from them are not the same sessions. Losing early-stage informational visits while keeping late-stage decision visits will raise your conversion rate and lower your traffic at the same time. Your dashboard will look like it improved.
What to do instead of quoting it
Measure your own. This is the only number that matters and almost nobody bothers. Segment AI referrers in your analytics and compare them against organic using the same conversion definition and a window long enough that a handful of sessions does not swing the rate.
Common AI referrers to segment: chatgpt.com, perplexity.ai, claude.ai, copilot.microsoft.com, gemini.google.com. Referrer data from AI surfaces is incomplete, because some traffic arrives with no referrer at all, so treat your counts as a floor.
Compare rates, but hold the definition fixed. A conversion has to mean the same event on both sides or the comparison is theatre.
Watch the absolute numbers, not just the rate. A rising conversion rate on falling sessions can mean you are being cited well. It can also mean you have lost the top of your funnel to answers that never send a click. Those look identical in a rate and completely different in revenue.
Do not buy a dashboard to answer this. Monitoring tools tell you whether you are cited, not how that traffic converts. Your own analytics already hold the conversion side.
Check you are reachable at all. None of this matters if AI crawlers cannot read your pages. Most of them do not document whether they execute JavaScript, which is covered in which AI crawlers render JavaScript and per crawler in our directory.
Why this is harder to measure than it sounds
Two structural problems sit under every number in this debate, including Seer’s.
Google’s AI traffic is not separable. Google’s AI features documentation says sites appearing in AI features “are included in the overall search traffic in Search Console”, under the Web search type. There is no AI Overviews or AI Mode breakout. So the 1.76% organic baseline in any study like this already contains AI-influenced traffic, and nobody can separate it out. The comparison is not clean on either side.
Referrers undercount. Chat interfaces do not reliably pass a referrer. Some AI-sourced visits arrive as direct traffic and land in the wrong bucket entirely, which understates AI sessions and inflates whatever the rate is computed on. Your AI session count is a floor, not a measurement.
The combined effect is that the true gap between AI and organic conversion is unknown in both direction and size. That is not a reason to ignore the question. It is a reason to distrust any confident ratio, including a memorable one.
The short version
15.9% versus 1.76% is a real finding from Seer Interactive, from one client, over seven months, on about 11,000 AI sessions, with the conversion event unspecified. Cite it that way or do not cite it.
The useful conclusion is not the number. It is that AI referrals tend to arrive later in the decision than search referrals, which means they should convert better and should also be far fewer. Measure both on your own site before you plan around either.