If you have spent any time in SEO circles lately, you have heard the pitch: optimise for AI answers now, or get left behind. A new study gives that pitch a reality check — not by dismissing AI search, but by measuring it. And the measurement is sobering.
SEO consultant Marie Haynes analysed 12 months of Google Analytics 4 data across 100 websites of varying sizes, covering more than 145 million sessions. Her finding, published on October 21, 2026, was blunt: in that dataset, sessions from ChatGPT amounted to just 0.62% of the volume driven by Google organic search. Adding every other large language model lifted the figure only to 0.73%.
That is the headline. But the headline alone would mislead you. The same study found that ChatGPT visitors converted far better than Google visitors — they just arrived in far smaller numbers. This post walks through what the data actually says, where it is limited, and what it should change about how you split your time.
Key Takeaways
- ChatGPT sent 0.62% as many sessions as Google organic search across 100 sites and 145M+ sessions; all LLMs combined reached 0.73%.
- Among 55 sites tracking conversions, ChatGPT visitors converted at 13.8% versus 3.02% for Google organic — more than 4.5 times the rate.
- Despite that higher rate, Google organic still drove 27 times as many total conversions as all LLMs combined.
- Haynes reports ChatGPT referral growth stalled after August 2025 and declined afterward.
- The sensible read: keep investing where the traffic is today, and treat AI visibility as a measured experiment, not a budget replacement.
What Did the Study Actually Measure?
How big was the dataset?
Haynes looked at 12 months of GA4 data across 100 different websites of varying sizes, totalling more than 145 million sessions. That scale is what makes the study worth paying attention to: it is not one site’s anecdote, and it is not a survey of opinions. It is observed referral traffic, counted the way most site owners count it.
What does 0.62% really mean?
This is the number most likely to be misquoted, so let us be precise. The 0.62% figure is ChatGPT sessions expressed as a share of Google organic sessions — not as a share of all website traffic. The same applies to the 0.73% figure for all LLMs combined. If someone tells you “AI sends 0.73% of all traffic,” they have read the study wrong. The study compares AI referrals against one channel, Google organic, and that channel is itself usually the largest single source of visits for these sites.
Why does the conversion rate matter so much?
Among the 100 sites, 55 were tracking conversions in GA4. In that group, 13.8% of ChatGPT visitors converted, compared with 3.02% of Google organic visitors. That is a striking gap, and it fits a plausible story: someone who has already had a back-and-forth conversation with ChatGPT about a problem may arrive at your site further along in their decision than someone who clicked a search result after a two-second glance.
But rates are not totals. Google organic still drove 27 times as many conversions as all LLMs combined, and 96.5% of the conversions across the channels Haynes analysed. A high conversion rate on a tiny number of visits produces a small number of leads. Both facts are true at once, and any honest summary has to hold both.
Has ChatGPT Referral Growth Stalled?
What happened after August 2025?
According to Haynes, the exponential rise in ChatGPT sessions across the 100 properties ended in August 2025, around the release of GPT-5, and ChatGPT has sent fewer visitors to those sites since. That is a claim about this dataset, not a universal law — but it cuts against the assumption that AI referral traffic is compounding month after month for everyone.
Why might growth have flattened?
The study itself does not settle the cause, and this post will not invent one. Plausible factors include changes in how ChatGPT answers questions (more answered in-chat, fewer clicks out), shifts in user behaviour, and the simple fact that early growth from a tiny base always looks exponential. What matters for your planning is the direction: in this dataset, the curve bent, and it bent a while ago.
Is GEO Really a “Distraction”?
What did Haynes argue?
Haynes called generative engine optimisation a “distraction” — her argument being that, for the vast majority of sites, AI referral traffic is minuscule compared with Google organic visitors, so effort spent chasing AI citations is effort taken away from the channel that actually pays the bills. In her words, as reported by Search Engine Land: “The AI search that is actually sending people to businesses today isn’t ChatGPT. It’s Google.”
What did Search Engine Land’s editors say back?
Search Engine Land’s editors pushed back, and their counterpoint deserves equal weight. Today’s referral numbers, they argued, do not tell us how important AI search visibility will become. Their advice: focus on the Google traffic you get today, but do not lose sight of where search is heading and the many ways customers may find your brand. That is not a contradiction of the data — it is a disagreement about what the data implies for the future.
So who is right?
Both, in different time horizons. If your question is “where should next quarter’s effort go?”, the data points firmly at Google organic. If your question is “should I ignore AI search entirely?”, the answer is no — the conversion-rate finding alone suggests AI-referred visitors are unusually qualified, and being cited in AI answers also builds brand familiarity that may show up later as branded searches or direct visits, even if it never appears as a referral in GA4. The mistake is treating either horizon as the whole story.
What Should You Actually Do With This Data?
How should you measure your own AI referral traffic?
Start with your own numbers before borrowing anyone else’s conclusions. In GA4, check session source/medium for chatgpt.com, perplexity.ai, and other AI referrers, and compare them with Google organic over the same period. If you track conversions, compare rates the way Haynes did — but also compare totals. A channel can be your best-converting source and still be your tenth-biggest source of leads. Both numbers belong in the same report.
One measurement warning: not all AI-driven visits arrive labelled. Earlier this month, ChatGPT briefly appeared to drop its UTM tags, sending some AI traffic into the “direct” bucket before OpenAI confirmed and fixed the issue — we covered that measurement story in ChatGPT’s Missing UTM Tags: Measuring AI Referral Traffic. If your AI numbers look oddly low, check your tagging before concluding the channel is dead.
How should you split your SEO effort?
A practical split follows from the data: the bulk of your effort still belongs to the fundamentals that drive Google organic traffic and AI citations alike — clear, well-structured content that answers real questions, strong topical coverage, and pages that load fast and work well. That is also why our earlier guides on getting cited by ChatGPT (link), getting cited by Claude (link), and getting cited by Perplexity (link) emphasise content quality over tricks: the same work serves both channels. What the Haynes data argues against is a separate, large GEO budget justified by traffic that has not arrived yet.
What should you track going forward?
Watch three things, quarterly: your own AI referral volume and conversion rate, whether your pages are being cited in AI answers (see our guide to tracking AI Overviews citations, link), and whether branded search volume moves after AI visibility changes. If your own numbers start diverging from the study — say AI referrals growing while Google stays flat — that is the signal to rebalance. Until then, let your own data, not the hype cycle, set the budget.
People Also Ask
How much traffic does ChatGPT send compared to Google?
In Marie Haynes’ 12-month study of 100 sites and 145M+ sessions, ChatGPT sent 0.62% as many sessions as Google organic search, and all LLMs combined sent 0.73% — measured against Google organic, not against total site traffic.
Do ChatGPT visitors convert better than Google visitors?
In the study’s 55 conversion-tracking sites, yes by rate: 13.8% of ChatGPT visitors converted versus 3.02% of Google organic visitors. But Google organic still drove 27 times more total conversions because its volume is vastly larger.
Is GEO worth investing in?
As a primary strategy, the data says no for most sites — AI referral volume is still tiny next to Google organic. As a measured side effort built on the same quality content that serves Google, yes: AI-referred visitors convert well, and AI visibility may matter more over time.
Is ChatGPT referral traffic still growing?
In Haynes’ dataset, no: exponential growth in ChatGPT sessions ended in August 2025 and referrals declined afterward. That describes her 100-site sample, so check your own GA4 data before assuming the same pattern holds for your site.
The Bottom Line
The Haynes study does not say AI search is irrelevant. It says AI search, as a referral channel, is currently small — and that smallness should discipline your budget. Google organic still delivers the visits and the conversions. AI-referred visitors convert at an impressive rate, which makes AI visibility worth cultivating, but not worth raiding your core SEO budget for. Measure your own numbers, keep the fundamentals strong, and let the data — yours, not the discourse — decide when the balance shifts.
Sources: Search Engine Land, “ChatGPT sends less than 1% of Google’s traffic: Data” (Oct 21, 2026); Marie Haynes via Search Engine Journal, “GEO Is A Distraction: Here’s Where The Real Opportunity Is”; MarTech; Search Engine Roundtable’s roundup of the Haynes, SEO Works, and Kevin Indig studies.
