AI assistants sent under 3% of organic traffic over 12 months

Two analyses published this week, alongside a 12-month benchmark published in September and updated on 14 October, put traffic from ChatGPT and other AI assistants at a small fraction of what Google organic search sends, even though the AI visitors often convert at a higher rate.

Marie Haynes published 12 months of GA4 data across 100 websites on 21 October. The SEO Works published its 12-month benchmark across 95 websites in September and updated it on 14 October. Kevin Indig went looking for AI traffic hidden inside the Direct channel across 12 sites on 19 October. The three datasets were built differently and still land in the same place: for most businesses, AI assistants are a minor channel next to Google.

Do ChatGPT visitors at 13.8% convert better than Google?

Haynes' dataset covers more than 145 million sessions. Across those 100 sites, ChatGPT sessions came to 0.62% of the volume Google organic delivered. Adding every other LLM lifted the figure to 0.73%.

Conversions are more nuanced. Of the 100 sites, 55 were tracking conversions in GA4. In that group ChatGPT visitors converted at 13.8% against 3.02% for Google organic, more than 4.5 times the rate. Google still drove 96.5% of the conversions across the channels she analysed, and Google organic drove 27 times more conversions than all LLMs combined.

Haynes also found the growth curve has flattened. Across the 100 properties, the exponential rise in ChatGPT sessions ended in August 2025, around the release of GPT-5, and ChatGPT has sent fewer visitors to those sites since.

One side finding should bother any owner paying for marketing: 45 of the 100 randomly sampled sites were not tracking any key events at all.

Measure AI referrals against organic search, not total sessions

The SEO Works benchmark runs from September 2025 to August 2026 across 95 GA4 properties. The sites received 2,825,594 organic search sessions and 72,095 AI referral sessions, so organic delivered 39 times more traffic. Put the other way, AI assistants sent 2.6% of what organic sent.

The share is rising. AI referrals sat between 2.0% and 2.8% of organic until April 2026, then climbed to 3.1% in May and 3.9% by August. The SEO Works links the May step to ChatGPT making brand names clickable inline in its answers around 7 May, and calls that a working hypothesis rather than a proven cause.

The study deliberately uses organic search as the denominator rather than total sessions, because total sessions move with paid campaigns, email and direct traffic that have nothing to do with AI visibility. It also excludes Google AI Overviews entirely, since those clicks arrive in GA4 as ordinary organic sessions and cannot be separated. Given how far AI Overviews on commercial keywords have spread, the measured AI contribution is best read as a floor.

AI traffic converted at 0.62x for B2B Products

Across the 63 sites with enough volume and recorded conversions to compare (at least 100 AI sessions and 1,000 organic sessions over the year), AI referrals converted at 7.18% against 6.08% for organic, a ratio of 1.18x pooled and 1.21x at the median. That sits far below the 3x, 4.4x and 23x multiples circulating elsewhere; The SEO Works says the two highest come from single companies measuring themselves or publishers counting newsletter sign-ups against a low search baseline.

The average hides a wide spread. On 23 of the 63 sites, AI traffic converted worse than organic. Five sites recorded no AI conversions at all from 1,487 AI sessions between them.

The comparison channel changes the answer too. The same data gives 0.67x against paid search, 1.18x against organic, 1.37x against total sessions and 2.48x against direct. Anyone quoting a multiple without naming the baseline has given a number that cannot be checked.

By customer type, B2C sites saw AI convert at 1.51x the organic rate, mixed sites 1.46x and B2B services 1.21x. B2B product businesses were the exception at 0.62x, and they also received the highest share of AI referrals relative to organic at 3.40%. The group is only ten sites, so the researchers removed each one in turn and recalculated. The worst case was still 0.71x, and AI visitors converted below the organic rate on six of the ten sites.

Direct grew on eight of 12 sites, much of it bots

Indig tested the popular theory that a lot of AI-driven visits arrive untagged in the Direct channel. He compared August and September 2026 across 12 GA4 properties, mostly US-targeted and skewed to B2B, after excluding homepages, pricing, login and other pages people tend to type in.

Eight of the 12 properties showed Direct growth, some over 100%. Much of it was bot traffic. One site logged 409,000 Direct sessions, of which 198 were engaged. From January to September, total Direct rose 140.9% while engaged Direct fell 1.0% in monthly counts. Direct sessions carrying any identifiable AI trace averaged 0.003% of Direct traffic.

A second test across four brands and 632 comparisons found one correlation of 0.72 between AI share of voice and Direct traffic at a three-week delay. When Indig shifted the weeks out of alignment, several wrong pairings produced equally strong or stronger results, so he could not establish a reliable link between AI visibility and later branded search or Direct visits.

He notes one company, Ramp, has attributed 30 to 40% of its Direct traffic to answer engines by matching visit behaviour on long blog URLs to search visitors and confirming with self-reported attribution. Across his own 12 sites, he could not detect the signal yet.

Organic growth of 3.2% matches doubling AI traffic

The SEO Works ran the arithmetic for a typical site in its set, holding conversion rates constant. Doubling AI referral traffic would add 6.6 conversions per site per month. Growing organic search by 5% would add 10.3. Growing organic by 3.2% matches the gain from doubling AI traffic.

The two channels also move together. Across 61 sites, the correlation between AI referral growth and organic growth was +0.557, and +0.534 after controlling for overall site growth. The sites gaining AI traffic were the sites whose organic search was improving. The same control applied to Direct traffic turned the relationship slightly negative, so the dataset offers no support for the idea that AI mentions quietly drive Direct visits.

Haynes draws the broader point. The AI search that sends people to businesses today is Google, through AI Overviews, AI Mode and a ranking system already driven by machine learning. She argues the industry has put too much money into tracking ChatGPT mentions, and that the time is better spent on conversion tracking, user satisfaction and preparing for agentic surfaces such as Universal Cart, the Universal Commerce Protocol and WebMCP. Both Haynes and Indig publish paid resources on AI search, so treat the conclusions as informed interpretation sitting on top of the data rather than settled fact.

Trust your own GA4 property, not 207 sites elsewhere

None of the three samples is Australia-specific. Indig's sites mostly target the US, and the other two do not break out country. The pattern is consistent enough to act on, but your own GA4 property is the number that matters. If the terminology is a hurdle, our guide to the difference between GEO and SEO covers the terms before you measure.

• Build an AI referral segment. Google only added an AI Assistant channel to GA4 in May 2026, so for a 12-month view filter session source with a regex such as bard|chatgpt|openai|claude|copilot|gemini|perplexity|grok, which is the method The SEO Works used.
• Compare AI sessions and conversions with organic search, then with paid search. State the baseline every time, because the multiple swings from 0.67x to 2.48x depending on which channel you pick.
• Confirm conversions are being tracked at all. Nearly half of Haynes' sample had no key events, which turns any channel comparison into guesswork.
• Check Direct for bots before reading anything into it. Break it down by landing page, browser version and screen size, and compare total sessions with engaged sessions.
• Allow for tagging gaps. The ChatGPT referral drop in GA4 earlier this month, caused by a missing utm_source tag, shows AI referrals can slide into Direct without warning.
• Ask new leads where they first heard of you, including which AI tool, and tie those answers to revenue. Indig recommends combining self-reported attribution with tracked AI mentions and revenue outcomes.

Budget follows that reading. On these numbers, AI visibility earns a small, measured line item, and the bulk of the effort stays with the organic search and paid search channels that deliver the conversions. If your GA4 shows AI referrals above 3% of organic with a conversion rate ahead of paid search, you have a case to spend more. If it shows what these samples show, keep the GEO work small and revisit the share each quarter.