Yes. Across 60+ websites, AI visitors converted at nearly twice the rate of other visitors to the same site. Industry decides the size of the gain, and for software companies the result reverses.
GO MO Group · 250+ Google Analytics 4 properties reviewed · 90 days of data
The short answer
How much better do AI visitors convert?
Nearly twice as well. AI visitors converted at about 2x the rate of the other visitors on the same website, and they did so on 8 in 10 sites.
They stayed about as long, close to a minute per session. Two in three AI visits were engaged, against three in five for other traffic.
AI assistants still send a small share of traffic: about 1% of all sessions. ChatGPT sent over 90% of those visits.
~2xConversion rateAI visitors compared with the same site's other visitors.
8 in 10Sites where AI converted betterThe typical result held on most sites.
~1%Share of all sessionsOver 90% of AI visits came from ChatGPT.
30 June to 27 September 2026. All figures are rounded.
250+ GA4 properties reviewed60+ websites with enough AI traffic to compareNordic B2B, education and e-commerce companies
The questions this article answers
- “Does traffic from ChatGPT convert better than traffic from Google?”
- “Is AI search traffic worth optimising for in B2B?”
- “Why does AI traffic in my GA4 look worse than other channels?”
These are the questions marketing teams ask us, and the way they ask them in ChatGPT and Google. Each section below opens with its answer.
What is an AI visit? A session that arrived from an AI assistant such as ChatGPT, Gemini, Copilot, Perplexity or Claude.
What is a conversion? A session in which the visitor completed one of the site's key events, such as a form, a purchase or a booking.
What is the baseline? The same site's other visitors in the same 90 days. Every site is compared with itself.
Four actions
What should marketing teams do with their AI traffic?
Check that it is real, compare it with your own site, and build the landing page for the stage the visitor is at.
01
Is your AI traffic real? Break AI sessions down by country, device and landing page before you report them. On one site in our study, 9 in 10 “ChatGPT” visits were automated.
02
What is your own baseline? Compare AI visitors with your site’s other visitors. Each site defines a conversion its own way, so benchmarks from other sites will mislead you.
03
What stage is the visitor at? E-commerce visitors from AI arrive ready to buy, so shorten the path to checkout. Software visitors arrive to research, so lead with comparison and pricing content.
04
Are you counting every assistant? GA4’s AI assistant channel catches ChatGPT, Gemini and Copilot. Add your own channel rule so Perplexity and Claude visits are counted as AI too.
Hypothesis
Why is our result lower than the 23x in the headlines?
Because the biggest numbers come from single websites. Across 60+ sites, the typical uplift is about 2x.
We expected to confirm the uplift everyone quotes. Ahrefs reported that AI visitors converted about 23 times better on its own site. Conductor's benchmark put the figure at about 4x.
A study of 973 e-commerce sites by the University of Hamburg pointed the other way: ChatGPT visitors converted slightly below organic search visitors.
Our result sits between the two. We believe the method explains most of the gap. A single site with a strong product can show a large uplift. Comparing each of 60+ sites with itself gives the typical result.
Plan on AI visitors converting at about twice your site's normal rate. Then test that figure on your own data.
Reported uplift for AI visitors, by study
StudyUplift
Hamburg, 973 sitesjust under 1x
Bars drawn to a linear scale. The studies use different comparison traffic: organic search (Ahrefs, Hamburg), all other traffic (Conductor) and the same site's other visitors (GO MO).
Industry groups
Which industries gain the most from AI visitors?
E-commerce gains the most, education and industry follow closely, and software and finance lose. The results fall into three groups, set by how far along the buying process the visitor is when they arrive.
AI visitors' conversion rate compared with the same site's other visitors
IndustryResult
Manufacturing and industryover 1.5x
Software and finance20% below
The typical (median) site in each industry. Manufacturing and industry includes construction and recycling. The e-commerce and the software and finance groups each have fewer than 10 sites.
GROUP 1 · E-COMMERCE
Decided buyers: they browse less and buy more
AI visitors spent about a third less time on e-commerce sites, and still converted at over twice the normal rate. The pattern held on every e-commerce site with purchase tracking.
Our hypothesis: the assistant has already compared the products. The visitor clicks through to buy the one it recommended.
What to do: make the product page an AI visitor lands on the shortest route to checkout.
GROUP 2 · EDUCATION AND INDUSTRY
Shortlisters: they stay longer and ask for more
In education, AI visitors converted at about twice the normal rate, on 9 in 10 sites. In manufacturing and industry the gain was over 1.5x, also on 9 in 10 sites. These are the two largest groups in the study, so the result is the most reliable.
Our hypothesis: the visitor asked the assistant for a shortlist of schools or suppliers. They arrive to confirm a choice, so they request information, apply or ask for a quote.
What to do: put the facts people check on a shortlist (entry requirements, specifications, lead times, prices) on the first page they see.
GROUP 3 · SOFTWARE AND FINANCE
Researchers: they compare and leave
This is the one group where AI visitors converted below the baseline, about 20% lower. Time on site was similar to other visitors.
Our hypothesis: people ask assistants to compare software tools and funds early in the process. They click through to check one detail. The conversion on these sites is usually a demo request or an account opening, which is a big step for someone still comparing.
What to do: give AI visitors comparison and pricing pages, and a smaller next step than a demo: a price calculator, a checklist or a short product video.
One caution: the software and finance group has fewer than 10 sites. Treat it as a strong signal to test on your own site, and give the education and industry results the most weight.
The outliers
Why does AI traffic look bad in some GA4 reports?
In our data, the main reason was automated traffic. On one e-commerce site, about 9 in 10 visits labelled as ChatGPT came from a bot.
That single site made up 60% of all AI sessions in the study. With it included, AI visitors looked shorter-staying and lower-converting than the average visitor. Once we removed it, the overall result reversed.
Our hypothesis: an automated tool, such as a scraper or a price monitor, sent the ChatGPT referrer with its requests. The visits came in one large burst, from one country outside the site's markets, and almost none engaged.
Three other sites had broken tracking: two recorded a conversion on every session, and one had a baseline full of bot visits. We removed all four sites before comparing.
Sign of automated trafficAre most AI sessions on desktop from one country?Compare the country split of AI visits with your other traffic. A country outside your markets, holding most AI sessions, needs a closer look.
Sign of automated trafficDid the sessions arrive in one short burst?Plot AI sessions by day. Human AI traffic grows steadily. A single month with ten times the usual volume points to a tool.
Sign of automated trafficIs the landing page "(not set)"?Real visitors land on a page. Large numbers of sessions with no landing page usually come from automated requests.
Sign of broken trackingIs every session a conversion?A conversion rate close to 100% on any channel means the key event fires on page load. Fix the event before comparing channels.
If AI traffic looks like your worst channel, check whether it is real before you change any pages.
By assistant
Does it matter which AI assistant sent the visitor?
Only in volume. ChatGPT sent over 90% of AI visits. Gemini, Copilot, Perplexity and Claude sent the rest.
Visitors from every assistant stayed about a minute per session. ChatGPT visitors converted slightly more often, but the other assistants sent too few visits for a reliable comparison.
For now, optimise for the questions people ask ChatGPT, and track every assistant so you see when the mix changes.
ChatGPTGeminiCopilotPerplexityClaude
Check your own data
How do you measure AI visitors in GA4?
Five steps, in this order. The whole check takes about an hour for one property.
01
Count every AI assistantIn GA4, create a custom channel group with an "AI assistants" channel above Referral. Use a session source rule that matches this pattern:chatgpt|openai|perplexity|claude\.ai|gemini\.google|copilot|deepseek|grok|mistral|kagiLeave out paid ChatGPT ad clicks, which arrive with sources such as "chatgpt_ad". Count them as paid media.
02
Remove automated and broken sessionsCheck AI sessions by country, device, day and landing page, using the signs in the section above. Exclude what fails the check and write down what you removed.
03
Compare AI with your own baselineCompare the AI channel's key-event rate, engagement rate and engagement time with all other sessions on the same property, over the same dates.
04
Split by page type and marketLook at which pages AI visitors land on and which markets they come from. This shows where the gain comes from, and which pages to improve first.
05
Match the page to the visitor's stageDecided buyers need a short path to purchase. Shortlisters need the facts they check. Researchers need comparisons and a small next step.
Methodology
How did we measure the difference?
What data did we use? First-party Google Analytics 4 data from every property GO MO manages, 250+ in total, pulled through the GA4 Data API. 60+ websites had enough AI traffic to compare.
Which period did we cover? 90 days, from 30 June to 27 September 2026.
How did we compare sites? For each site, we compared AI visitors with the same site's other visitors. We then took the typical (median) site. This method holds up because each site defines its own key events.
How were industries grouped? Using SNI 2025, Sweden's industry classification, merged into broader themes. Client names are withheld. Health and care and business services had too few sites to report separately.
What did we exclude? Sites with fewer than 100 AI sessions, one site with automated ChatGPT-referred traffic, and three sites with broken tracking.
What are the limits? This is GO MO's client sample of Nordic companies, weighted towards education and industry. AI visits with no referrer appear as Direct in GA4, so the true AI share is likely higher than 1%.
A conversion is a session with at least one GA4 key event. Ratios compare the AI conversion rate with the baseline conversion rate on the same site. All figures are rounded.
Related questions
What else should marketing teams know?
Does ChatGPT traffic convert better than Google traffic?
In our data, yes. AI visitors, over 90% of them from ChatGPT, converted at about twice the rate of the same site's other visitors, Google included. The gain is largest in e-commerce and smallest in software.
Is AI search traffic worth optimising for if it is only 1% of visits?
Yes, if those visits convert at twice the normal rate. In that case 1% of sessions brings about 2% of conversions. Start with the pages AI visitors already land on.
Why do AI visitors spend less time on e-commerce sites?
They arrive having already compared products in the assistant. They go straight to the product and buy it, which shortens the visit and raises the conversion rate.
Where does GA4 put ChatGPT traffic?
Since May 2026, GA4 files recognised AI referrals under the AI assistant channel, with the medium "ai-assistant". Some Perplexity and Claude visits still land in Referral, and visits with no referrer land in Direct. A custom channel rule catches more of them.
How can you tell real AI visits from bots?
Real AI visitors land on a page, come from your markets, use a mix of devices and grow steadily over time. Automated traffic arrives in bursts, from one country and one device, with no landing page.
What should software companies change for AI visitors?
Change what AI visitors find when they arrive. In our data, software visitors from AI came to research. Give them comparison pages, clear pricing and a smaller first step, such as a price calculator, alongside the demo form.
Research source: GO MO analysis of anonymised, first-party Google Analytics 4 data, 30 June to 27 September 2026.
Other studies cited: Ahrefs, Conductor benchmark (via Lantern), University of Hamburg study of 973 e-commerce sites, Clique Studios on the GA4 AI assistant channel. Reviewed 29 September 2026.
Want to know what your AI visitors are worth?
GO MO can clean your AI traffic data, compare it with your own baseline and show which pages to change first.