Venbit Intelligence Research · tracked websites · bots excluded
ChatGPT1.8 times the search rate
Search
29 times the visits, 16 times the leads
The short answer
Yes, on the pooled numbers: in our latest research on tracked websites, ChatGPT visitors sent 6.4 leads per 100 visits over 90 days, against 3.6 from search. They read less, 70 active seconds against 99, and got in touch more. The evidence is thin, though: 157 leads, and the median website with enough AI visits saw 1.5 per 100.
What to know
- ChatGPT visitors sent 6.4 leads per 100 visits across tracked websites over 90 days.
- Search brought 3.6 leads per 100 visits, so the pooled ChatGPT rate is 1.8 times the search rate.
- Visitors from ChatGPT read less: 70 active seconds and 2.17 pages a visit against 99 seconds and 2.71 pages from search.
- Website by website the order turns over: the median website with 100 or more AI visits got 1.5 leads per 100, against 4.8 for search.
- On the local service websites alone, AI assistants reached 9.9 leads per 100 visits and search 4.1.
- Gemini, Copilot, Perplexity, Meta AI and Claude sent 188 visits between them, too few for any lead rate.
Yes, on the pooled numbers: ChatGPT visitors sent 6.4 leads per 100 visits, and visitors from search sent 3.6. Venbit Intelligence Research followed visitors on tracked small business websites for 90 days, with every bot removed, to see which source turns them into leads. The gap has a catch in its size: 157 ChatGPT leads in all, most AI leads on a few websites, and a median of 1.5 per 100 on the websites with enough AI visits to measure one by one. Put your own two sources into the tool below and it tells you whether your gap is real yet.
Test two sources
Is the difference real?
Enter visits and leads for two sources over the same stretch of time. The tool works out each lead rate and tests whether the gap between them is larger than chance at your numbers.
5.9 leads per 100 visits
3.6 leads per 100 visits
Leads per 100 visits
Over 90 days
- ChatGPT5.9
1 lead from 17 visits
- Google search3.6
16 leads from 440 visits
You cannot tell these two apart yet
With 17 visits and 1 lead from ChatGPT, a gap this size turns up by chance alone. If both rates hold, you would need about 350 visits from ChatGPT and 9,100 from Google search, about 5 years of traffic at your current pace.
Example: a website of median size over 90 days, at the rates we measured. Type your own numbers over it.
Fisher's exact test, two-sided: p = 0.481. Under 0.05 counts as larger than chance.
- Gap
- 1.6×
- higher for ChatGPT
- Visits behind it
- 457
- over 90 days
- Needed to tell
- About 350
- visits from ChatGPT
- At your pace
- About 5 years
- of traffic
The test is Fisher's exact test, two-sided, run in your browser. Both fill-in sets come from Venbit Intelligence Research, tracked websites, over 90 days in autumn 2026, bots excluded; search there counts unpaid visits from every search engine. Nothing you type leaves your browser.
The tool opens on a website of median size over 90 days. At 609 visits a month, with 1.0% from AI assistants and 24% from search, that website gets about 17 ChatGPT visits and 440 search visits in three months, and I set its leads close to the rates we measured. Even so, the test cannot tell the two sources apart, and at that pace the website would need about five years of traffic to find out.
Half the tracked websites get less traffic than that. For most owners, then, the answer about their own website is that it is too early to tell, whatever the pooled figures say. The websites behind these figures are small and local businesses in the Venbit Approved System, and none is named here.
ChatGPT visitors sent 6.4 leads per 100 visits
- 6.4
- ChatGPT leads per 100 visits (157 from 2,435)
- 3.6
- Search, with 2,476 leads on 69,697 visits
- 1.8×
- how far the ChatGPT rate sits above search
- 6.6
- All AI assistants together, 172 leads on 2,623 visits
These figures pool every visit. I added the ChatGPT visits across all tracked websites, added the leads those visits produced, and divided, then did the same for search. Search here means unpaid visits from Google and the other search engines, with paid search ads counted in their own channel. A lead is a form sent, a phone number tapped, or the goal that website set.
Rate and volume point in opposite directions. Search brought 29 visits for every visit from ChatGPT and 16 leads for every ChatGPT lead, so ChatGPT's better rate sits on a small base. Put through the same exact test the tool uses, the pooled gap is far past chance, with a probability under one in a billion. That test treats every visit as independent of every other, and the sections below show why that does not hold here.
| Source | Visits | Leads | Leads per 100 visits |
|---|---|---|---|
| ChatGPT | 2,435 | 157 | 6.4 |
| All AI assistants together | 2,623 | 172 | 6.6 |
| Organic search, every engine | 69,697 | 2,476 | 3.6 |
AI visitors read less and contact more
I expected the visitors who convert better to read more as well. They read less. A ChatGPT visit averaged 70 seconds of active time and 2.17 pages, against 99 seconds and 2.71 pages for a visit from search, and fewer ChatGPT visits counted as engaged.
| ChatGPT | Search | |
|---|---|---|
| Leads per 100 visits | 6.4 | 3.6 |
| Active seconds a visit, average | 70 | 99 |
| Pages a visit, average | 2.17 | 2.71 |
| Visits counted as engaged | 73% | 79% |
My reading of it: a person who asks ChatGPT for a plumber or an electrician gets a short list with reasons attached, so the comparing is done before they arrive. They check the page they were sent to, find the phone number or the form, and use it. A visitor from search starts with a page of results and opens several sites to compare, which takes more reading on each. The data cannot prove that story, but it fits every row of the table.
ChatGPT visitors read less and get in touch more: 70 active seconds against 99, and 6.4 leads per 100 visits against 3.6.
That pattern decides what the page an assistant links to should do. Of the 2,435 ChatGPT visits, 1,855 came from a phone, 76%, so the visitor who has already decided wants a number to tap near the top of a small screen. A long page that saves its contact details for the bottom wastes the head start these visitors bring. For what normal reading time looks like across all sources, see average time on site for a small business website.
Per website, AI visits sent 1.5 leads per 100
A pooled total counts every visit once, so a website with many AI leads pulls the whole figure toward itself. To see what one website can expect, I took each website with at least 100 visits from a source, worked out that website's own lead rate for the source, and found the median across websites. For AI assistants, few websites cleared the 100-visit bar.
| Source | Median leads per 100 visits |
|---|---|
| Search | 4.8 |
| Direct | 3.9 |
| Ads | 3.0 |
| AI assistants | 1.5 |
| Social media | 1.4 |
| Other websites | 1.1 |
Two things explain the turn. A few websites carry most of the AI leads, and inside a pooled total they lift the figure for everyone. The median gives each website one vote, so at least half of the websites in the AI row sit at 1.5 leads per 100 or below. The AI row also holds few websites, so read it as an early signal and not as an industry average.
Venbit Intelligence Research · tracked websites · bots excluded
Every visit pooled
Every tracked website added together
AI assistants172 leads from 2,623 visitsSearch2,476 leads from 69,697 visits
The median website
Websites with 100 or more visits from the source
AI assistantsFew websites, an early readSearchMedian of each website's rate
Both numbers are correct, and they answer different questions. The pooled 6.4 says that across a large set of small business websites, a visit from ChatGPT produces more leads than a visit from search. Per website, the 1.5 says your own website should not count on it, since the median website with real AI traffic got fewer leads per AI visit than per search visit. Yours is one website, which is why the tool tests your numbers and not ours.
Treat 157 leads as an early reading
Every figure on this page rests on 157 ChatGPT leads, and 172 from all assistants together. That is enough to show the pooled gap clears chance and too few to split by kind of business, by device or by month with confidence. Below ChatGPT, the assistants sent so few visits that I will not give a lead rate for any of them.
Visits from the other AI assistants, 90 days
- Gemini96
- Copilot50
- Perplexity33
- Meta AI5
- Claude4
Gemini's 96 visits would swing from a strong rate to a weak one on two or three leads either way, and Claude's four visits are four people. ChatGPT sent 93% of all assistant visits, so for now the question of AI visitors on a small business website is the question of ChatGPT visitors. The share of traffic each assistant sends, month by month, is covered in how much website traffic comes from ChatGPT.
Local service websites saw 9.9 leads per 100 AI visits
Some tracked websites are not local services at all, and on those a visitor can sign up or read without ever contacting the business. Leave them out and the gap between AI and search grows wider on the local service websites that remain.
| Source | Leads per 100 visits | Engaged | Active seconds a visit |
|---|---|---|---|
| AI assistants | 9.9 | 75% | 71 |
| Search | 4.1 | 80% | 113 |
| Ads | 3.8 | 47% | 63 |
| Direct | 3.6 | 58% | 76 |
| Other websites | 3.0 | 70% | 87 |
| Social media | 2.5 | 59% | 86 |
AI assistants top the table at 9.9 leads per 100 visits, more than twice the 4.1 from search, while their visitors spent 71 active seconds against 113. It is the same pattern as the full set, only sharper. These websites got about a third of all assistant visits, and 87 leads is a small number to build on, so the per-website caution above applies here with more force.
Count ChatGPT ads apart from ChatGPT answers
ChatGPT sells ads now, and their clicks arrive tagged utm_source=chatgpt with a paid medium such as utm_medium=cpc. A click like that was paid for, and no answer sent it, so Venbit Intelligence files it as an ad under the name ChatGPT Ads. Its AI search figures count only the people an answer sent.
Microsoft Adsnow/emergency-repair?msclkid=8f2c41…
ChatGPT Adsnow/?utm_source=chatgpt&utm_medium=cpc
Metanow/?utm_source=facebook&utm_medium=paidGoogle Adsnow/care-plans?gclid=Cj0KCQjw…
Google AdsBy its click id
MetaBy its link tags
Microsoft AdsBy its click id
TikTokBy its click id
LinkedInBy its click id
XBy its click id
PinterestBy its link tags
RedditBy its click id
SnapchatBy its click id
NextdoorBy its link tags
YelpBy its link tags
ChatGPT AdsBy its link tags
If your own analytics files every visit tagged chatgpt as AI traffic, the first ad campaign you run there will move your AI lead rate for reasons that have nothing to do with what assistants say about your business. Check how your tool treats utm_medium before you compare the two sources.
Decide what to do about AI visitors
- 1
Measure your own two rates
Take visits and leads from ChatGPT and from search for the same 90 days and put them in the tool above. If it says you cannot tell them apart yet, your website has not answered the question, whatever a study says, this one included.
- 2
Put the way to reach you near the top
ChatGPT visitors read less, and 76% of them arrive on a phone. On the pages assistants send people to, keep the phone number and a short form within the first screen. Mobile and desktop traffic for a small business covers the device side.
- 3
Keep search as the main channel
Search brought 16 leads for every ChatGPT lead in this data. A higher rate on a small base does not replace the source that fills the calendar, so keep the pages that rank in good order.
- 4
Check the pages assistants read
An assistant recommends from what it has read about you. Make sure the pages it reaches say what you do, where you work and how to contact you, in words it can repeat. Which AI bots read your website covers the crawlers behind those answers.
- 5
Tie each lead to its source
Count leads by where the visit came from, so a ChatGPT lead is counted as ChatGPT and not as direct. The guide on tracking where website leads come from shows how.
See AI visitors and their leads in Venbit Intelligence
Venbit Intelligence names every assistant that sends a visitor and marks the visit the moment it becomes a lead. In the Analytics tab, the AI search section gives each assistant its own row, with visits and leads beside the period before, and opening a row lists the people it sent, the pages it sent them to and what its crawlers read. The Feed and Analytics page shows the whole tab.
GPTBot, ClaudeBot and most other AI crawlers read pages without running scripts, so the pages read here are the ones loaded the way a browser does.
To compare AI with search the way this page does, open Traffic sources and tap AI assistants, then Search. Each shows its visits, its leads with the share of visits they make, and the share of visits that stayed to read, so the two rates sit a tap apart. In the Inbox, a lead an assistant sent carries that assistant as its source, so you know which enquiry came out of a ChatGPT answer before you reply; the Inbox page shows how leads arrive.
You can also ask your own AI assistant. Venbit MCP connects ChatGPT, Claude and other assistants to your Venbit Intelligence numbers, read-only and one website per link, so a question about whether ChatGPT visitors turned into leads more than search visitors this quarter gets an answer from your own figures.
visitor_journeysad_reportYour contact form loses the most. 411 visits clicked into it and 79 sent it: 127 stopped at the Message box, against 78 at Phone. Your retargeting ad lands on the home page, where 79% of its clicks leave within five seconds.
When the Feed compares one source with the rest of your website, it runs the same family of exact test as the tool above, holds like against like on the same devices and new visitors against new, and shows the finding only when the gap clears chance. Visitors sent by AI assistants are one of the sources it tests, so if they bring more than their share of your leads, the Feed says so once your numbers can carry it.
How we measured this
Every number here comes from Venbit Intelligence Research: small business websites tracked in the Venbit Approved System from 21 August to 18 November 2026. Bots were identified and removed before anything was counted. A visit counts as ChatGPT when the browser reports chatgpt.com or chat.openai.com as the page it came from, or when its link carries a utm_source tag naming ChatGPT or OpenAI, and the other assistants are found by the same two prints. Visits with neither print land in direct, the catch-all explained in what direct traffic is and why it is so high.
157 ChatGPT leads
Enough to show the pooled gap clears chance, too few to split by kind of business, device or month.
Few websites in the AI median
Few tracked websites got 100 or more AI visits, so read the AI median as a comparison and not as an industry average.
Search means every engine
The search figures add unpaid visits from Google and the other search engines together.
Some AI visits look direct
From 8 to 15 October, links inside ChatGPT's answers opened with no referrer and no tag, so some of its visits were filed as direct. Search Engine Journal reported the bug.
A lead is a form sent, a phone number tapped, or the goal the owner set; a tap shows intent, and no website can see whether the phone was answered. Leads per 100 visits counts the visits that produced a lead. Convert on this page means got in touch, not bought, because owners mark outcomes on too few leads to publish a close rate.
The published comparisons of AI and search visitors I found measured B2B companies and publishers, and they disagree with each other. I found none on local service businesses, which is the reason to publish these numbers with their limits attached. Websites tracked this closely may be more active online than the average shop, so treat the figures as a well-measured sample of small business websites and not as a census.
See which assistant sends you leads
Venbit Intelligence names every AI assistant that sends you a visitor, counts its leads beside your search leads, and tests each gap on your own numbers. One line to install, free for 14 days, no card needed.
In Venbit Intelligence
More website analytics guides
Every question in this series, from Website Analytics for Small Business Owners.
Leads inbox4
Ad performance4
SEO tracking3
Visitor behaviour4
Feed and notifications3
Goals and measurement4
Privacy, AI and teams3
Venbit Intelligence Research19
- Average Website Visitors per Month for a Small Business
- What Is a Good Conversion Rate for a Small Business Website?
- What Time of Day Do Website Leads Come In?
- Which Day of the Week Brings the Most Website Leads?
- How to Tell If a Website Lead Is Spam
- How Much Website Traffic Comes From ChatGPT?
- Do ChatGPT Visitors Convert Better Than Google Visitors? (you are here)
- Which AI Bots Read Your Website?
- Why Ad Clicks Don't Match Website Visits
- Mobile vs Desktop Website Traffic for Small Business
- How Long Do Website Visitors Take to Contact a Business?
- Average Time on Site for a Small Business Website
- Average Pages per Visit for a Small Business Website
- How Far Do Visitors Scroll on a Website?
- Why Is My Website Getting Traffic From China and Singapore?
- What Is Direct Traffic and Why Is It So High?
- Where Does Small Business Website Traffic Come From?
- How Many Visitors Reach Your Contact Page?
- New vs Returning Visitors for a Small Business Website
Tony Carnevale
Founder, Venbit
Tony started Venbit in Mill Creek, WA in 2011 and still does the work: web design, SEO, and the website analytics product the company builds and runs for its own clients. The numbers and screens in his articles come from that product, not from a content mill.