
- Across five of my sites, ChatGPT sent 84 of 477 people (17.6%) from 1 to 12 Nov 2026, and 62 of 299 (20.7%) in the October days the counters were running. Source: my own server-log counters, read 13 Nov 2026.
- The share is not spread evenly. On my Polish blog madrydochod.pl it was 30.9% in November, on tillradar.co.uk 29.3%, on mainstreetradar.com 19.6%, on kontowtescie.pl 1.1% and on gregciupek.com 0%.
- Other AI assistants barely register: 1 person from Gemini in the whole period and none from Perplexity, Claude or Copilot.
- For every person who clicked through from ChatGPT, ChatGPT's user agent fetched my pages 15 to 17 times on madrydochod.pl and tillradar.co.uk in November. Most ChatGPT reading never turns into a visit.
Contents · 8 sections
- 01ChatGPT traffic share report: what share of our visitors came from ChatGPT?
- 02ChatGPT traffic share by site: why is it 31% on one blog and 0% on another?
- 03ChatGPT traffic share week by week: is it growing?
- 04How often does ChatGPT read a page for each person it sends?
- 05What about Perplexity, Gemini, Claude and Copilot traffic?
- 06How do I count ChatGPT traffic share, and what can this report not tell you?
- 07Frequently asked questions
- 08Sources
This ChatGPT traffic share report is for you if you run a young site and want to know whether ChatGPT is a real traffic source or a talking point. I read the daily statistics files of five sites in my fleet on 13 Nov 2026 and added them up by site and by month. No estimates, no third-party panels: just people my counters saw, with bots and my own visits removed.
The numbers are small, because the sites are small. madrydochod.pl launched on 6 Oct 2026, the others between 19 and 27 Oct 2026. That is the point: it shows ChatGPT's share on sites that are weeks old. I will repeat it month by month, and the live fleet total sits on the Empire Counter.
ChatGPT traffic share report: what share of our visitors came from ChatGPT?
From 1 to 12 Nov 2026, ChatGPT sent 84 of the 477 people who visited five of my sites, a share of 17.6%. In the October days the counters covered, it was 62 of 299 people, or 20.7%. Both figures come from server-log counters read on 13 Nov 2026, with bots and my own visits removed.
| Site | Period | Days | People | From ChatGPT | Share | ChatGPT-User requests |
|---|---|---|---|---|---|---|
| madrydochod.pl | 10–31 Oct | 22 | 191 | 55 | 28.8% | 1,222 |
| madrydochod.pl | 1–12 Nov | 12 | 165 | 51 | 30.9% | 773 |
| tillradar.co.uk | 19–31 Oct | 13 | 55 | 5 | 9.1% | 144 |
| tillradar.co.uk | 1–12 Nov | 12 | 75 | 22 | 29.3% | 363 |
| mainstreetradar.com | 25–31 Oct | 7 | 21 | 2 | 9.5% | 229 |
| mainstreetradar.com | 1–12 Nov | 12 | 51 | 10 | 19.6% | 460 |
| kontowtescie.pl | 26–31 Oct | 6 | 29 | 0 | 0% | 0 |
| kontowtescie.pl | 1–12 Nov | 12 | 93 | 1 | 1.1% | 48 |
| gregciupek.com | 27–31 Oct | 5 | 3 | 0 | 0% | 0 |
| gregciupek.com | 1–12 Nov | 12 | 93 | 0 | 0% | 2 |
| Fleet (5 sites) | October days | 53 | 299 | 62 | 20.7% | 1,595 |
| Fleet (5 sites) | 1–12 Nov | 60 | 477 | 84 | 17.6% | 1,646 |
"People" is the sum of daily visitors, so someone who came back on two different days counts twice. Each site starts on the first day its counter ran on the current method (madrydochod.pl on 10 Oct 2026), which is why the October rows cover different numbers of days. "ChatGPT-User requests" is a different thing: page fetches by ChatGPT's user agent, explained further down.
The fleet share fell from October to November even though ChatGPT sent more people (84 against 62). The reason is the denominator: gregciupek.com and kontowtescie.pl grew from Google and direct visits, and they get almost nothing from ChatGPT. On 10 Nov 2026 alone, gregciupek.com counted 58 people, 48 of them direct, on desktop, each opening a different page, which looks more like one person or tool going through the site than 48 readers. The counter still counts them, so I do too. Without that day, the November fleet share would be 84 of 419, or 20.0%.
ChatGPT traffic share by site: why is it 31% on one blog and 0% on another?
ChatGPT's share of people ranged from 30.9% on madrydochod.pl to 0% on gregciupek.com in 1–12 Nov 2026. The sites that answer narrow, practical questions (bank accounts, card readers, business banking) get ChatGPT clicks; this site, a few weeks old and full of broad guides, gets its visitors from Google and direct visits instead.
Three things separate the high-share sites from the rest, as far as I can see in the logs:
- Age and volume. madrydochod.pl had been publishing six articles a day for a month before November. More pages means more chances that one of them is the source ChatGPT links to.
- Question-shaped topics. On 12 Nov 2026, when ChatGPT sent 4 of tillradar.co.uk's 6 people, the pages read that day included a November list of UK business account offers and a comparison of corporation tax filing software: specific questions with a dated answer.
- ChatGPT's search index. The sites with a high share also get the most visits from OpenAI's search crawler (OAI-SearchBot, labeled "ChatGPT Search" in my counter): 902 requests on madrydochod.pl in November, 31 on gregciupek.com.
kontowtescie.pl is the odd one. It gets steady Google traffic (79 of its 93 November people came from Google) and only 1 person from ChatGPT, on 10 Nov 2026. It is mostly an index with 12 bank pages, which may be why ChatGPT rarely cites it; the logs do not prove that.
How I set the sites up for ChatGPT is in How to get cited by ChatGPT. The full first month of madrydochod.pl, posts and money included, is in the affiliate blog case study.
ChatGPT traffic share week by week: is it growing?
On madrydochod.pl, ChatGPT's weekly share of people went from 22.5% (12–18 Oct 2026) to about a third from the second half of October on, with 35.2% in the four days of 9–12 Nov 2026. The number of people from ChatGPT roughly doubled, but so did everything else, so the share stopped climbing.
| Week | People | From ChatGPT | Share |
|---|---|---|---|
| 12–18 Oct 2026 | 40 | 9 | 22.5% |
| 19–25 Oct 2026 | 71 | 23 | 32.4% |
| 26 Oct–1 Nov 2026 | 80 | 26 | 32.5% |
| 2–8 Nov 2026 | 104 | 29 | 27.9% |
| 9–12 Nov 2026 (4 days) | 54 | 19 | 35.2% |
As a cross-check, the fact register I keep for this site has madrydochod.pl at 80 people and 27 from ChatGPT (34%) for 17–25 Oct 2026. The fresh read on 13 Nov 2026 gives exactly the same 80 and 27 for those days, so the files have not been rewritten since.
The UK and US sites moved faster in relative terms. tillradar.co.uk got its first ChatGPT visitor on 27 Oct 2026 and went from 9.1% in October to 29.3% in November. mainstreetradar.com went from 9.5% to 19.6%. With 5 and 2 people in October, one person more or less moves those percentages by several points.
First person from ChatGPT on each site: madrydochod.pl 13 Oct 2026 (4 people that day), tillradar.co.uk 27 Oct 2026, mainstreetradar.com 28 Oct 2026, kontowtescie.pl 10 Nov 2026, gregciupek.com none yet (as of 12 Nov 2026).
How often does ChatGPT read a page for each person it sends?
In 1–12 Nov 2026, ChatGPT's user agent fetched pages on madrydochod.pl 773 times while 51 people clicked through, about 15 requests per visitor. On tillradar.co.uk it was 363 requests for 22 people, about 17 per visitor, and on mainstreetradar.com 460 for 10. Most of the time ChatGPT reads a page, the user never visits.
OpenAI's documentation says ChatGPT may visit a web page with the ChatGPT-User agent when a user asks ChatGPT a question, and that this agent is not used for automatic crawling. So each of those requests is tied to someone's conversation. My counter records them separately from people, because no browser ping comes with them.
Two caveats. A single answer can fetch several pages, and the home page shows up often (38 of the 72 ChatGPT-User requests on madrydochod.pl on 12 Nov 2026 were for "/"), so requests are not conversations. Still, the gap between 1,646 fleet requests and 84 people in November is the clearest thing in this report: being read by ChatGPT and getting a visitor from ChatGPT are two very different numbers.
If you only look at Google Analytics, you see neither number well. ChatGPT-User requests never run your analytics script, and some ChatGPT clicks arrive without a referrer and land in "direct". Server logs are the only place I can see both.
What about Perplexity, Gemini, Claude and Copilot traffic?
Across all five sites and every day counted up to 12 Nov 2026, other AI assistants sent exactly 1 person: a visit from Gemini to madrydochod.pl on 27 Oct 2026. Perplexity, Claude and Copilot sent nobody my counters could see, so for my fleet "AI traffic" currently means ChatGPT.
Those assistants do read the sites. In November, Claude's user agent fetched pages on madrydochod.pl 36 times and Perplexity's user agent fetched one page on mainstreetradar.com. One visit on madrydochod.pl on 3 Nov 2026 carried an OpenAI campaign tag; the counter filed it under campaigns, and I left it there.
How do I count ChatGPT traffic share, and what can this report not tell you?
A person in my counter is a browser that loaded a page and sent a small ping back to my server; requests from known bots, data-center addresses, my own addresses and pages with no ping are dropped. The source is read from where the visit came from, and "ChatGPT" means the visit was attributed to ChatGPT.
If you want the same report for your own site, this is the order I would follow:
- Count people, not hits. Use a measure that needs a real browser (a script ping or a consented analytics hit), otherwise bots swamp everything. On 12 Nov 2026, madrydochod.pl dropped 635 bot-like visitors against 22 people.
- Split sources per day. Store each day with people per source, so you can add up any period later instead of trusting a dashboard's preset ranges.
- Log AI user agents separately. Count ChatGPT-User, OAI-SearchBot and similar agents by page. They tell you what ChatGPT reads even when nobody clicks.
- Report with a date and a denominator. "31%" means nothing without "of 165 people, 1–12 Nov". Small denominators move fast.
What this report cannot tell you: it does not cover dziennikfirm.pl and the two Companies Daily sites, which joined in November, are kept out of Google and only count readers who accept statistics. It does not show what ChatGPT traffic is worth; for one site, that is in the case study linked above. And five young sites are a lab, not a market sample, so do not read 17.6% as "the share ChatGPT has on the web". If you are at the very start, how to get traffic to a new website covers the other channels.
The live total for the whole fleet, updated from the same logs, is below. Other reports are collected under Reports.
| Site | People | Page views | From Google and Bing | From ChatGPT and other AI |
|---|---|---|---|---|
| The Online Empire gregciupek.com · this site, English, since 28 Oct 2026 | 275 | 310 | 94 | 29 |
| Internetowe Imperium (PL) internetoweimperium.pl · Polish edition of The Online Empire, since 26 Oct 2026 | 142 | 367 | 22 | 0 |
| Mądry Dochód (Poland) madrydochod.pl · affiliate blog, since 6 Oct 2026 | 402 | 613 | 148 | 132 |
| TillRadar (UK) tillradar.co.uk · UK business banking, since 20 Oct 2026 | 278 | 314 | 128 | 51 |
| Konto w Teście (Poland) kontowtescie.pl · bank reliability index, since 26 Oct 2026 | 145 | 183 | 97 | 9 |
| MainStreetRadar (US) mainstreetradar.com · US business banking, since 27 Oct 2026 | 156 | 167 | 61 | 22 |
| Dziennik Firm (Poland) dziennikfirm.pl · Polish business daily from the company register (KRS), since 5 Nov 2026 (counter since 6 Nov, counts only people who accepted statistics) | 214 | 498 | 176 | 3 |
| Companies Daily (UK) companiesdaily.co.uk · UK business daily from Companies House, since 6 Nov 2026 (counter since 7 Nov, counts only people who accepted statistics) | 173 | 244 | 148 | 11 |
| Companies Daily US companiesdaily.com · US business daily from SEC EDGAR filings, since 11 Nov 2026 (counter since 12 Nov, counts only people who accepted statistics) | 10 | 11 | 10 | 0 |
| Total | 1,795 | 2,707 | 884 | 257 |
Source: web01 server logs, counted by a script every 30 minutes. A person is a browser that ran our script; bots, data-centre addresses and our own visits are removed. Google Analytics shows less because it runs only after consent. How we count
Frequently asked questions
What share of traffic comes from ChatGPT on a new site?
On my five sites, 17.6% of people came from ChatGPT in 1–12 Nov 2026 (84 of 477), according to server logs read on 13 Nov 2026. The range was wide: 30.9% on one Polish blog and 0% on gregciupek.com. It depends on the site, not on a general rate.
Is ChatGPT traffic share growing?
On madrydochod.pl the weekly share went from 22.5% in mid-October 2026 to roughly a third after that, with 35.2% in 9–12 Nov 2026. On tillradar.co.uk and mainstreetradar.com it rose from about 9% in October to 29.3% and 19.6% in November, but from very small numbers.
Why does Google Analytics show less ChatGPT traffic than server logs?
Some clicks from ChatGPT arrive without a referrer and are counted as direct, and visitors who block or decline analytics are missing. ChatGPT's own page fetches never run analytics scripts at all. Server logs see the requests, so they show both the reading and the clicks.
Do Perplexity or Gemini send traffic to small sites?
Not much to mine yet. Across five sites up to 12 Nov 2026, Gemini sent 1 person and Perplexity, Claude and Copilot sent none my counters could see. Their user agents do fetch pages, just rarely.
What is the ChatGPT-User agent?
It is the agent ChatGPT uses when it visits a page during a user's conversation, according to OpenAI's bot documentation (checked 13 Nov 2026). It is not used for automatic crawling. On my sites it made 15 to 17 requests for each person who clicked through from ChatGPT in November 2026.
Sources
- My server-log counters on web01, daily files for madrydochod.pl (md-stats), gregciupek.com (oe-stats), tillradar.co.uk (tr-stats), mainstreetradar.com (msr-stats) and kontowtescie.pl (kw-stats), 10 Oct–12 Nov 2026, read 13 Nov 2026: live totals on the Empire Counter
- The Online Empire fact register, madrydochod.pl traffic 17–25 Oct 2026 (80 people, 27 from ChatGPT)
- OpenAI, Overview of OpenAI Crawlers (checked 13 Nov 2026): platform.openai.com
As of 13 Nov 2026. I read the daily statistics files of five sites on 13 Nov 2026 and added up every full day from the start of each counter's current method to 12 Nov 2026 (the last full day). People are summed daily visitors; UK and US sites cut days in London and New York time. OpenAI's crawler documentation checked 13 Nov 2026. Next update: December 2026, with November complete.



