ChatGPT Business Hub Counts Clicks. My Logs Counted 3.
On this page, 9 sections
ChatGPT Business Hub counts clicks. I counted mine too.
On 10 November, OpenAI opened Business Hub in beta. You claim your business, and after identity verification an Insights tab shows how often ChatGPT mentioned your website and how many people clicked through, over the last 28 days.
I wanted to know what that number looks like from the other side, so I read my own server log before looking at anyone’s dashboard.
Short version: in 9 days, ChatGPT-labelled requests fetched my pages at most 85 times, and I can find 3 pages where a visitor arrived from ChatGPT.
Every outside fact here was read on 22 November 2026, and the log numbers are mine.
This post shows how I counted, what is shaky about it, and what I would put next to Business Hub numbers if you claim your site.
1. What my log says
The log covers 12 November 23:58 UTC to 22 November 14:35 UTC, about 9.6 days, and 327,055 requests.
My site sits behind Cloudflare, and the user agent still reaches the server. I counted by user agent string only. I did not check IP ranges, so every number below is “labelled as”, not “proven to be”.

OpenAI’s crawler page lists three agents you will see. GPTBot and OAI-SearchBot crawl on their own. ChatGPT-User is the one that visits a page when a person asks ChatGPT something, and OpenAI says it is not used to crawl the web automatically.
So ChatGPT-User is the closest thing to “someone asked about my page”. It showed up 137 times.
2. Why 137 became 85
Of those 137 requests, 52 asked for paths like /secrets.env, /key.json and /service_account.json. None of those exist on my site. They all returned 404.
ChatGPT does not go hunting for credential files. A scanner is borrowing the name. That is the whole reason I stopped trusting user agent counts by themselves.
Strip those out and 85 requests remain, 83 with a 200 and 2 redirects. 57 of them were the home page. The rest were single posts, for example my pages on YubiKeys, Cloudflare Scrape Shield and robots.txt for AI crawlers.
I would still call 85 an upper bound. Fake user agents that behave politely look exactly like real ones. If you want to tell them apart, I wrote up the check in how to tell a real GPTBot from a fake one.
3. How many visitors came back
ChatGPT adds utm_source=chatgpt.com to links it opens. I searched the log for it.
Nine requests carried it. Most were page assets and a plugin ping. The real landing pages were three: a post on converting a video to a Live Photo, a post on RoutineHub alternatives, both on 13 November, and a post comparing memory layers for agents on 17 November.
Three landing pages in 9.6 days, on a site that served about 34,000 requests a day.
I also searched the referrer field for claude.ai, perplexity.ai, gemini and copilot. Nothing. The only AI referrer in the whole log was that one ChatGPT visit and the assets it loaded.
4. The other bots, for scale
The same log shows GPTBot 889 times and OAI-SearchBot 443.
ClaudeBot 463, Claude-User 100 and Claude-SearchBot 68. PerplexityBot 658 and Perplexity-User 67.
For comparison, Googlebot made 1,928 requests, Bingbot 2,323, and Meta’s external agent 6,097.
Read that as “who is looking”, not “who is sending traffic”. Meta’s crawler is by far the busiest visitor in this log, and it tells me nothing about readers.
I went deeper on what each crawler does in which AI crawlers visit my site and on the OpenAI trio in GPTBot, OAI-SearchBot and ChatGPT-User.
5. What Business Hub measures, and what it does not
Here is the part of OpenAI’s help page that matters for comparing numbers.

For an online business, a mention counts when ChatGPT references your website. The brand name alone does not count.
Website clicks show how often people click through. Insights cover 28 days and appear only after enough conversations to pass a threshold. OpenAI also says a blank metric does not always mean zero activity.
Search Engine Journal reported one early example: Search Engine Journal itself saw 353,356 mentions and 2,663 website clicks between 14 October and 10 November. That is roughly 0.75 clicks per 100 mentions. One site, one window, and OpenAI’s own definitions, so I would not treat it as a rate for anyone else.
I have not put my own Business Hub numbers in this post. Everything above comes from my server log.
6. How I would compare the two
If you claim your site, do this once the first 28 days have passed:
- Take the Business Hub clicks for the last 28 days.
- Count page loads in your own log or analytics with utm_source=chatgpt.com in the same 28 days.
- Count ChatGPT-User requests that returned 200 on real pages, and drop the ones asking for files that do not exist.
- Put the three numbers side by side before you tell anyone the result.
OpenAI says a recorded click is not a confirmed visit. If the log shows fewer visits than Business Hub shows clicks, that gap is normal and worth knowing, not a bug to chase.
If the log shows more, check that the referrer is not being stripped somewhere between ChatGPT and your analytics. One roundup of this month’s analytics studies reports AI-shaped visits landing in search and direct traffic, so a low count can hide real readers.
7. What I would do next
Save utm_source=chatgpt.com as a segment in your reporting, and add a weekly count of ChatGPT-User hits that return 200. Nothing fancy. A 10 minute check on Monday mornings is enough.
The bigger lesson is a small one. Bots read my pages all day, and a very small number of people arrive from an AI answer. Writing for the reader who clicks still matters more than writing for the crawler.
If you run a site, look at your own log before you read anyone’s benchmark, including mine. Mine is 9 days and one site.
Frequently asked questions
Does ChatGPT-User prove a person asked about my page?
Not by itself. OpenAI says ChatGPT-User visits a page when a user asks ChatGPT something, but anyone can send that name in a request. In my log, 52 of 137 ChatGPT-User requests were probing for files that do not exist.
How do I count visitors that came from ChatGPT?
Search your log or analytics for utm_source=chatgpt.com on page requests, not on images or scripts. In my log, 9 requests carried it, and 3 were real landing pages.
Is claiming my site in ChatGPT Business Hub worth it?
OpenAI says you do not need an Ads account or an Enterprise agreement to claim a business, and the Insights tab shows mentions and website clicks for 28 days after you pass identity verification. Whether it brings visitors is a different question, and my own log suggests the numbers are small for a site like mine.