ChatGPT-User

Budgius messengerus

Cursores, the couriers. AI assistant, sent by a person.

A joke about the drawing A budgie with something to say, on behalf of someone.

The line above plays on the name. It says nothing about how the bot behaves: everything below is read from the request log, for today and the 29 days before it, by UTC date.

Alignment

Insufficient data. A judgement needs at least 200 requests and 5 sessions in the trailing 30 days. This one would rest on unverified sessions, so it is a statement about the species and not about any operator: 7 requests in 1 session over the trailing 30 days.

How the score is worked out, and the count behind every input, is on the Census page.

Who runs it, and how that is checked

Who runs it: OpenAI
A user-agent string it sends: Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot
What it says it is for: ChatGPT-User fetches a page when a person using ChatGPT asks something that needs it. OpenAI says it is not used to crawl the web automatically.
How its real one can be told, as its operator says: OpenAI publishes the addresses this bot uses.

Everything above is a claim: the operator’s own account where it has published one, and a third party’s where it has not. This site has not tested it. What the bot has done is in the sections below, read from the request log.

Customs has no documented check for it. A session that says it is ChatGPT-User can be neither confirmed nor refuted, so every one is shown as unverified and none is put down to OpenAI.

The name it is matched on

A request is filed under this species when its user-agent string contains the product name ChatGPT-User as a whole word, in any mix of capitals. Where a string holds more than one product name, the longest wins, and a named bot wins over a library it is built on.

How to address this bot in robots.txt

A group in robots.txt addresses this bot by the line:

User-agent: ChatGPT-User

An example group. The path is a placeholder.

User-agent: ChatGPT-User
Disallow: /a-path/
Allow: /a-path/one-page

This is reference, not advice. Systema Machinae is not a bot-blocking product.

This site’s own robots.txt names it, in the group that keeps the guest list off the Private Beach.

Landings

1 landing in the trailing 30 days. A landing is the first request of a session.

Verified sessions passed Customs. Unverified sessions said they were ChatGPT-User and could not be checked, or have not been answered yet: they are shown for the species and are not put down to any operator. Sessions that failed are not on this page at all.

By day

VerifiedUnverified

The same figures as a table

Landings on each of the last 30 days, UTC
Day	Unverified	Total
9 Sep	0	0
10 Sep	0	0
11 Sep	0	0
12 Sep	0	0
13 Sep	0	0
14 Sep	0	0
15 Sep	0	0
16 Sep	0	0
17 Sep	0	0
18 Sep	0	0
19 Sep	0	0
20 Sep	0	0
21 Sep	0	0
22 Sep	0	0
23 Sep	0	0
24 Sep	0	0
25 Sep	0	0
26 Sep	0	0
27 Sep	0	0
28 Sep	0	0
29 Sep	0	0
30 Sep	0	0
1 Oct	0	0
2 Oct	0	0
3 Oct	0	0
4 Oct	0	0
5 Oct	0	0
6 Oct	0	0
7 Oct	0	0
8 Oct	1	1

By hour of the day, UTC

VerifiedUnverified

The same figures as a table

Landings in each hour of the UTC day, over the last 30 days
Hour	Unverified	Total
00:00	0	0
01:00	0	0
02:00	0	0
03:00	0	0
04:00	0	0
05:00	0	0
06:00	0	0
07:00	0	0
08:00	0	0
09:00	0	0
10:00	0	0
11:00	0	0
12:00	0	0
13:00	0	0
14:00	0	0
15:00	0	0
16:00	0	0
17:00	0	0
18:00	0	0
19:00	0	0
20:00	0	0
21:00	0	0
22:00	1	1
23:00	0	0

Busiest hour: 22:00 UTC, with 1 landing.

What it asks for

Part of the site	Verified requests	Unverified requests
The island, on the home page	0	1
Species pages	0	1
The Private Beach	0	1
The Turnstile	0	1
The Side gate	0	1
The vault, behind the Forbidden Door	0	1
The Pitfall	0	1

Parts of the site, from a fixed list. The address of a request is never shown.

Its record with each rule

Trailing 30 days	Verified	Unverified
Sessions	0	1
Requests	0	7
Sessions that asked for anything besides robots.txt	0	1
read robots.txt first	0	0
had read it in the 24 hours before	0	0
read it only afterwards	0	0
did not read it	0	1
Sessions that entered the Pitfall	0	0
having read robots.txt	0	0
Requests turned back at the vault, with a 403	0	0
Highest Climb level reached	Has not climbed	Has not climbed
Private Beach visits while on the guest list	0	0
Turnstile waits kept	0	0
Turnstile waits broken	0	0
Requests over the rate limit	0	0
sessions that went over it	0	0
Fetches for a disallowed path, sent on someone’s behalf (not counted)	0	3

These are counts, not a verdict. A session is judged on what it did inside the 30 days.

Customs

Sessions that said they were this bot	Trailing 30 days
Passed: verified	0
Could not be checked: unverified	1
Not yet answered: unverified	0
Failed, and filed under Googlebot (unverified)	0

Nothing a failed session did is counted anywhere else on this page.

Where it comes from

Unverified sessions, the busiest 1
Network owner	Country	Sessions	Requests
Microsoft Corporation	United States	1	7

Network owner and country only. An address is never shown.

Sent on someone’s behalf

A bot in this order fetches a page because a person, or a schedule a person set, asked it to. When that page is one robots.txt disallows, the fetch is shown here and is not counted as breaking a rule: 3 such fetches in the trailing 30 days, in 1 session. The order is scored on identity and rate only.