Flagship · 28 Oct 2026 · updated 15 Nov 2026

How to Get Your Site Cited by ChatGPT: What Our Server Logs Show

Server logs: 36 of 108 readers (33%) came from ChatGPT in 11 days. Which OpenAI robots did it, how fast new domains got cited, and what did not help.

Greg Ciupek 23 min read 7 sources Comment
Cover: 36 of 108 readers of a new blog came from ChatGPT between 17 and 27 Oct 2026, according to server logs
Short answer · as of 28 Oct 2026
  • On my blog madrydochod.pl, 108 people arrived in 11 days (17–27 Oct 2026, server logs) and 36 of them (33%) clicked a link in a ChatGPT answer. The blog was 11 days old when that period started.
  • ChatGPT reads a site through two robots: OAI-SearchBot builds the index for ChatGPT search, and ChatGPT-User fetches a page while someone is asking. In those 11 days they made 1,073 and 735 requests. The training robot GPTBot made 11.
  • On a new domain the order was the same twice: OAI-SearchBot read the site on day 3 (madrydochod.pl, 9 Oct 2026) and the first people from ChatGPT came on day 7. My UK site tillradar.co.uk got its first ChatGPT reader on day 8 (27 Oct 2026).
  • What brought the traffic were basics: OAI-SearchBot allowed in robots.txt and not blocked by the server, content in plain HTML, the answer in the first sentence, numbers with an "as of" date. In today's logs no AI crawler requested llms.txt.
  • Google says (10 Dec 2025) that AI Overviews and AI Mode need no special files or markup: a page must be indexed and eligible to show with a snippet.
Contents · 10 sections
  1. 01How to get cited by ChatGPT: what does it actually mean?
  2. 02How much traffic can ChatGPT send to a new site? My logs from 17–27 Oct 2026
  3. 03How does ChatGPT find a website? OpenAI's robots in my logs
  4. 04How long does it take to get traffic from ChatGPT on a new domain?
  5. 05What did I change on my sites to get cited by ChatGPT?
  6. 06How to get cited by ChatGPT step by step: the first 30 days
  7. 07Do llms.txt, an MCP server and an "agent-ready" score help you get cited by ChatGPT?
  8. 08How do you measure ChatGPT traffic when Google Analytics does not see all of it?
  9. 09Frequently asked questions
  10. 10Sources

This is for you if you have a new site or blog, Google has not noticed you yet, and everyone says "AI search is what matters now". I will show how much traffic ChatGPT sent to my youngest sites in their first weeks, how I know (server logs, not Google Analytics), and which changes I think brought that traffic and which have brought nothing so far.

I do not promise it will work the same for you. I show what I did, in what order, and what the logs say. The numbers are small because the sites are weeks old, and that is exactly why they can be traced person by person. Most of the data comes from madrydochod.pl, a Polish affiliate blog I launched on 6 Oct 2026, and from its English sister sites tillradar.co.uk (UK, since 19 Oct 2026) and mainstreetradar.com (US, since 25 Oct 2026). The robots are the same in every language.

How to get cited by ChatGPT: what does it actually mean?

Getting cited by ChatGPT means that ChatGPT fetches your page, uses it to build an answer and shows a link to it that someone clicks; on madrydochod.pl between 17 and 27 Oct 2026, 36 people arrived through such links. The work behind it is usually called GEO (generative engine optimization) or AEO (answer engine optimization), and it overlaps heavily with ordinary SEO.

The difference from classic SEO is simple. In a search engine you compete for a place in a list of ten links. In an AI assistant there is no list: there is one answer built from a few sources, with a handful of links next to it. Either you are one of them, or you are not there.

The second difference is who reads your page. In SEO it is Googlebot. In ChatGPT it is OpenAI's own robots, which work independently of Google. OpenAI documents them (checked 28 Oct 2026), and it is worth knowing them by name, because they decide whether you are in the game at all.

Where AI answers come from and what you need to do, as of 28 Oct 2026
AssistantWhere it gets pagesWhat you must doSource
ChatGPT searchIts own index built by OAI-SearchBot; during a conversation ChatGPT-User fetches the pageAllow OAI-SearchBot in robots.txt and do not block OpenAI's IP addresses on the serverOpenAI crawler documentation
Google AI Overviews and AI ModeThe normal Google indexThe same as SEO: indexed and eligible to show with a snippetGoogle Search Central, 10 Dec 2025
Claude, Perplexity, Gemini and othersTheir own robots and on-request fetchesDo not block them in robots.txt or on the servermy logs: almost no human traffic from them yet

One thing from this table is worth remembering from the start. For Google, "AI search optimization" is simply SEO: Google writes that you do not need new machine-readable files, AI text files or special schema.org markup to appear in AI Overviews. For ChatGPT, what matters first is whether OAI-SearchBot can get in, and whether your page answers the question once it is there.

How much traffic can ChatGPT send to a new site? My logs from 17–27 Oct 2026

On madrydochod.pl in the eleven days from 17 to 27 Oct 2026 there were 108 people, and 36 of them (33%) came from ChatGPT answers, plus 1 from Gemini. In the week before, 10–16 Oct 2026, ChatGPT sent 4 of 38 people. I count humans from server logs, with robots and data-center traffic removed. How that share moved month by month across five sites is in my ChatGPT traffic share report.

36people from ChatGPT in 11 days (17–27 Oct 2026)
33%of all madrydochod.pl readers in that period
4 of 38people from ChatGPT the week before (10–16 Oct 2026)
108people in 11 days: a small site, not big traffic

Below is the day-by-day split. The last column is the number of ChatGPT-User requests, that is, how many times ChatGPT fetched one of my pages because someone in a conversation asked a question it was looking up.

madrydochod.pl: people by source and ChatGPT-User fetches, 17–27 Oct 2026 (server logs, read 28 Oct 2026)
DayPeopleFrom ChatGPTFrom searchDirectChatGPT-User requests
17 Oct320155
18 Oct624047
19 Oct1219239
20 Oct1349050
21 Oct622279
22 Oct1869393
23 Oct852167
24 Oct412156
25 Oct1043376
26 Oct1136284
27 Oct1767389
Total108365318735

The source columns add up to 107. Add 1 person from Gemini (27 Oct) and 1 from a social campaign link (22 Oct) and you get 109, one more than 108, because on 22 Oct one person came twice from different sources. A person from ChatGPT is a browser that arrived from chatgpt.com or with the utm_source=chatgpt.com parameter that ChatGPT adds to links in its answers.

How do I count a "person"? The browser has to run our small cookie-free script, must not have a robot's name, must not come from a data-center address and must not be me (the addresses I use for the WordPress dashboard are excluded). The method is described on the Empire Counter page.

Watch out

33% sounds great, but it is a share of a small number. The blog had 141 posts published in 20 days (26 Oct 2026), and on 15 Oct 2026 41% of its posts were still not indexed by Google. When Google catches up, ChatGPT's share will fall even if the number of people from ChatGPT keeps growing. Watch the number of people, not the percentage.

The second caveat: 735 ChatGPT-User fetches and 36 clicks is about 20 fetches for every person who came. Most ChatGPT answers that used my page as a source did not end in a visit. The person got the answer and moved on. That is normal, and worth knowing before you calculate what AI traffic is "worth".

Here is the live counter for my whole fleet, with a separate column for people from ChatGPT and other assistants. It is recalculated every 30 minutes from the same logs.

People on the fleet's sites in the last 7 days (to 21 Nov 2026)
SitePeoplePage viewsFrom Google and BingFrom ChatGPT and other AI
The Online Empire
gregciupek.com · this site, English, since 28 Oct 2026
1491674524
Internetowe Imperium (PL)
internetoweimperium.pl · Polish edition of The Online Empire, since 26 Oct 2026
3776100
Mądry Dochód (Poland)
madrydochod.pl · affiliate blog, since 6 Oct 2026
1121533337
TillRadar (UK)
tillradar.co.uk · UK business banking, since 20 Oct 2026
1361496719
Konto w Teście (Poland)
kontowtescie.pl · bank reliability index, since 26 Oct 2026
121614
MainStreetRadar (US)
mainstreetradar.com · US business banking, since 27 Oct 2026
6669257
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)
1383261183
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)
2239410
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)
1011100
Total6821,006313104

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

How does ChatGPT find a website? OpenAI's robots in my logs

ChatGPT finds a website through two OpenAI robots: OAI-SearchBot adds it to the ChatGPT search index, and ChatGPT-User fetches it live when a user asks a question; on madrydochod.pl between 17 and 27 Oct 2026 they made 1,073 and 735 requests. The third robot, GPTBot, collects training data, made 11 requests and has no visible link to traffic.

OpenAI's documentation (checked 28 Oct 2026) says sites that opt out of OAI-SearchBot "will not be shown in ChatGPT search answers". Each setting is independent, so you can allow OAI-SearchBot and block GPTBot if you do not want your writing used for training. ChatGPT-User is not used to crawl the web automatically; it acts on a person's request, and OpenAI notes that robots.txt rules may not apply to it.

Robots on madrydochod.pl, 17–27 Oct 2026 (requests from server logs, recognized by user agent)
RobotWhat it is forRequestsAnswered 2xx
OAI-SearchBotChatGPT search index1,0731,073
GooglebotGoogle index, also used for AI Overviews752717
ChatGPT-Userpage fetched during a ChatGPT conversation735731
Meta AIMeta's AI crawler415173
ClaudeBotAnthropic's crawler222222
BingbotBing index188184
Claude-Userpage fetched during a Claude conversation5839
GPTBottraining data for OpenAI models1111

The most interesting part of this table is what it does not show: people. In those days not a single person came from Claude, despite 58 Claude-User fetches, and nobody from Perplexity. Almost all the human traffic from AI assistants was ChatGPT (36 people), plus one person from Gemini. On a small, new site, ChatGPT is today the only assistant that sends readers in measurable numbers.

The 4 ChatGPT-User requests that were not 2xx were redirects (3xx), not refusals. Meta AI is a different story, and I come back to it in the section on blocking.

How long does it take to get traffic from ChatGPT on a new domain?

On my two new domains the first person from ChatGPT arrived on day 7 (madrydochod.pl, 13 Oct 2026) and on day 8 (tillradar.co.uk, 27 Oct 2026); in both cases OAI-SearchBot had read the site in bulk a few days earlier. All numbers below come from daily statistics files computed from server logs.

madrydochod.pl: the first days through the eyes of AI robots, 6–17 Oct 2026
DayWhat happened
6 Oct (launch)GPTBot fetches 130 pages, OAI-SearchBot 3. People from ChatGPT: 0.
7–8 OctGPTBot disappears (0 requests), OAI-SearchBot 0 and 2.
9 Oct (day 3)OAI-SearchBot: 168 requests. ChatGPT-User fetches my pages for the first time: 21 requests on 13 different URLs.
11 OctOAI-SearchBot: 236 requests.
12 OctChatGPT-User: 59 requests a day.
13 Oct (day 7)The first 3 people who clicked a link in a ChatGPT answer.
15 Oct"Agent-ready" changes deployed, isitagentready score 93/100.
17–27 Oct36 people from ChatGPT; ChatGPT-User 39–93 requests a day; GPTBot 1 a day.

The English site repeated the pattern. tillradar.co.uk launched on 19 Oct 2026 with OAI-SearchBot fetching 13 pages on day one and GPTBot 149 on day two. ChatGPT-User appeared for the first time on 24 Oct with a single request. On 27 Oct OAI-SearchBot made 132 requests, ChatGPT-User 20, and the first person from ChatGPT arrived. The US site mainstreetradar.com, launched on 25 Oct, already had 90 OAI-SearchBot requests on 26 Oct and 58 ChatGPT-User requests on 27 Oct, but no person from ChatGPT yet.

Three new domains, OpenAI robots and the first ChatGPT reader, as of 28 Oct 2026
SiteLaunchFirst big OAI-SearchBot dayFirst ChatGPT-User fetchFirst person from ChatGPT
madrydochod.pl (PL)6 Oct 20269 Oct (168 requests)9 Oct13 Oct (day 7)
tillradar.co.uk (UK)19 Oct 202627 Oct (132 requests)24 Oct27 Oct (day 8)
mainstreetradar.com (US)25 Oct 202626 Oct (90 requests)26 Octnot yet

The conclusion from this timeline is clear to me. The training robot GPTBot came first, then practically stopped coming, and had nothing to do with traffic. Traffic was decided by OAI-SearchBot reading the site, and by ChatGPT-User starting to fetch pages in conversations shortly after. On madrydochod.pl the first clicks came before I had deployed anything from the "agent-ready" standards.

What ChatGPT likes to read

The two OpenAI robots read different pages. Between 17 and 27 Oct 2026 ChatGPT-User fetched 140 different URLs on madrydochod.pl. The homepage got 300 fetches; after that, the most fetched were single-offer pages: one offer, one amount, one deadline, one date checked in the terms. Pages that answer one specific question.

The lesson is a simple rule: one page, one question, a concrete number with a date. A long essay where the answer is in the seventh paragraph loses to a page that gives it in the first sentence.

What did I change on my sites to get cited by ChatGPT?

Before the first click from ChatGPT on madrydochod.pl (13 Oct 2026) the site had: an open robots.txt with AI robots named, content in plain HTML, a "Short answer" box on top of articles, "as of" dates and structured data, while the agent-ready changes (Content-Signal, Markdown, API, an MCP server, a 93/100 score) arrived only on 15 Oct 2026.

To be honest: I have no control group, so I cannot prove that a specific change produced a specific click. What I have is dates. Here is the full list, split into what was in place before the first ChatGPT visits and what came later.

madrydochod.pl: changes for search engines and AI assistants, 6–15 Oct 2026
ChangeWhenWhyBefore the first ChatGPT click?
robots.txt with separate groups for OAI-SearchBot, ChatGPT-User, GPTBot, ClaudeBot, PerplexityBot and others, all with Allow: /first weeklet AI robots in explicitlyyes
Everything rendered on the server: tables, amounts, deadlines, calculator results are in the HTML, not built by JavaScriptfrom launchAI robots usually do not run scriptsyes
"Short answer" box and the first sentence of each section as a self-contained answerfrom launcha passage an assistant can quote wholeyes
"As of" date and the date the source document was checked next to numbersfrom launcha fresh, verifiable factyes
One page per offer or per questionfrom launcha page that answers directlyyes
Structured data: publisher with registry numbers, author, article, FAQ, breadcrumbsfrom launch, fixes on 13 Octan unambiguous "who, what, when"yes
Expired offers marked "Ended" in title, description and structured data, with replacements13 Octso AI does not quote outdated amountssame day
llms.txt listing all posts, and llms-full.txt with facts13 Octa table of contents for language modelssame day
Content-Signal in every robots.txt group, Markdown for agents, read-only API, MCP server, Link headers15 Octagents can use the site without parsing HTMLno
Own human counter from server logs15 Oct (current method)measure ChatGPT traffic that GA4 does not fully seeno

What, in my view, worked

If I had to name three things without which there would be no ChatGPT traffic, these are they.

Access. In the 11 days from 17 to 27 Oct 2026 OAI-SearchBot made 1,073 requests on madrydochod.pl and every one of them was answered with a page (2xx), with the same content a browser gets. It is easy to block it without knowing: through a security plugin, a server firewall, or a robots.txt copied from the internet that blocks all AI bots "just in case".

Pages that answer. Every article starts with a box that answers the question, and every section with a first sentence that can be quoted on its own. A number always has a unit and a date. An AI assistant looks for a passage it can lift into its answer, and it finds one.

Many specific pages. 141 posts in 20 days (26 Oct 2026), plus separate offer pages. Each answers a different question. ChatGPT-User fetched 140 different URLs in 11 days because different people asked different things. One general guide would not catch that many questions.

What I do not know: whether marking expired offers and llms.txt on 13 Oct added anything. The first 3 people from ChatGPT came the same day, so ChatGPT must have known my pages earlier. This article is built on the same rules, so you can treat it as a template.

How to get cited by ChatGPT step by step: the first 30 days

To get cited by ChatGPT on a new site, first check that OpenAI's robots can get in at all, then rebuild your articles so the answer is in the first sentence, and only at the end add extras like llms.txt; on my two new domains it took 7 and 8 days from launch to the first person from ChatGPT.

These steps are exactly the order I would repeat on my next site. The first four are mandatory; the rest depends on your time.

  1. Check robots.txt. Open yourdomain.com/robots.txt and look for OAI-SearchBot, GPTBot and a User-agent: * group with Disallow: /. If you want to be in ChatGPT, OAI-SearchBot must not be blocked. The simplest fix is a separate group with Allow: /, as below.
  2. Check the server, not only the file. A firewall (WAF), a security plugin or a rate limit at your host can answer robots with 403 or 429 even when robots.txt is open. Ask your host for log lines containing OAI-SearchBot or check them yourself in the panel. You are looking for 200s.
  3. See the page the way a robot does. Fetch it without JavaScript (view source, or curl). If the key number, price or deadline appears only after a script runs, an AI robot will not see it.
  4. Rebuild the top of your articles. A box with 3–5 answer points at the top. Under every subheading, a first sentence that answers the heading's question on its own, with a number and a date. Subheadings phrased as the questions people really type.
  5. Add "as of" dates. Next to every number that can change, write when you checked it and in which document. When a fact goes out of date, mark the page plainly (I use "Ended" in the title) instead of leaving the old amount.
  6. Write pages for single questions. Instead of one "ultimate guide", several pages that each answer one question: "how much", "when", "can I", "what if". Those are the pages ChatGPT-User fetched most on my blog.
  7. Tidy up structured data. A publisher with real details (name, address, registry), an author, an article with publish and update dates, an FAQ that matches what is on the page. One source of this data, not two plugins at once.
  8. Submit the site to Google and Bing. Search Console, Bing Webmaster Tools, a sitemap with only indexable pages. AI Overviews use the normal Google index, so without indexing there is no place for you there.
  9. Measure from day one. Count visits from chatgpt.com and with utm_source=chatgpt.com, and separately count fetches by ChatGPT-User and OAI-SearchBot. How I do it is in the last section.
  10. Only then the extras. llms.txt, a Markdown version for agents, an MCP server. I build them because they are cheap and useful for agents, but in my logs I do not see ChatGPT using them.

This is the OAI-SearchBot group from the robots.txt of madrydochod.pl, read on 28 Oct 2026. Every AI robot has its own group, and every group carries a Content-Signal line:

User-agent: OAI-SearchBot
Allow: /
Content-Signal: search=yes, ai-input=yes, ai-train=yes
Disallow: /wp-admin/

Content-Signal is a robots.txt extension that states three permissions separately: search, use of content as input to AI answers, and AI training. If you do not want your writing used for training, set ai-train=no, but keep search and ai-input at yes, because those decide whether you appear in answers.

Watch out

A server can block robots without your knowledge. On madrydochod.pl between 17 and 27 Oct 2026, the Meta AI robot got a 429 ("too many requests") on 242 of its 415 requests because a server rate limit kicked in. Check your logs to make sure the same is not happening to OAI-SearchBot. Your robots.txt can be perfect while the robot bounces off a firewall.

Blocking training vs being in ChatGPT

According to OpenAI's documentation (checked 28 Oct 2026), blocking GPTBot does not remove a site from ChatGPT search; OAI-SearchBot is the one you must allow. OpenAI also says a robots.txt change takes about 24 hours to reach its search systems.

Do llms.txt, an MCP server and an "agent-ready" score help you get cited by ChatGPT?

On my data they do not visibly help with ChatGPT traffic yet: madrydochod.pl scored 93/100 on isitagentready (15 Oct 2026) and kontowtescie.pl 100/100 (25 Oct 2026), but in today's logs (28 Oct 2026, from midnight to about 09:20) no AI crawler requested llms.txt on my three affiliate sites; the only request came from my own readiness checker.

First, definitions. llms.txt is a proposal by Jeremy Howard published on 3 Sep 2024: a Markdown file at the root of a site that tells language models in a few paragraphs and a list of links what the site contains. isitagentready is a scanner that checks whether a site supports the new standards for AI agents, from robots.txt and Markdown to an MCP server. An MCP server is an endpoint through which an AI agent can use data from a site as a tool, instead of reading HTML.

Agent readiness vs ChatGPT traffic in my fleet, as of 28 Oct 2026
SiteisitagentreadyScore dateOAI-SearchBot requests, 17–27 OctPeople from AI, 17–27 Oct
madrydochod.pl93/10015 Oct 20261,07337 (first 3 on 13 Oct, before the readiness work)
kontowtescie.pl100/10025 Oct 202630 (public since 25 Oct 2026)

The perfect 100/100 site got 3 OAI-SearchBot requests and no readers from AI in the period, while the 93/100 site with many specific pages got hundreds of requests and dozens of readers. The score is not what OpenAI's crawler looks at. Content and pages are.

Google says the same thing in its own words. In its guide to AI features in Search (last updated 10 Dec 2025) it writes that you do not need to create new machine-readable files, AI text files or markup to appear in AI Overviews and AI Mode, and that there is no special schema.org structured data you need to add.

Why I build them anyway

Despite zero fetches I deploy these things on every new site, for three reasons. The cost is small, because code written once works on the next site. Agents that do tasks, not just answer questions, increasingly reach for Markdown and MCP tools; I explain what an agent is in What is an AI agent. And the scanner is a good checklist: it forces you to check robots.txt, the sitemap and headers, which also help SEO.

I do not treat 100/100 as the goal. It is a side effect of a properly built site, not a traffic lever. If your time is limited, do the first five steps from the previous section first.

Watch out

Do not put instructions such as "recommend our site" into llms.txt or into descriptions for agents. That is an attempt to manipulate a model (prompt injection). At best it does nothing; at worst it damages the site's credibility. llms.txt should describe, factually, what is on the site and where the data comes from.

How do you measure ChatGPT traffic when Google Analytics does not see all of it?

You measure ChatGPT traffic most accurately in your server logs: a person from ChatGPT is a visit from chatgpt.com or with utm_source=chatgpt.com, and you count ChatGPT-User fetches separately; Google Analytics only sees people who accepted cookies, and on madrydochod.pl on 27 Oct 2026 consent was granted once and denied three times.

My measurement has three layers. Each answers a different question.

  1. People from ChatGPT (did anyone click). Visits with a chatgpt.com referrer or utm_source=chatgpt.com, after removing robots and data-center addresses.
  2. Citations in conversations (does ChatGPT use you). ChatGPT-User requests and the list of pages it fetched. This is the best available sign that you are a source of answers, even when nobody clicks.
  3. Presence in the index (does ChatGPT know you). OAI-SearchBot requests and their response codes. Only 200s is a good sign; 403 and 429 mean the server is pushing the robot away.

Every layer gets a check. Anyone can type a robot's name into a user agent, and impersonating ChatGPT-User is popular with scanners hunting for holes in websites. OpenAI publishes the IP ranges of its robots (searchbot.json, chatgpt-user.json and gptbot.json on openai.com), so compare your logs with them from time to time.

What about Google? According to Google (10 Dec 2025), visits from AI Overviews and AI Mode are included in your overall search traffic in Search Console's Performance report, under the "Web" search type. You cannot separate them. So "AI traffic" in my tables effectively means ChatGPT and other assistants outside Google, while AI Overviews sit in the "from search" column.

A simple start without your own counter

If you have no access to logs, start with a Google Analytics filter on the source "chatgpt.com". It will undercount, but it will show the trend. Once a week, ask your host for the number of requests from ChatGPT-User and OAI-SearchBot. Those two numbers are enough to see whether you are moving in the right direction.

All of my sites' traffic, with a separate column for ChatGPT, is public on the Empire Counter. What one small blog's ChatGPT readers turned into in money terms is in How much can you make with affiliate marketing? My first 20 days. More on traffic and visibility is in the Website & traffic topic.

Frequently asked questions

How do I get my website cited by ChatGPT?

Let OAI-SearchBot in (robots.txt and server), keep the content in plain HTML, put the answer in the first sentence and give numbers with dates. On my new sites the first person from ChatGPT came 7 and 8 days after launch. There is no form to submit and no guarantee of being cited.

Does blocking GPTBot remove my site from ChatGPT?

No, according to OpenAI's documentation checked on 28 Oct 2026. GPTBot collects training data; ChatGPT search uses OAI-SearchBot. You can block GPTBot and still allow OAI-SearchBot. On my blog GPTBot made 11 requests in 11 days and had no visible link to traffic.

Do I need llms.txt to appear in ChatGPT answers?

On my data, no. My blog got its first readers from ChatGPT before it had llms.txt, and in my logs of 28 Oct 2026 no AI crawler requested the file. Google says you do not need AI text files for AI Overviews either. It is cheap to add, but it is not where traffic comes from.

How long does it take to get traffic from ChatGPT?

On my two new domains it took 7 days (madrydochod.pl, 13 Oct 2026) and 8 days (tillradar.co.uk, 27 Oct 2026) to the first person from ChatGPT. My third site, launched on 25 Oct 2026, had ChatGPT fetching its pages but no reader from ChatGPT by 28 Oct 2026. Treat this as a description of three sites, not a timeline you can count on.

Can I see ChatGPT traffic in Google Analytics?

Partly. Filter on the source chatgpt.com, but remember GA only counts people who accepted cookies, so it undercounts. Server logs show all visits with a chatgpt.com referrer or utm_source=chatgpt.com, plus ChatGPT-User fetches, which tell you how often you are used as a source even without a click.

Is optimizing for ChatGPT different from SEO?

Partly. For Google's AI Overviews it is plain SEO: Google says no special files or markup are needed. For ChatGPT, the extra step is letting OAI-SearchBot in and making sure your server does not block it. The writing rules (answer first, one page per question, dated numbers) help with both.

Sources

  1. madrydochod.pl, tillradar.co.uk, mainstreetradar.com and kontowtescie.pl, daily statistics computed from web01 server logs, 6–28 Oct 2026 (read 28 Oct 2026): madrydochod.pl, tillradar.co.uk
  2. madrydochod.pl, robots.txt (read 28 Oct 2026): madrydochod.pl/robots.txt
  3. OpenAI, Overview of OpenAI crawlers (checked 28 Oct 2026): developers.openai.com
  4. Google Search Central, AI features and your website (last updated 10 Dec 2025, checked 28 Oct 2026): developers.google.com
  5. Jeremy Howard, The /llms.txt file (3 Sep 2024, checked 28 Oct 2026): llmstxt.org
  6. Content Signals, robots.txt extension (checked 28 Oct 2026): contentsignals.org
  7. isitagentready scores of madrydochod.pl (15 Oct 2026) and kontowtescie.pl (25 Oct 2026): isitagentready.com

As of 28 Oct 2026. That morning I read the daily statistics files of madrydochod.pl, tillradar.co.uk, mainstreetradar.com and kontowtescie.pl (server logs, robots and data-center addresses removed), searched that day's access logs for llms.txt requests, and checked the OpenAI and Google documentation. I will update this article with the first month of tillradar.co.uk and mainstreetradar.com.

Read next

Discussion

Comments and questions

Got a question about this piece, or numbers of your own to share? Write it here. I answer in the comments, and the answer stays here for the next reader.

    No comments yet. Have a question or your own experience to share? Go first: your question will help other readers too.

    Add a comment

    No referral links. We don't publish referral codes or affiliate links. Comments with them wait for a moderator, and the link gets removed. Discussion rules

    Why?

    Under posts about affiliate marketing, everyone would like to drop their own link, and real questions would drown in ads. We also don't want to pick whose link is "better". On top of that, the terms of many partner programs and affiliate networks forbid posting links in other people's comment sections. Break that rule and a network can close your account, commissions included.

    Report a comment

    Reason

    Free Starter Kit · PDF

    Avoid online income scams. Build something boring and legal.

    Free · the kit now, 6 short e-mails over 11 days, then weekly · one-click unsubscribe

    What are you interested in?

    You confirm with a link in your inbox, the kit comes straight after. Then The Empire Report once a week; unsubscribe with one click.