How to show up in ChatGPT. And in Copilot, Gemini, Perplexity and Claude, which read the web through different doors.
A site can rank first on Google and still be missing from ChatGPT’s answers, because each assistant draws on its own index. Here is which door to open for each one, and a 10-point checklist in Word and PDF.
To show up in ChatGPT, a site has to let in OAI-SearchBot, the crawler OpenAI uses to find pages for search, and be in the search engines’ indexes. Blocking GPTBot, which is used for training, does not take the site out of search.
The other assistants read different indexes. Microsoft documents that Copilot uses the Bing search service; Google says that for AI Overviews and AI Mode a page only needs to be indexed and eligible for a snippet; Perplexity and Anthropic run their own search crawlers.
Once in the index, a page gets cited when its facts can be checked. In the GEO study presented at KDD 2024, sources, quotations and statistics raised visibility in answers by up to 40%.
The 10-point checklist can be downloaded below, in Word and PDF, with no email required.
Written by: Mattia Esposito, AI and marketing consultant, founder of Itria AI. From Puglia, based in Turin since 2023, he works remotely in Italy and abroad. LinkedIn profile.
The 10-point checklist, to download
One sheet with the table of assistants, the ten checks with where to look, and the robots.txt lines, in Word to fill in and in PDF to print. You need access to your site’s admin panel, to your CDN if you use one, to Search Console and to Bing Webmaster Tools. It takes an hour.

| File | Contents | Link |
|---|---|---|
| Wordto fill in, in Italian | The table of assistants, the ten checks with where to look, the robots.txt lines. | |
| PDFto print, in Italian | The same checklist, to tick off with a pen. |
| N. | What to check | Where to look |
|---|---|---|
| 1 | robots.txt doesn’t block OAI-SearchBot, bingbot, Googlebot, PerplexityBot or Claude-SearchBot. | yoursite.com/robots.txt |
| 2 | The CDN or firewall doesn’t stop search crawlers before robots.txt is read. | CDN or hosting dashboard |
| 3 | Your important pages are in Google’s index. | Search Console, Pages report |
| 4 | The site is in Bing’s index, which feeds Copilot. | Bing Webmaster Tools |
| 5 | New or changed pages are flagged straight away: an up-to-date sitemap and IndexNow. | Bing Webmaster Tools, Search Console |
| 6 | Copy, prices and contact details are in the HTML, readable without JavaScript. | Page source (Ctrl+U) |
| 7 | Every important page answers one question in its first lines, in the words people search with. | Your main pages |
| 8 | Facts can be checked: figures with their source, dates, organisations and documents named in full, attributed quotations. | Your main pages |
| 9 | Who is behind the site is stated clearly and consistently everywhere: name, address, VAT number, people, profiles. | About page, footer, external profiles |
| 10 | Measure it: visits from chatgpt.com, citations in Bing, Search Console. | Analytics, Bing, Search Console |
On 7 August a friend sent me what ChatGPT was saying about me: «I can’t find a Mattia Esposito connected to itria.io as founder». I had already written the file for assistants and the structured data, the things everyone usually recommends. As far as AI was concerned, I didn’t exist.
It mattered to me because at the start of September a food producer told me an order had reached him through AI: someone was looking for a supplier and an assistant had named him. From that day the question for me became a different one: when a buyer asks, who gets named?
From 11 October to 7 November 2026, itria.io pages appeared 3,818 times in Google’s and Microsoft’s AI answers. In between there are more than a hundred pages written and plenty of mistakes, some of them told below. The part that applies to any site is in the public documents of the companies that run the assistants, and this page puts it in order.
Which index each assistant reads
When it searches the web, each assistant draws on a different source, and a page missing from that source cannot be cited. The table brings together the documentation of OpenAI, Microsoft, Google, Perplexity and Anthropic, read on 10 November 2026. The last column is the name to look for in robots.txt.
| Assistant | Where it gets the web from | Search crawler |
|---|---|---|
| ChatGPT | OpenAI’s own crawler and, at times, third-party search providers that OpenAI doesn’t name. | OAI-SearchBot |
| Copilot | The Bing search service, according to Microsoft’s documentation. | bingbot |
| AI Overviews and AI Mode | Google Search’s index, under the same rules. | Googlebot |
| Gemini | Google Search’s index, to ground its answers. | Googlebot |
| Perplexity | Its own index. | PerplexityBot |
| Claude | Its own search crawler, plus outside providers including Brave Search. | Claude-SearchBot |
Hence the first practical rule: get indexed by Google and Bing together. Search Console and Bing Webmaster Tools are free, and Bing lets you import a site from Search Console without verifying it again. Ranking first on Google isn’t enough: in an Ahrefs study of 15,000 prompts (August 2025, data from an SEO tool vendor) only 12% of the pages cited by assistants were in Google’s top ten for the same query.
How to get found by ChatGPT: OAI-SearchBot
OpenAI runs three crawlers with three jobs, and for search what matters is OAI-SearchBot. GPTBot collects pages that may be used to train models; ChatGPT-User opens a page when a user asks it to. OpenAI’s documentation is blunt about the first: «Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers».
The settings are independent. A site can block GPTBot, if it doesn’t want its text used for training, and let OAI-SearchBot in to stay in the answers. OpenAI says it takes about 24 hours for a robots.txt change to reach its search systems.
User-agent: OAI-SearchBot
Allow: /
User-agent: GPTBot
Disallow: /
A robots.txt with no rules for these names already lets them in: the lines are needed when there is a blanket block higher up. The links ChatGPT shows in its answers usually carry utm_source=chatgpt.com, so the visits that come from there show up in your site analytics. OpenAI used to say so in its publishers’ FAQ, since withdrawn and still readable in the archived copy of 4 October 2026.
The CDN, the door before robots.txt
Many sites sit behind a CDN such as Cloudflare, which has its own rules for AI crawlers and applies them before robots.txt is read. In its documentation updated on 2 November 2026, Cloudflare splits access policies into three behaviours, Search, Agents and Training, set in the security settings.
For new domains all three start at «Allow». A site configured earlier, or by a cautious provider, may have search blocked without anyone remembering. The check is quick: open the dashboard, find the AI crawler policies and set Search to «Allow».
How to show up in Copilot: go through Bing
Microsoft says that Copilot, with web search on, gets its information from the Bing search service: it writes a short query, sends it to Bing and composes the answer from what comes back. So to show up in Copilot, the site has to be in Bing’s index.
You register in Bing Webmaster Tools, submit the sitemap and flag new pages with IndexNow, the protocol created by Microsoft Bing and Yandex in 2021 that lets a site tell participating search engines when a page changes: up to 10,000 URLs in a single call.
Since 10 February 2026 Bing Webmaster Tools has also had the AI Performance report, in public preview: it counts how often your pages were cited as a source in Copilot answers, in Bing’s AI summaries and with some partners. It measures citations, not position or clicks.
AI Overviews, AI Mode and Gemini: Google’s rules apply
For Google, a page can appear in AI Overviews and AI Mode if it is indexed and eligible to show a snippet in Search. Google Search Central puts it in one line: «There are no additional technical requirements.» Meeting the requirements doesn’t guarantee you will appear.
The controls are Search’s own: nosnippet, data-nosnippet and max-snippet limit the text shown, noindex removes the page. Google-Extended, the name used to deny the use of your text for training Gemini and grounding its answers, does not affect inclusion in Search and is not a ranking signal, according to Google.
How to show up in Perplexity and Claude
Perplexity builds its own index with PerplexityBot, which it doesn’t use to train models: to show up, just don’t block it. There is also Perplexity-User, which opens a page when a user asks and which, Perplexity says, generally ignores robots.txt.
Anthropic runs three crawlers: ClaudeBot for training, Claude-User for user requests and Claude-SearchBot for the quality of search results. Blocking Claude-SearchBot, Anthropic says, stops indexing for search and can reduce visibility.
On 25 September 2026 Brave Search appeared in Anthropic’s list of providers, with the web search function: what Brave’s index knows about a site is part of what Claude can find. Brave has a public form to ask for a URL to be recrawled.
What gets a page cited, according to research
The most cited study on the subject, «GEO: Generative Engine Optimization» by Aggarwal and others (KDD 2024), rewrote pages in nine different ways and measured how much space they won in a generative engine’s answers. From a baseline of 19.3, the three best techniques were quotations (27.2), statistics (25.2) and cited sources (24.6).
| Technique (GEO study, table 1) | What you do to the page | Visibility |
|---|---|---|
| None | The page as it is. | 19.3 |
| Quotations | Sentences attributed to a person or an organisation. | 27.2 |
| Statistics | Figures instead of general claims. | 25.2 |
| More fluent writing | The same text, written better. | 24.7 |
| Cited sources | A reference to the source of every fact. | 24.6 |
| Unique words | A more varied vocabulary. | 20.5 |
| Keyword stuffing | SEO’s oldest technique. | 17.7 |
Repeating keywords did worse than doing nothing. And for a site in fifth place in the results, citing sources brought 115% more visibility, while the top result lost 30% on average. The study uses an engine built for the test and warns that techniques may change over time: read it as a direction, not a promise.
On my own site I saw it before I read the study. The pages I had sweated over most, in fifth and sixth place, were getting zero clicks, and their AI boxes cited the body that writes the rule and a courier. This is how I put it on 21 September: «AI cites the body that issues the rule or the company doing the work. Never the people explaining it.» From then on I changed the way I write: the full post.
Then there is a technical requirement people forget. Assistants’ crawlers read the HTML as it arrives and don’t run JavaScript: Vercel measured it across about 1.3 billion requests in a month. Copy, prices and contact details that only appear after a script don’t exist for an assistant.
The email address that wasn’t there. Cloudflare’s obfuscation removed the address from the HTML and put it back with a script: to an assistant, the contact page had no contact details. The case, and the same flaw on 14 out of 121 producers’ sites, is in the piece on the email address AI assistants can’t see.
From one page to five. On the morning of 25 September 2026, for Claude’s web search, itria.io had a single page; by the evening, after asking Brave to recrawl seven, it had five. Before and after, with the limits of the data, in Brave and Claude.
The pages that didn’t exist. Without a real 404 page, five URLs that had never existed were getting impressions in first position: the case of the phantom URLs.
The result. Between 11 October and 7 November 2026 Bing’s AI Performance report counted 268 citations of itria.io pages in Copilot, and Search Console 3,550 impressions in AI Overviews and AI Mode.
What you don’t need to show up in ChatGPT
The best-selling shortcuts don’t appear in any official document read for this page. No company sells a paid form to get into the answers; llms.txt is a 2024 proposal you can write, but Google, in its May 2026 guide to AI features, says Search ignores it: creating one neither helps nor hurts. OpenAI, Anthropic and Perplexity, in the documents read, don’t say they use it to choose sources.
In the same guide Google names AEO and GEO, answer engine optimization and generative engine optimization, and warns that many of the tricks suggested online are neither effective nor supported by how Search actually works. To appear in its generative answers, the basics of SEO apply.
What counts is the dull stuff: open doors, the index, a readable page, the fact with its source and the name of the person or company behind it, written the same everywhere. How to build it for a real company, with the right questions to own, is on the page about presence on search engines and AI assistants.
Questions and answers
How do I show up in ChatGPT?
By letting in OAI-SearchBot, the crawler OpenAI uses to find pages for ChatGPT search, and by getting the site indexed by Google and Bing. Then content matters: pages that answer a precise question, with sources, figures and names written out in full.
There is no form to register a site with ChatGPT search. Since 10 November 2026, though, a business can claim its listing in Business Hub: it does not lift you in the answers, but it keeps your details right and shows mentions and clicks.
If I block GPTBot, do I disappear from ChatGPT?
No. GPTBot collects pages to train models, while ChatGPT search uses OAI-SearchBot. OpenAI keeps the two settings separate: you can opt out of training and stay in search. After a change to robots.txt it takes about 24 hours.
How do I show up in Copilot?
Microsoft documents that Copilot, when it uses the web, gets its information from the Bing search service. So the site has to be in the Bing index: you register it in Bing Webmaster Tools, which can import it from Search Console, and flag new pages with IndexNow.
In the same dashboard, the AI Performance report counts how often your pages have been cited.
How do I show up in Perplexity and Claude?
Perplexity builds its own index with PerplexityBot: just don’t block it. Anthropic uses Claude-SearchBot to index pages for Claude’s search; blocking it, Anthropic says, can reduce visibility. Brave Search is also listed among Anthropic’s web search providers.
How can I tell if ChatGPT cites me?
In your site analytics: the links ChatGPT shows in its answers usually carry utm_source=chatgpt.com, as OpenAI used to say in its publishers’ FAQ, since withdrawn; visits from there can be counted. For Copilot there is the AI Performance report in Bing Webmaster Tools; for Google, Search Console.
Does ChatGPT Atlas still exist?
No. On 9 July 2026 OpenAI announced that Atlas would stop working on 9 August 2026, and the launch announcement now carries a notice that the browser has been deprecated. Browsing with the agent has moved into the ChatGPT desktop app, the browser extension and the cloud browser of the paid plans.
For a website the basic rule does not change: OAI-SearchBot must be able to get in, and buttons, menus and forms with correct ARIA labels help the agent use them, as OpenAI advised for Atlas.
Notes on sources
- OpenAI, Overview of OpenAI Crawlers: OAI-SearchBot, GPTBot, ChatGPT-User, robots.txt timing; source of the quotation. Read on 10 November 2026.
- Google Search Central, AI features and your website (updated 10 December 2025) and Google common crawlers, content licensed under Creative Commons Attribution 4.0.
- Microsoft Learn, Data, privacy, and security for web search in Microsoft Copilot, updated 4 November 2026; Bing Webmaster Blog, AI Performance, 10 February 2026; IndexNow documentation.
- Perplexity, Perplexity Crawlers; Anthropic, crawler page, updated 30 October 2026, and list of providers, read on 25 September and again on 10 November 2026.
- Google Search Central, Optimizing your website for generative AI features on Google Search, 15 May 2026, updated 10 July 2026, CC BY 4.0: llms.txt and the position on AEO and GEO.
- Cloudflare, documentation on AI crawlers, updated 2 November 2026.
- Aggarwal and others, GEO: Generative Engine Optimization, KDD 2024, CC BY 4.0 licence: table 1 (position-adjusted word count) and table 2.
- Vercel, The rise of the AI crawler. Ahrefs, AI search overlap, 11 August 2025: a study by an SEO tool vendor, 15,000 prompts.
- The itria.io figures come from Search Console (generative AI report) and Bing Webmaster Tools, window 11 October to 7 November 2026.
Opening the doors is the first step. Getting chosen over your competitors is the work that follows. Positioning is built starting from the market.
Mattia Esposito, who wrote this guide, also works with Itria on positioning: in-depth market analysis to find where competitors are exposed and take that ground. He did it on his own site first. In August, asked who had founded itria.io, ChatGPT couldn’t find him. From 11 October to 7 November 2026 itria.io pages appeared 13,803 times in Google and Bing results and 3,818 times in AI answers: Google’s AI Overviews and AI Mode, and Microsoft Copilot.