Updated on 18 November 2026, by João Leite, Founder of murmur.
Content on a subdomain appears in AI answers when each AI’s search bot can read the subdomain. AI models cite page by page, by each page’s URL, and a page at guia.suaempresa.com.br enters the answer like any other, if it answers the searcher’s question. What the subdomain asks for beyond the main site is its own robots.txt, with the search bots allowed, and the link that leads from the main site to it.
murmur, a GEO agency in Brazil that puts brands into the answers of ChatGPT, Claude, Google AIO and other AI agents, found pages on subdomains among the sources of the four AI models it tracks. Every week, it asks the same 90 questions, about GEO, AI search and GEO agencies, to ChatGPT, Claude, Google AIO and Perplexity. In the 10 rounds between 4 September and 2 November 2026, 273 of the 2,437 answers with a source cited a page on a content subdomain, such as blog., guia. or resources., in 75 of the 90 questions.
- Content on a subdomain appears in AI models, page by page. In murmur’s measurement, pages on content subdomains were a source in 273 answers, in 75 of 90 questions, across the four AI models measured.
- The subdomain has its own robots.txt. Anthropic, which makes Claude, asks for the bot rule on every subdomain, and each AI’s search bot needs to be allowed there.
- The main site leads to the subdomain. A link in the menu or on the main pages helps readers and the bot reach it.
- A folder on the site also works. AI models cite the page that answers, and the work is the same at both addresses.
- Mindo’s guide, at guia.mindo.com.br, became a source. In Google AIO, it was a source in 36% of answers in July 2026.
Content on a subdomain appears in AI answers, page by page
Content on a subdomain appears in AI answers because AI models cite the page, not the whole site. When ChatGPT, Claude, Perplexity or Google AIO show a source, they show the link to that page, and the link leads to the address with the subdomain in front. The guide at guia.suaempresa.com.br is read, chosen and cited as a page, for what it answers.
A subdomain is an address that comes before the site name, such as blog.suaempresa.com.br or guia.suaempresa.com.br. A folder is an address that comes after, such as suaempresa.com.br/blog/. For readers, both are pages of the same company. For AI crawlers, the programs with which AI companies read sites, the subdomain is an address with its own rules.
The subdomain appears across the four AI models, and what it asks for extra is configuration, not different content. The page that answers the buyer’s question well is still the page that answers well, at either address.
In murmur’s measurement, pages on content subdomains were a source across the four AI models
In murmur’s measurement, pages on content subdomains were a source in 273 answers, in 75 of the 90 questions, across the four AI models. The question set is the same in every row: the 90 questions about GEO, AI search and GEO agencies, repeated across the 10 rounds between 4 September and 2 November 2026. Each row shows how many answers with a source from that AI model cited a page on a content subdomain.
| AI model | answers with a page on a content subdomain as a source | answers with a source | share of answers with a source |
|---|---|---|---|
| Perplexity | 108 | 787 | 14% |
| Google AIO | 80 | 745 | 11% |
| ChatGPT | 45 | 463 | 10% |
| Claude | 40 | 442 | 9% |
All four AI models cited pages on subdomains, from 9% of Claude’s answers with a source to 14% of Perplexity’s. Added together, that was 273 answers out of 2,437.
Among the subdomains cited are blog.google, from Google, blogs.bing.com, from Microsoft, and blog.hubspot.com, from HubSpot, which sells marketing software. These are companies with a large presence on the internet, and what all three show is that a subdomain address enters the answer like any other. For a company that is not their size, the example is Mindo, a B2B motion design studio from São Paulo, with the guide at guia.mindo.com.br.
Subdomain or folder: what changes for each AI’s bot
For each AI’s bot, the subdomain is a separate address, with its own robots.txt, and the folder follows the main site’s robots.txt. In robots.txt, the owner of each address writes what Google’s and the AI models’ bots may read there. When explaining how to block its bots, Anthropic, which makes Claude, asks for the rule on every subdomain: “Please do this for every subdomain that you wish to opt out from.” The rule that applies to suaempresa.com.br does not apply on its own to guia.suaempresa.com.br.
Each AI model has its own search bot, and that is the one that needs to get into the subdomain. In the documentation for OpenAI’s bots, OAI-SearchBot is the one that feeds ChatGPT search: “OAI-SearchBot is used to surface websites in search results in ChatGPT’s search features.” Anthropic separates Claude-SearchBot, which reads the web to improve answers with search, and writes that blocking it can reduce the site’s visibility in those answers. Perplexity recommends allowing PerplexityBot in robots.txt, the bot that shows and links sites in its results.
Each bot has its own rule. The complete list, with each company’s search bot and training bot, and the case of the subdomain that needs its own file, is in the guide about llms.txt, robots.txt and AI bots.
What content on a subdomain needs in order to appear in AI
Content on a subdomain needs four things in order to appear in AI, and three of them are configuration:
- Each AI’s search bot allowed in the subdomain’s robots.txt.
- A sitemap for the subdomain, with the list of its pages.
- A link from the main site to the subdomain, in the menu or on the pages that cover the subject.
- The page that answers the buyer’s question, with the answer in the first sentence.
Even with the bot allowed, it is worth checking whether it receives the page: the subdomain’s access log shows whether the search bots arrive, and this matters more when it runs on a tool or hosting different from the main site’s. The index, the sitemap and the links between pages are in the technical site checklist for AI models, and the conditions for a page to become a citation are in the guide to how the company is cited by AI models.
When a folder on the main site also works: the same work at another address
A folder on the main site also works, because AI models cite the page that answers, at any of the company’s addresses. The work is the same in both cases: the page that answers the buyer’s question, the search bot allowed, the sitemap and the links between pages. What changes is who handles what.
In the folder, the content goes into the main site’s tool, with the same robots.txt, the same sitemap and the same team that takes care of the site. On the subdomain, the content sits at an address alongside, with its own tool, robots.txt and sitemap, and the main site only gains the link. The choice comes out of a practical conversation: who can publish, in which tool and how often.
murmur publishes on a brand subdomain, and Mindo’s guide became a source with its name in the link
The guide on a brand subdomain becomes a source with the brand name in the link itself. murmur publishes the guides on a subdomain of the client’s brand, which it sets up and maintains without touching the main site, and the content remains the client’s after the contract. The site team carries on with the site, and the guide gets the address with the company name in front.
The proof is Mindo, a B2B motion design studio in São Paulo. Its guide went live at guia.mindo.com.br and was fully online by 22 June 2026. The following month, July, it was a source for Google AIO in 36% of answers; in the answers where the guide’s text appeared, the AI gave credit with a link in 90% of them, and that link led to the address with Mindo’s name in front.
For each client, one person at murmur studies the company and its market, sees in the AI answers who occupies the brand’s place and takes care of the content that changes that answer. It is that same person who tracks the result every week, with the same questions re-run in ChatGPT, Claude, Google AIO and Perplexity. GEO, which is the work of putting the company into the AI answers, comes on top of what the company already does on its site, as shown in the guide on adding GEO to SEO.
Frequently asked questions
Do I need a robots.txt for the subdomain?
Yes: robots.txt applies to the address where it sits, and the subdomain has its own. Anthropic, which makes Claude, asks for the crawlers rule on every subdomain. In the subdomain’s file, let in each AI’s search crawler, such as OAI-SearchBot, from OpenAI, Claude-SearchBot, from Anthropic, and PerplexityBot, from Perplexity. The role of each crawler is explained in the guide to llms.txt and crawlers.
Does Search Console show the subdomain together with the site?
It does, if the site is registered in Search Console as a domain property. According to the help pages of this free Google tool, this type of property covers all subdomains, such as m and www, and more than one protocol, such as http and https. The AI numbers in each report are explained in the guide to Search Console and Bing Webmaster Tools.
Can I move the subdomain guide into a folder later?
You can, as long as each old URL redirects to the new one, so that anyone arriving through the link in an answer finds the page. What changes when the cited page’s URL changes is in the frequently asked questions of the guide to how the company is cited by AI models.
Does the guide on the subdomain remain the company’s if the contract ends?
It does: in murmur’s work, all the content belongs to the client, even after the contract ends. The subdomain belongs to the client’s brand, as guia.mindo.com.br belongs to Mindo, and the pages stay at its URL. murmur’s contract is monthly, with no minimum contract term and no penalty to end it.
Where to start
Before choosing between a folder and a subdomain, check whether any page of your company already appears in the AI answers, at either URL. murmur’s free diagnostic starts from the site URL and puts on screen, with no sign-up and in less than two minutes, the AI answers for anyone searching for what your company sells; the full report arrives later by email. murmur’s sector studies show the brands that the AI models named in each market measured.
Read also
- llms.txt: What It Is, What It Is For, and Whether It Works to Get Cited by AI
- Technical checklist to appear in ChatGPT and other AI models: what to check on your site
- How to get your company cited by AI models: ChatGPT, Claude and Google AIO
- GEO and SEO: which to invest in first, and how to add GEO without cutting SEO
References
- Anthropic, “Does Anthropic crawl data from the web, and how can site owners block the crawler?”, Claude Help Center, undated page. The crawlers rule applies per subdomain (“Please do this for every subdomain that you wish to opt out from.”); Claude-SearchBot reads the web for answers with search. https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler (accessed on 9 November 2026).
- OpenAI, “Overview of OpenAI Crawlers”, undated page. OAI-SearchBot feeds ChatGPT’s search results. https://developers.openai.com/api/docs/bots (accessed on 9 November 2026).
- Perplexity, “Perplexity Crawlers”, undated page. Perplexity recommends allowing PerplexityBot in robots.txt. https://docs.perplexity.ai/docs/resources/perplexity-crawlers (accessed on 6 October 2026).
- Google, Search Console Help, page about the domain property, undated in the section read. The domain property includes all subdomains. https://support.google.com/webmasters/answer/34592?hl=pt-BR (accessed on 9 November 2026).