Do Subdomains Matter Less for AEO and GEO Than for SEO?

Why is the subdomain debate different for AI answers?

In traditional SEO, the argument is about whether a subdomain shares the main site's ranking signals. Google's guide to its ranking systems says site diversity generally treats subdomains as part of the root domain, though sometimes treats them as separate sites when relevant. It also says site-wide signals contribute alongside page-level ones. Practitioners still disagree about how much a blog on blog.example.com benefits from example.com.

AI answers change the question in three ways:

• Citations are page-level. An assistant picks passages that answer the question, and the URL it cites is wherever that passage lives.
• Brand recognition is entity-level. Assistants describe "Example", not "example.com". What connects a subdomain to the brand is its name, its content and the sources that mention it, not the DNS structure.
• Discovery depends on search indexes and crawlers, which handle subdomains as separate hosts with separate rules.

For Google's own AI features, which draw on Google's index, whatever Google's ranking systems do with subdomains carries over. For other assistants, test rather than assume. How these disciplines relate is covered in the difference between SEO, AEO and GEO.

Does hosting a website on a subdomain hurt AI citations?

Not by itself. Technically, www.example.com is a subdomain too. Help centres on help. or support. subdomains and developer documentation on docs. subdomains are standard, and they are exactly the kind of specific, factual pages assistants quote.

What does hurt is what often comes with a subdomain:

• A separate robots.txt. Google's robots.txt specification says the file applies only to the host, protocol and port where it is served. example.com/robots.txt does not cover help.example.com. A subdomain with its own blanket block, or none of the rules you set on the main site, behaves differently.
• A different platform. Hosted help centres, landing page builders and community tools often control robots.txt, sitemaps and rendering themselves, and some do not let you change them.
• Different CDN or firewall settings, which may challenge AI crawlers the main site allows.
• A missing sitemap or no Search Console and Bing Webmaster Tools verification.
• JavaScript-only rendering, common on app-built pages, which most AI crawlers cannot read.
• Thin brand context, such as a help centre with no link back to the main site and a different name in its title.

Does putting marketing content on an app subdomain split authority in AI search?

The authority-splitting argument is weaker for AI than the technical one. Marketing pages on app.example.com usually fail for plainer reasons:

1. The app subdomain is set up for logged-in users. It often has Disallow: / in robots.txt, a noindex header, or redirects anonymous visitors to a login page. Marketing pages there inherit those settings.
2. Pages are rendered by the app's JavaScript framework. Crawlers that fetch raw HTML see an empty shell.
3. The content is not linked from the main site's navigation, so it is rarely discovered or treated as important.
4. Duplicate or conflicting versions exist on the main site, such as an older pricing page on example.com and a newer one on app.example.com.

If you move marketing content to the root domain, redirect every old URL with a server-side 301 to its equivalent. Update internal links and sitemaps, and keep the app subdomain for the app. If you keep it on the subdomain, fix the four problems above: allow crawlers on marketing paths, render the content in HTML, link it from the main site, and remove duplicates. Server-side rendering is covered in how to make JavaScript-rendered pages readable to AI crawlers.

Should a new sub-brand use its own domain or a folder for AI search visibility?

It depends on whether you want assistants to treat the sub-brand as part of the parent or as its own thing.

A sub-brand on its own domain starts almost from zero in AI answers. Assistants learn about brands from how often and how consistently they are mentioned across independent sources. So a new domain needs its own coverage, reviews and listings before it is recommended, whatever the parent's standing. If you choose a separate domain, state the relationship clearly on both sites, such as "Subbrand, from Example". Use it consistently in listings, press releases and profiles, and link the two sites. If you choose a folder, give the sub-brand its own clearly named pages, so assistants can describe it specifically rather than folding it into the parent's generic description. How topical depth compares with domain strength is covered in whether topical authority matters more than domain authority for GEO.

How do ChatGPT and Perplexity choose between country domains, subfolders and subdomains?

Neither OpenAI nor Perplexity documents any preference among URL structures. Both select pages from their own indexes and from search results. OpenAI says ChatGPT search uses third-party search providers, and Perplexity runs its own crawler and index. So the structure matters indirectly. It decides which URLs rank in those indexes for a question in a given language and location, and whether crawlers can reach every version.

What tends to decide which version is cited:

• Language. A German-language page gives the engine a version that matches a question asked in German, whatever the domain.
• Local detail. Prices in local currency, local contact details, regulations and delivery terms mark a page as relevant to a country.
• Access. A version a crawler cannot reach, because of a location-based redirect or a blocked subdomain, cannot be chosen.
• Search ranking. Whatever ranks in the underlying search provider for that locale is more likely to be retrieved.

How ccTLDs, redirects and hreflang fit in is covered in how redirects, canonicals and ccTLDs affect AI search visibility.

What should you check on every subdomain?

1. robots.txt: fetch https://sub.example.com/robots.txt and check the rules for OAI-SearchBot, Claude-SearchBot, PerplexityBot and Googlebot.
2. Indexing: verify the subdomain in Search Console, or use a domain property that covers all subdomains, and in Bing Webmaster Tools.
3. Sitemap: list one in the subdomain's robots.txt.
4. Rendering: fetch a page without JavaScript and confirm the main content is in the HTML.
5. CDN rules: confirm AI crawlers you allow are not challenged.
6. Brand ties: consistent name in titles, a link to the main site, and the same organisation details.
7. Duplicates: make sure the same content is not live on both the subdomain and the main site.
8. Logs: look for AI crawler requests to the subdomain, and for 403 or challenge responses.

What might an illustrative case look like?

Consider an illustrative project management company whose feature pages lived on app.example.com/features, built in the app's front-end framework. Assistants described its features using an old review and a two-year-old page on the main site. The app subdomain's robots.txt disallowed everything, a rule written to keep dashboards out of search, and the feature pages rendered only with JavaScript. The company moved the feature pages to example.com/features as static HTML, with 301 redirects from the old URLs, and linked them from the main navigation. It retired the old duplicate page. Within two months, assistants described current features and cited the new pages.

What do people ask about subdomains and AI search?

Is www or non-www better for AI search?

Neither. Pick one, redirect the other to it, and use it consistently.

Should the blog be on a subdomain or a folder?

Either can work. A folder is simpler, because it shares the main site's robots.txt, sitemap and navigation. A subdomain needs its own setup checked.

Do AI assistants treat help.example.com as the same company?

Usually, if the name, links and content make the tie clear. Help centres are often cited for product questions.

Will moving from a subdomain to a folder improve AI citations?

Only if it fixes a real problem, such as blocked crawling, JavaScript rendering or duplicates. Moving well-working pages risks a temporary dip.