How Do Redirects, Canonicals, and ccTLDs Affect AI Search Visibility?

What does each one do?

How should redirects be set up for AI crawlers?

Use server-side permanent redirects. Google's redirect documentation orders the options by how reliably Google interprets them. Server-side 301 and 308 redirects come first, then meta refresh, then JavaScript. Google advises using JavaScript redirects only when nothing else is possible. AI crawlers make the gap wider. A 2024 Vercel and MERJ study found that none of the major AI crawlers it tracked rendered JavaScript, so a script-based redirect leaves them on the old page.

Other rules:

• Redirect straight to the final page. Google's site move guidance says Googlebot follows up to 10 hops but advises redirecting to the final destination directly. Other fetchers may give up sooner, and some systems reject redirects altogether.
• Redirect to the closest equivalent page, not the home page. An assistant that cites an old pricing page should land users on current pricing.
• Keep redirects for a long time. Google's guidance says generally at least a year after a move. For AI systems, longer is safer, because old URLs live on in training data and in answers.
• Use 302 only for genuinely temporary moves, such as a short promotion. A 302 left in place for months keeps the old URL as the one Google shows.

The temporary dip that can follow a large migration is covered in why a website migration with 301 redirects can reduce AI visibility temporarily.

Should you redirect URLs that AI assistants invent?

Sometimes. Assistants send users to URLs that do not exist far more often than search engines do. An Ahrefs study of 16 million URLs found they sent visitors to 404 pages 2.87 times as often as Google. Some of these were old URLs, and others were invented URLs that fit a site's pattern. Check your logs for 404s requested by ChatGPT-User, Claude-User, Perplexity-User and similar fetchers, and for 404s with AI referrers. Where a requested URL has an obvious real equivalent, such as /pricing-plans for /pricing, a redirect helps the user. Where it does not, a helpful 404 page with search and key links is better than redirecting everything to the home page.

How do location-based redirects hide pages?

Many international sites redirect visitors to a country version based on their IP address or browser language. Google advises against this. Its multi-regional guidance says automatic redirects between language versions can prevent users and search engines from seeing all versions. It also says IP-based content adaptation is generally unreliable. Google also notes that Googlebot's default crawling IPs appear to be US-based and that it sends no Accept-Language header. So a site that redirects US IP addresses to its US version may show Googlebot only that version.

AI crawlers have the same problem. If a crawler requesting your German page is redirected to the US page, the German content is never read, and German-language answers cite someone else. Instead:

• Give each version its own URL and let all crawlers reach every version.
• Suggest a version with a banner, rather than forcing a redirect.
• Use hreflang to link language and regional versions, each listing itself and the others.

How do canonicals interact with international sites?

The biggest mistake is a cross-country canonical. If example.co.uk/pricing declares example.com/pricing as canonical, Google may fold the UK page into the US one. The UK page then stops being shown or cited, even for UK users. Google's canonicalization documentation says regional variants in the same language are duplicates, and recommends using canonicalization together with hreflang so Google can show the right regional URL. In practice, each regional page should declare itself canonical and list its alternates in hreflang. Its content should also differ where it matters, in prices, currency, contact details and regulations.

For other assistants, which do not document canonical handling, the regional differences on the page are what distinguish the versions. How canonical problems cause wrong citations generally is covered in how indexing, canonical and schema issues prevent LLM citations.

Do ccTLDs help AI visibility in a country?

For Google's AI features, a ccTLD is one of several country signals. Google's multi-regional guidance lists ccTLDs first among the signals it uses to determine a page's target audience. Others include hreflang, server location, and local addresses, phone numbers, language and currency on the page. Links from local sites and Business Profile signals count too. Because AI Overviews and AI Mode draw on Google's index, these signals shape which country's pages are used for local questions.

Google treats some ccTLDs as generic, because people use them that way. Its documentation names .tv and .me as examples. Google added .ai to that list in May 2023. As Google's Gary Illyes put it at the time, it will no longer infer a target country from .ai. A .ai, .io or .co domain is therefore not a country signal in Google.

For other assistants, there is no documented use of ccTLDs. They answer in the user's language, and some use the user's approximate location, so on-page signals probably matter more than the domain. Those signals include content in the local language, local prices and currency, local contact details, and mentions on local sources.

ccTLDs or subfolders?

For most businesses, subfolders are simpler for AI visibility, because the brand stays one clearly identified entity with one domain. If you already have ccTLDs, keep them. Make sure each site states clearly that it belongs to the same company, and use the same brand name. Link the sites to each other with hreflang where content corresponds. Wider structural choices are covered in how site architecture and metadata affect AI search visibility.

What should you check?

1. Crawl your redirects and remove chains, loops and JavaScript-only redirects.
2. Search logs for AI fetcher 404s and redirect those with clear equivalents.
3. Request each country version from a US IP address, with no language header, and confirm you are not redirected.
4. Check canonical tags on every regional page point to themselves.
5. Validate hreflang, making sure each page lists itself and all alternates with valid language and region codes.
6. Test country-specific questions in each assistant, and note which domain or folder is cited.

What might an illustrative case look like?

Consider an illustrative furniture retailer with sites on .com, .co.uk and .de. Its UK and German pages were rarely cited, even for local questions. Testing showed three problems. The .co.uk product pages declared the .com pages canonical. The .de site redirected any visitor with an English browser setting to .com. And discontinued product URLs used JavaScript redirects. The retailer made every regional page self-canonical with hreflang, replaced the forced redirect with a country banner, and moved discontinued products to server-side 301s. It also added local prices and delivery terms to the UK and German pages. Over the next two months, UK and German answers began citing the local sites.

What do people ask about redirects, canonicals and ccTLDs?

Does a 301 redirect pass AI citations to the new URL?

It sends anyone following an old citation to the new page, and Google moves signals over time. Assistants update their cited URLs as they recrawl.

Should I buy a ccTLD for each country to be cited there?

Not for AI visibility alone. Local content and signals matter more than the domain, and one domain keeps your brand's authority together.

Is a .ai domain bad for AI search?

No. Google treats it as generic, like .com, and no assistant documents treating it differently.

Do hreflang tags matter for ChatGPT or Perplexity?

They are not documented as signals there. They matter for Google's AI features, and the separate URLs they describe help every crawler.