Canonicals, Redirects and Migrations for AI Product Sites

Updated 2026-09-06 · guide · SEO, technical, launch

Canonicals, redirects and migrations are the plumbing that tells crawlers and AI engines which version of a page is the real one — and when you get them wrong, you split signals or send search engines into dead ends. A page reachable at several URLs (with and without trailing slash, http vs https, parameters) looks like several pages unless you say otherwise. This matters twice over for AI products: classic crawlers and AI crawlers both follow redirects and read canonicals, and a migration done badly can wipe out months of accumulated citations in one release. This guide covers the mental model, the rules that never go out of date, and the migration checklist that protects your existing rankings and citations.

Why this matters more for AI product sites

URL decisions affect whether launch demand consolidates or fragments. This feature announcement SEO guide pairs with the canonicals and redirects guide for long-term cleanup. Two audiences depend on your URL hygiene:

• Search engines consolidate signals to the canonical URL — split versions mean split ranking weight.
• AI engines remember the URL they cited. If you migrate and don't redirect, the old cited URL breaks, and the AI answer now points at a dead page.

The second point is the new one: in 2026, broken URLs aren't just a 404 problem for users — they're a citation-loss problem. A stale link in an AI answer that leads nowhere is a small but real reputation hit.

The three tools, and what each is for

The rule of thumb: canonicals for duplicates that both exist, redirects for pages that moved. They're not interchangeable, and using the wrong one costs you.

Canonical rules that never change

Redirect rules that never change

• One canonical per page. Point every duplicate variant at the single URL you want indexed.
• Use absolute URLs. A relative canonical is ambiguous and often ignored.
• Self-referencing canonicals are fine — they tell engines "this URL is the canonical for itself," which prevents confusion on www/non-www and protocol variants.
• Don't canonicalize to a page that redirects. Pick the final destination; engines hate chasing loops.
• Make canonical and internal links agree. If internal links point at URL B but canonical says URL A, you've told engines two different things.

Legacy URLs need equivalent destinations; the deprecation and docs-churn SEO guide adds lifecycle rules and dates.

The migration checklist (protect your citations)

• Redirect to the nearest real page. Deleted content should point at its closest living successor, not the home page, unless there's no good match. Deciding between redirect, restore, or a clean 404 for each dead URL is the error-handling and link-rot workflow.
• One hop only. Chain redirects (A → B → C) waste crawl budget and risk getting dropped. Update the intermediate link instead.
• Keep old URLs alive long enough. A 301 is permanent, but changing the old page to a 404 before the crawler has re-crawled the redirect loses the signal.
• Update internal links too. Redirects are a safety net, not a long-term strategy — fix the links in your own site.

Use that checklist inside the site migration SEO playbook, which adds owners, gates, rollback, and daily monitoring.

A site move — new domain, new URL structure, or a CMS change — is the riskiest operation in technical SEO. The checklist:

Common mistakes

1. Map every old URL to its new one before you cut over. No guesswork at migration time.
2. Redirect 1:1 with 301s — every old URL gets a 301 to its direct new counterpart. This is what preserves ranking and citation equity.
3. Update the sitemap and llms.txt to the new URLs on the same release, so both engine types discover the new canonicals fast — see the llms.txt guide for what belongs in it.
4. Update internal links and canonicals to the new structure — don't rely on redirects for your own site.
5. Submit the new sitemap in Search Console and request re-indexing of the money pages.
6. Verify the AI-cited URLs — search your old URLs in the AI engines, and confirm they now resolve (via redirect) to the new pages.
7. Monitor for a dip, don't panic. Expect a temporary ranking dip during re-crawl; it recovers if the redirects are clean.

• Canonicalizing everything to the home page. This tells engines "only the home page matters" and buries your content pages.
• 302-ing a permanent move. Engines may treat it as temporary and keep indexing the old page — the worst of both worlds.
• No redirect map before migrating. Doing a migration without a 1:1 map is how sites lose months of rankings overnight.
• Forgetting llms.txt and sitemap. AI crawlers read these first; stale URLs there mean the new canonicals get discovered last.
• Deleting old URLs without a successor. A 404 on a previously-cited page destroys the citation.

URL, canonical, and redirect rules should be diagnosed in the SEO discovery phase before a redesign or migration scope is signed.

Bottom line

Canonicals, redirects and migrations decide whether crawlers and AI engines see one clean version of your site — and a botched migration can erase accumulated citations in one release. Keep one canonical per page, redirect moved pages with a 1:1 map, and update sitemap, llms.txt and internal links in the same release. Your next step: audit your own URLs for duplicates — check http vs https, trailing slash and parameter variants — and canonicalize any you find today.

FAQ