How Do I Protect AI Visibility During a Site Migration?

Protect AI visibility in a migration the same way you protect SEO — plus four AI-specific steps: 301-redirect every URL that AI engines currently cite (not just high-traffic ones), update llms.txt and sitemaps on launch day, re-test access for AI crawlers like GPTBot and PerplexityBot against the new infrastructure, and monitor citations for at least eight weeks.

Before the migration: baseline what AI actually cites

Standard migration prep ranks URLs by organic traffic. AI adds a different priority list: the specific URLs appearing as citations in ChatGPT, Perplexity, Gemini, and Copilot answers. These are often deep pages — glossary entries, statistics pages, old blog posts — that a traffic-sorted redirect map would deprioritize. Pull a citation baseline with citation tracking before you freeze the redirect map, and treat every cited URL as must-redirect.

Export your server logs too. Filter for AI user agents (GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, Bingbot) to see which paths they fetch most. That crawl pattern predicts which URLs will break first in retrieval if redirects fail.

Launch-day checklist, AI edition

1. One-to-one 301s for every mapped URL. Redirect chains longer than two hops get abandoned by some fetchers; flatten them.
2. Update llms.txt. The llmstxt.org format is a curated URL list — a stale one becomes a curated list of dead links.
3. Regenerate and resubmit sitemaps in Google Search Console and Bing Webmaster Tools. Bing matters disproportionately because ChatGPT and Copilot retrieve through its index.
4. Re-verify crawler access on the new stack. New CDN, new WAF rules, new bot management defaults — Cloudflare has blocked AI crawlers by default on new zones since July 2025. Curl your key pages with AI user-agent strings and confirm 200s.
5. Keep robots.txt permissive for the crawlers you previously allowed; a template robots.txt from the new platform can silently reintroduce blocks.

After launch: the 8-week watch

AI engines lag search engines in processing migrations. Expect old URLs to keep appearing in answers for weeks — that is fine as long as they 301 correctly. What you are watching for is citation loss: prompts where you were cited pre-migration and no longer are. Run your prompt set weekly, compare against the baseline, and trace any loss to a specific URL. Nine times out of ten the culprit is a missed redirect, a 403 from new bot rules, or a page whose content was "consolidated" away. The tracking workflow is the difference between a two-day fix and discovering the damage a quarter later.