By Mehdi HassanDoes a website migration affect AI Overviews and ChatGPT citations?
Yes, a website migration can affect AI Overview and ChatGPT citations, because answer engines cite specific URLs and specific passages. If those URLs break, the passages change, or the new site is harder for crawlers to read, the engines may cite someone else instead.
Most migration advice only covers classic rankings. Our website migration SEO checklist handles those. This guide adds the layer on top: keeping the answers that Google AI Overviews, ChatGPT and Perplexity already pull from your site. Few migration guides cover it, yet for B2B and SaaS sites these citations increasingly shape who gets considered.
How do you find which pages answer engines cite today?
Find the pages answer engines cite by asking them the questions your buyers ask and recording which of your URLs appear. Do this before migration, because once the site changes you can't recreate the baseline.
List your target questions
Pull them from your top Search Console queries, sales calls and the H2s on your best pages. Phrase them the way a buyer would.
Ask each engine
Run the questions through Google, ChatGPT with search, and Perplexity. Note every citation of your domain and the URL it points to.
Capture the passage
For each citation, find the sentence or table on your page the answer draws from. Save it with the page's raw HTML.
Check referral traffic
In analytics, filter referrals from AI assistants to see which landing pages they send visitors to.
Mark the pages
Flag these URLs in your redirect map and content audit as protected: keep URL, keep passage, keep schema.
Results vary by session and over time, so run each question more than once and keep dates. Our answer engine optimization guide explains how these engines choose sources in more depth.
What makes a passage citable, and how do you keep it?
A citable passage answers a question directly, makes sense on its own, and sits under a heading that matches the question. Answer engines lift sentences like these because they can quote them without extra context.
Redesigns often break this without anyone noticing. Copy gets shortened for layout, direct answers get turned into teaser lines, and tables get replaced with icons. Keep the cited passages word for word unless you're making them clearer. If you're rewriting copy during the move, our guide to a content audit before migration shows how to mark which passages to protect.
Element | Keep | Risky change |
|---|---|---|
Direct answer under an H2 | One or two sentences that answer the heading | Replaced by a tagline or hidden in a tab |
Comparison or price table | Real HTML table with clear headers | Turned into an image or a carousel |
Question-style headings | Headings that match buyer questions | Rewritten as clever slogans |
FAQ section | Visible questions and answers | Removed or loaded only on click |
Specific facts and prices | Stated plainly in text | Moved to a PDF or a chat widget |
How should URLs and redirects handle cited pages?
Keep cited pages at the same URL whenever you can, and redirect them precisely when you can't. A redirect passes signals to the new page, but an unchanged URL removes any question about which page the engine should cite.
If a cited URL has to change, redirect it to the page that contains the same passage, never to a category or the homepage. Avoid merging a cited page into a longer one unless the passage survives intact near the top. Build these rows first in your 301 redirect map, and test them with the platform guides for Framer redirects or redirects in Next.js.
What technical settings decide whether AI crawlers can read the new site?
Three settings decide whether AI crawlers can read the new site: robots.txt rules, CDN or firewall bot settings, and how the page renders. Any of them can quietly block answer engines after a platform change.
- robots.txt: check rules for the crawlers you want, such as Googlebot, GPTBot, OAI-SearchBot and PerplexityBot. New platforms ship their own defaults.
- CDN bot controls: some CDNs and hosts, Cloudflare included, offer settings that block AI crawlers. Make a deliberate choice rather than inheriting a default.
- Server rendering: main copy, headings, tables and FAQs should be in the initial HTML.
- Schema: carry over Organization, Article and FAQ markup with updated URLs. Our guide to schema markup and AI citations covers what the evidence says about its effect.
Hypothetical example: a SaaS pricing page that lost its citation
For example, suppose an AI assistant cites a SaaS company's pricing page when buyers ask what the tool costs. The page has a plain-text sentence stating the plans and a table underneath. During a move to a new builder, the designer replaces the table with animated cards that load from a script and swaps the sentence for a headline.
The URL stays the same and the redirect map is perfect, yet the assistant starts citing a review site instead, because the new HTML no longer contains the answer. The fix is to restore the sentence and render the plan details as real HTML. Nothing about the redirects needed to change.
How do you check AI citations after launch?
Check AI citations after launch by re-running your saved question list at weeks two, four and eight, and comparing which URLs appear. Pair that with AI referral traffic in analytics and normal Search Console monitoring.
If a citation disappears, check the page in order: does the URL resolve, is the passage in the raw HTML, is the crawler allowed, is the schema intact? Most losses trace back to one of those. For the wider picture, our guide on generative engine optimization summarizes what research says about how these engines pick sources, and our article on AI-assisted website migration covers what coding agents tend to drop.
