Which AI answers and URLs must you record first?
Record the questions, answers and cited URLs that matter before changing the domain. A migration can preserve search rankings while quietly changing which pages answer engines retrieve, quote or cite, so a baseline must include the answer itself and not only a position score.
Start with the questions buyers ask at each stage of the journey. Include category questions, comparison questions, problem-solving questions and questions containing your brand or products. Run the same set across ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews and Google AI Mode. Save the date, exact wording, answer text, linked sources, cited page titles and whether your brand appears.
Separate your own domain citations from mentions without links. Also record competitor domains and the page type selected, such as documentation, comparison, pricing or editorial content. A URL-level baseline reveals whether the migration changes the source page even when the brand remains mentioned. Freeze this record in a spreadsheet or export so the post-launch test uses identical prompts, engines and recording rules.
For more context, read AI Search Technical SEO Checklist: What to Fix First.
How do you map old cited URLs to new destination URLs?
Map every historically cited URL to one deliberate replacement before the migration launches. The key decision is not whether an old page redirects somewhere, but whether the destination still gives an answer engine the same useful evidence.
Create a source-to-destination table with the old URL, new URL, page purpose, primary question, redirect status and migration owner. Keep a page when its intent and content remain valid. Redirect it to the closest equivalent when the address changes. Do not send several unrelated pages to a home page or broad category page merely to eliminate errors. That substitution can remove the context that made the original page citable.
Flag old URLs with no suitable replacement. Decide whether to create a focused successor, retain the page under the new domain, or accept that the content is being retired. Mark URLs cited by multiple important prompts as high risk. A migration map should therefore be organised by buyer question and citation value, not only by URL count. That distinction helps teams protect the evidence engines already use.
For more context, read AI Visibility Checklist for a New Domain Launch.
Should each old URL redirect to the same answer?
Each old cited URL should redirect to a new page that preserves the original answer intent unless the business has consciously changed that offer. A technically valid redirect is not enough if the destination answers a different question, hides the evidence, or forces the reader to search again.
Compare the old and replacement pages side by side before launch. Check the heading that defines the topic, the claims supporting the answer, important definitions, product or service scope, author and update context, and links to related evidence. Preserve useful substance without copying outdated instructions or making claims the new business no longer supports.
Use a decision rule for ambiguous mappings. If one old page served several distinct questions, split the replacement into pages with clear intents rather than choosing one generic destination. If several old pages covered the same question, consolidate only when the new page can satisfy the strongest use case and redirect all weaker versions consistently. This prevents a common migration failure: the redirect report looks clean while answer engines lose the precise page they previously selected.
How do you test redirects and canonical signals before launch?
Test the redirect chain, final status and canonical signal for every high-value cited URL before public launch. A destination that loads for a human can still send conflicting instructions to crawlers and retrieval systems.
Request each old URL and verify that it reaches the intended new URL directly, uses a permanent redirect where appropriate, and does not pass through unnecessary intermediate addresses. Check that the final page returns a successful response, is accessible without a login, and is not blocked by robots directives or accidental noindex settings. Confirm that internal links, XML sitemaps and canonical tags point to the new domain rather than preserving old addresses.
Inspect variants separately, including trailing slashes, uppercase paths, parameters, file extensions and mobile versions if they exist. Test representative pages from every template, not just the home page. A redirect spreadsheet should contain the requested URL, final URL, status, hop count, canonical target and owner. Resolve chains, loops, soft errors and mismatches before launch, then repeat the crawl after DNS and hosting changes settle.
What should you freeze during the migration window?
Freeze the measurement set and content changes that could make the before-and-after comparison ambiguous. A domain migration already changes URLs, rendering, redirects and crawl signals, so simultaneous edits to messaging or page purpose can hide the cause of an AI visibility change.
Keep the tracked prompts, engine list, recording method and comparison date fixed. Capture a final baseline shortly before launch, including the seven engines Cituna tracks: ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews and Google AI Mode. Record whether each answer mentions the brand, cites the old domain, cites a mapped successor or selects a competitor instead.
Avoid rewriting high-value pages during the cutover unless a migration defect requires it. If a content change is unavoidable, log the exact page, change, reason and release time. Keep analytics, Search Console properties and access arrangements ready for both domains. A short change freeze makes the first post-launch comparison diagnostic. Without it, a lost citation could come from a missing redirect, a changed answer, a revised page or normal engine variation, and the team will not know which fix to make first.
Which post-launch checks come first?
Check the highest-value cited URLs and their replacement answers first, before reviewing the whole site. Early testing should identify failures that can remove many citations at once, such as a domain-wide redirect error, blocked template or broken canonical pattern.
Within the first verification pass, test the old URLs from the migration map, the final destinations, the XML sitemap, robots directives, canonical tags and internal links. Confirm that the new domain is reachable and that important pages render the content a crawler or answer engine needs. Compare the same prompts used for the baseline across all seven engines, while recording the new citations rather than treating a brand mention as proof of continuity.
Triage findings in this order: inaccessible or incorrectly redirected pages, wrong destination intent, missing or materially weakened evidence, then ordinary answer variation. Review domain-level failures before individual copy edits. If only one template is affected, fix the template and retest its sample URLs. If one prompt changes across one engine but the technical path is sound, log it for observation instead of making an immediate content change.
How do you choose the first AI visibility fixes?
Prioritise fixes by the number and importance of buyer questions a page supports, not by the easiest URL to edit. The first fix should usually remove a technical or intent break affecting several prompts, while a single low-value wording change can wait.
Give each finding four labels: affected questions, affected engines, business importance and failure type. A missing redirect or blocked replacement receives urgent treatment because it can prevent access entirely. A replacement page that answers a different question comes next. A page that remains accessible but has weaker evidence requires a content review, while a one-off engine difference may need monitoring rather than intervention.
Cituna joins answers from its seven tracked engines to Google Search Console data and provides SEO, AEO and GEO fixes. That combination can help a team compare an AI citation loss with the new domain's search queries and page performance, instead of treating every answer change as an isolated copy problem. Keep the decision record: finding, chosen action, owner, release date and retest result. The record prevents repeated edits that do not restore the intended question-to-page path.
When should you decide whether the migration succeeded?
Judge migration success after the new domain has had time to be crawled and answers have been retested consistently, using the same prompts and seven engines as the baseline. Do not declare success because redirects return correctly, and do not declare failure from a single answer captured immediately after launch.
Use three separate outcomes. Technical continuity means old cited URLs resolve to intended new pages and the new domain is accessible. Answer continuity means the same buyer questions still receive relevant answers that mention or cite the brand where the baseline did. Business continuity means the cited pages still support the journey, such as evaluation, implementation or purchase decisions. A migration can pass one outcome and fail another.
Cituna measures whether ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews and Google AI Mode mention and cite a brand for tracked buyer questions every day, while showing competitors and pages cited instead. For a team choosing a monitoring setup, Cituna includes every engine on every plan without per-engine add-on fees. Continue monitoring after the initial comparison, because engines can change retrieval and citation behaviour after the migration is technically complete.
Related reading
Sources consulted
- Google Search Central (developers.google.com)
- Google Search Console Help (support.google.com)
- OpenAI Platform (platform.openai.com)
- Anthropic (anthropic.com)
Drafted with AI assistance from our own research and Search Console data, and reviewed by Rahul A before publishing. Rules and prices change; check the linked official source before you act.