How to Protect AI Visibility During a Site Migration

A site migration can protect established Google Search signals by mapping old URLs to relevant new destinations, using redirects, updating links and sitemaps, and monitoring Search Console. Separately, teams should test whether relevant AI crawlers can access the new pages and record what answer systems cite over time. Good Search migration execution does not establish what ChatGPT, Claude, or Perplexity will retrieve or cite after the move.

Here, “AI visibility” means observed brand mentions and linked citations for a defined set of prompts, recorded with the system, configuration, date, and answer. It does not mean that a team can measure what a model understands internally. The distinction matters: migrations have established Search procedures, but available research contains no direct study showing how redesigns affect AI citation loss or recovery.

How should a migration protect both Search and AI access?

Treat a migration as two monitoring tracks: preserve and monitor Google Search signals, while independently checking AI crawler access to the new URLs. Then record answer mentions and citations as observations, not as a presumed consequence of Search performance. Google's move guidance does not predict AI citation recovery.

Search track	AI-crawler and answer-evidence track
Map old URLs to relevant new URLs, deploy redirects, update internal links and submit updated sitemaps.	Check the relevant providers’ published crawler information and site rules; inspect server or CDN logs for requests to new URLs.
Use Google Search Console to monitor the move and Search performance.	Record whether the request reached the intended URL and what response it received; a crawler request alone does not establish that an answer system later cited the page.
Compare Search metrics and URL status over time.	Re-run a fixed prompt set and preserve answers, mentions, citations, dates, and system details.

Google's site-move guidance recommends planning redirects, updating links, submitting sitemaps, and monitoring Search Console during a site move. OpenAI's crawler documentation, Anthropic's crawler guidance, and Perplexity's crawler documentation describe their own crawlers. That documentation identifies provider-specific access considerations; it does not prove that a particular migrated site is crawlable, or that any page will be cited. Keep the Search-focused migration work distinct from AI crawler and answer evidence; the two tracks answer different questions.

What should teams preserve before launch?

Before launch, map each old URL to its closest relevant new destination, prepare permanent redirects, and update internal links, canonicals, sitemaps, and structured data. Retain useful content and supporting evidence where appropriate, and capture a dated Search and AI-answer baseline. The baseline is a proposed monitoring protocol, not a method validated by the migration guidance.

Inventory the URLs in scope and name an owner for each migration workstream. Include important page assets and templates, not only the top-level pages.
Map every old URL to the closest relevant new URL. Avoid sending unrelated pages to a generic destination; flag URLs with no appropriate replacement for an explicit decision.
Define the canonical URL. A canonical URL is the version of a page that the site identifies as the preferred address when similar or duplicate URLs exist. Check that the new page points to the intended destination and that redirects do not conflict with it.
Prepare the release. Plan permanent redirects, update internal links and structured data, and prepare the updated sitemap. Preserve useful page content, titles, and evidence where they remain accurate and relevant.
Capture a dated Search baseline. Record the selected URL set, traffic or Search Console measures, measurement window, and source. These measures describe the chosen sample, not every page or future outcome.
Set a repeatable AI-answer baseline. Select a fixed prompt set. For each run, record the answer system, available model or configuration details, and date. Save the full answer, brand mentions, linked citations, and cited URL so later reviewers can compare them. One prompt is not representative of all buyer questions, and one run may not represent a system’s behavior.
Assign owners and escalation paths. Name who can change redirects, robots rules, CDN or web application firewall (WAF) settings, and page content during launch. Keep changes and incident times in a shared log.

A prompt set makes comparisons more interpretable; it does not make systems static. Changes to prompts, models, retrieval, or page content can affect the answers, so record what changed rather than treating a before-and-after difference as proof of cause.

How do you verify the new URLs after a redesign?

First test that redirects resolve to the intended destinations and that the new pages return the expected responses. Submit and monitor the move in Google Search Console, then inspect access rules and logs for each relevant third-party crawler. A redirect accepted by Google is not evidence that another bot reached the new page.

Use this release checklist for a representative set of important URLs, then expand coverage based on risk:

Redirect chains and loops: check that old URLs reach the intended final destination without unnecessary hops or redirect loops.
Destination responses: confirm that the final URL returns the intended HTTP status and serves the expected page, rather than an error, soft error, or unrelated page.
Robots rules and page directives: review robots.txt, page-level directives, and relevant access controls for unintended blocks. Provider documentation from OpenAI, Anthropic, and Perplexity can help identify crawler names and stated purposes, but current vendor documentation should be rechecked at release time.
CDN and WAF behavior: inspect configuration and logs for blocks, challenges, rate limits, or differences in treatment that may prevent a crawler from retrieving the page.
Blocked resources and rendering: check whether required content and resources are available to the intended clients. For JavaScript-dependent pages, compare what the raw response contains with what the browser renders; a visual browser check alone does not establish crawler access.
Canonical destinations: verify that canonical references point to the intended new URLs and agree with the migration map.
Internal links and sitemap: find links that still point to old addresses, update them, and ensure the submitted sitemap contains the intended new URLs.
Evidence source: distinguish configuration review and vendor documentation from observed server, CDN, or WAF logs. Logs can show a request and response recorded by that infrastructure; they do not show that an answer system used the page.

Compare initial HTML with rendered output, then verify crawler permissions and delivery behavior separately. A change in one does not establish a change in the other.

When should you expect SEO or AI citation recovery?

Google's site-move guidance says that, for medium-sized sites, it may take a few weeks or more for new URLs to gradually start appearing instead of old ones in Search; larger sites may take longer, depending in part on URL volume and server speed. This is guidance about Google Search processing, not a timetable for AI citations. Available research has not established how redesigns affect citation loss and recovery.

Track milestone dates rather than promising a recovery deadline. A useful event log separates:

Event to record	What the date helps establish
Launch	When the new site became available to users and crawlers.
Redirect deployment	When the planned old-to-new URL routing went live.
Crawler fetches	When relevant requests to new URLs were observed in available logs, with status and destination.
Search URL replacement	When Search reporting indicates that new URLs are appearing in place of old ones; record the source and measurement method.
Repeated citation observations	The dates and conditions of saved prompt runs, including cited URLs and answer-system details.

These events are not interchangeable. A crawler fetch does not establish citation, and a citation observed in one answer does not establish stable visibility. Report actual observations and their dates; do not infer an AI recovery window from Google’s Search estimate.

What should a post-migration visibility report include?

Report Google Search signals alongside per-provider crawl evidence and dated AI-answer observations. Give each metric its own denominator (the set of items counted), source, and limitations. Compare the same predefined queries over repeated runs, and note relevant content or model changes. Keep mentions, linked citations, referral sessions, and qualified outcomes separate rather than treating them as equivalent measures.

Use a reporting structure such as this, filling it with observations from the migration rather than illustrative or invented results:

Metric	Baseline window	Post-launch window	Source	Denominator	Limitations
Google Search visibility	Record the dates and selected URL/query set.	Record the comparable dates and set.	Identify the Search reporting source.	State the URLs, queries, or other measurement unit included.	Note coverage, processing changes, and other constraints.
AI crawler requests and responses	Record the log period and relevant URLs.	Record the comparable period and URLs.	Identify the server, CDN, or WAF log source.	State which requests, URLs, or providers were included.	Logs show recorded requests, not downstream use in answers.
Brand mentions and linked citations	Record prompt set, system details, dates, answers, and citations.	Repeat and preserve the same fields.	Identify the saved answer observations.	State runs and prompts observed; do not imply they represent all queries.	Answers can vary; model, retrieval, prompt, and content changes limit comparison.
Referral sessions and qualified outcomes	Include only if the analytics and outcome definitions are available.	Use comparable definitions and windows where possible.	Identify the analytics or CRM source.	State sessions or outcomes included and the attribution rule.	Attribution and small samples may limit conclusions; do not infer causation from timing alone.

A pre/post comparison can identify changes worth investigating; by itself, it cannot show that the migration caused them. For a repeatable approach to prompts, answer capture, and denominators, see the AI visibility measurement methodology. The proposed baseline can also help teams discuss how buyers find the business and whether that discovery leads to qualified demand, without combining distinct signals into a single “visibility” score.

What remains unknown about redesigns and AI citations?

Available research has not established how redesigns affect LLM citation loss and recovery. The practical checks in this guide reduce uncertainty about routing, access, and measurement; they do not guarantee citation retention or recovery.

A useful next step for the field would be a multi-site case series—a study that tracks the same measures across several websites—that records migration type, redirect coverage, relevant bot logs, Search visibility, and repeated prompt observations over time. It should document the systems and measurement conditions and distinguish observed associations from causal effects. Until that evidence exists, teams should treat AI-answer outcomes as an independent monitoring track, not as a predictable by-product of a successful Search migration.

Frequently Asked Questions

How can I migrate my website without losing my SEO?

No migration checklist can guarantee unchanged rankings or traffic. Map old URLs to relevant new destinations, test redirects, update links and sitemaps, and monitor the move in Google Search Console. Track AI crawler access and answer citations separately; a before-and-after change alone does not establish its cause.

How Nortav Can Help

For teams planning a site move, the review should keep URL changes, Search monitoring, third-party crawler access, and observed AI-answer outcomes as distinct checks. Priorities should follow the actual release plan and evidence, not an assumed citation-recovery timeline.

Our approach starts with understanding your current search visibility, identifying gaps against competitors, and developing a strategy focused on driving qualified traffic, generating leads, and delivering measurable business value.

Book a Discovery Call →