Site Migration AI Citations: Domain Move Checklist

Site migration AI citations are the links, brand mentions and source references that AI answer engines use after a domain, CMS, protocol or URL migration. To preserve them, treat the move as more than an SEO redirect project: migrate the evidence graph that AI systems use to describe, rank and cite your brand.

What are site migration AI citations?

Site migration AI citations are the source URLs, brand facts and third-party references that ChatGPT, Perplexity, Gemini, Copilot and Google AI features may use after a site move. The goal is not only to redirect old pages. The goal is to make the new domain discoverable, citable and correctly understood.

Traditional migration work asks: “Can Googlebot reach the new URL?” AI citation migration adds three questions:

1. Can answer engines fetch the new source?
2. Do they recognize the new domain as the same entity?
3. Do buyer prompts still mention and cite the brand after the move?

That distinction matters because AI answers often combine owned pages, search indexes, third-party profiles, comparison pages, reviews, community discussions and cached source memories. A technically clean 301 migration can still leave AI systems citing old URLs or describing the company with outdated facts.

Quick answer: how to preserve AI citations during a site migration

To preserve AI citations during a domain migration, run the migration in five layers:

1. Baseline prompt visibility before launch. Export mentions, citations, answer positions, competitors and screenshots for high-intent prompts.
2. Map old cited URLs to equivalent new URLs. Prioritize pages already cited by AI systems, not only pages with traffic or backlinks.
3. Use one-hop permanent redirects. Old cited URLs should resolve directly to matching new pages, not to the homepage.
4. Keep crawlers unblocked. Test robots.txt, server responses, CDN rules and WAF rules for Googlebot, OAI-SearchBot and PerplexityBot.
5. Repair the external evidence graph. Update third-party profiles, review sites, partner pages, directories, documentation links and brand facts.
6. Track prompt recovery for 90 days. Measure new-domain citation rate, stale URL citation rate, answer accuracy and AI share of voice by engine.

Google’s own site move guidance recommends URL mapping, permanent redirects, updated internal links, sitemaps and monitoring in Search Console. For AI visibility, those steps are the base layer, not the whole job.

What most site migration checklists miss

Most site migration checklists are built for organic rankings, not AI-generated answers. They usually cover redirects, canonicals, sitemaps and Search Console. Those are necessary, but they do not prove that AI answer engines have switched their supporting evidence from the old domain to the new one.

Migration area	Standard SEO check	AI citation check
Redirects	Old URLs return 301 or 308	Old AI-cited URLs resolve to equivalent new URLs in one hop
Indexing	New URLs are crawlable and indexable	New URLs are eligible to appear as supporting sources in AI answers
Internal links	Navigation and sitemaps use new URLs	Cited source clusters point to the new domain
Robots and WAF	Googlebot is allowed	Googlebot, OAI-SearchBot, PerplexityBot and relevant fetchers are not blocked where visibility is desired
Canonicals	New pages canonicalize to themselves	Old and new domains do not create conflicting entity signals
Reporting	Rankings, clicks and index coverage	Prompt-level mentions, citations, stale URLs, competitor replacement and answer accuracy
External references	Backlinks are monitored	Review sites, partner listings, app marketplaces, docs, comparison pages and community sources are updated

Google says AI Overviews and AI Mode can use query fan-out, issuing multiple related searches across subtopics and data sources before showing supporting links in its AI features documentation. That means a migrated homepage can be fine while hidden subtopic searches still surface old docs, old marketplace profiles or stale comparison articles.

Why AI citations vanish after a domain change

AI citations usually vanish because the evidence chain becomes inconsistent. The new domain may be live, but the sources that support AI answers are split between old URLs, blocked crawlers, stale third-party pages and conflicting brand descriptions.

Failure pattern	What it looks like in AI answers	Likely cause	Fix
Old URL still cited	Perplexity or an AI Overview links to the retired page	Redirect exists, but the old URL was not recrawled or the third-party source still links to it	Confirm one-hop redirect, keep old sitemap temporarily, update high-value external references
Brand disappears from shortlist prompts	Competitors replace the brand in “best tools” prompts	Source graph lost trust or new pages are harder to discover	Rebuild internal links, update category pages, refresh earned sources
Wrong product or category appears	AI answer uses old positioning	About page, schema, review profiles or partner pages still contain old facts	Update entity facts across owned and third-party sources
ChatGPT misses the site but Google ranks it	Search crawler access differs by engine	OAI-SearchBot is blocked by robots.txt, CDN or WAF	Audit AI crawler access and server logs
AI Overview recovers but Perplexity does not	Engine-specific index lag or crawler block	Different retrieval systems and crawl patterns	Monitor by engine instead of using one blended metric

This is why a migration needs both standard SEO QA and AI crawler access checks. Crawl access determines whether engines can fetch the evidence. Entity consistency determines whether they trust it. Prompt tracking determines whether buyers still see it.

The Citation Continuity Score

The Citation Continuity Score is a practical migration readiness framework for AI visibility. Score six signals from 0 to 100, then track the weighted score daily from launch through day 45 and weekly through day 90.

Signal	Weight	What to measure	Pass condition
Redirect parity	20%	Old cited URLs resolve to equivalent new URLs	95%+ of high-value cited URLs redirect in one hop
Index eligibility	15%	New pages return 200, are indexable and have self-referencing canonicals	No critical cited target is noindexed, blocked or canonicalized away
AI crawler access	20%	Search and answer-engine bots pass robots.txt, CDN and WAF controls	Verified access in logs for priority templates
Source graph parity	15%	Third-party pages update links, names and descriptions	Top external evidence sources no longer point users to the old domain
Entity consistency	15%	Schema, title tags, profiles and About pages describe the same company	No conflicting brand, product or category facts remain in priority sources
Prompt visibility	15%	AI answers mention, rank and cite the new domain across target prompts	New-domain citation rate trends up and answer accuracy returns near baseline

Use these thresholds:

Score	Migration risk	Editorial decision
0-69	High	Do not launch unless the migration is unavoidable
70-84	Moderate	Launch only with daily monitoring and a rollback plan
85-94	Low	Acceptable for most B2B migrations
95-100	Strong	Appropriate for high-dependence AI search categories

The score is intentionally weighted toward crawler access and redirects because those failures can suppress every downstream signal. Source graph and prompt visibility become more important after launch, when technical issues are fixed but answers still lag.

Baseline AI visibility before the move

Baseline AI visibility before redirects go live. Without a prompt baseline, teams cannot tell whether post-migration loss came from the domain move, normal model variance, competitor changes or a separate visibility problem.

A useful baseline includes:

1. 40 to 100 prompts for category, alternatives, pricing, integrations, security, implementation, reviews and vendor due diligence.
2. Every cited URL and cited domain shown in AI answers.
3. Brand mention position in shortlists and comparison answers.
4. Competitor co-mentions for prompts where buyers compare vendors.
5. Answer accuracy notes for outdated product names, pricing claims, categories or acquisition history.
6. Screenshots or saved answer exports for the prompts that affect revenue or reputation.

This is where rankings are insufficient. A site can rank well in Google and still be missing from AI answers, which is a separate visibility gap explained in why ranking #1 on Google does not guarantee AI answer visibility.

Use a prompt-to-URL matrix before launch:

Prompt cluster	Example prompt	Pre-move cited URL	New target URL	Risk
Category research	“best SOC 2 automation tools for startups”	/features/security	/platform/security	High
Alternatives	“best alternatives to Vendor A”	/compare/vendor-a	/compare/vendor-a-alternative	High
Implementation	“tools with API documentation for compliance workflows”	/docs/api-v1	/docs/api	Medium
Pricing	“which compliance tools publish pricing?”	/pricing	/pricing	High
Reviews	“most recommended compliance automation platforms”	G2 profile	Updated G2 profile	Medium

Add an “AI cited” flag to the migration inventory. A page with modest traffic can be mission-critical if it is the page AI systems use to explain pricing, integrations or security posture.

Build redirects around cited evidence

A redirect map for site migration AI citations starts with evidence, not traffic alone. Inventory the URLs that AI answers cite, then map each old source to the most equivalent new source.

Prioritize these URL groups:

1. Pages cited in ChatGPT, Perplexity, Gemini, Copilot or AI Overviews.
2. Pages with high-intent commercial content: pricing, security, comparison, alternatives, integrations and documentation.
3. Pages linked from third-party sources that AI systems may use for corroboration.
4. Pages with backlinks from review sites, partner pages, analyst content or software directories.
5. Pages that define the entity: homepage, About, product, category, contact and schema-bearing pages.

Use this redirect decision table:

Old URL type	Best action	Why
Equivalent page exists	One-hop 301 or 308 to the matching new URL	Preserves continuity and reduces ambiguity
Page merged into a broader page	Redirect to the closest section-level equivalent	Keeps users and crawlers near the same intent
Page retired with no replacement	Return 404 or 410	Avoids soft 404 patterns and irrelevant evidence
Old comparison page replaced	Redirect to the new comparison or alternatives page	Preserves answer relevance for competitor prompts
Old docs page superseded	Redirect to current docs if behavior is still supported	Prevents implementation answers from citing obsolete docs

Google’s site move documentation recommends preparing a URL mapping, redirecting old URLs to new URLs and monitoring both properties. Google’s redirect documentation also states that permanent redirects such as 301 and 308 are signals that the target should be treated as canonical in Search.

Do not redirect a large group of old cited pages to the homepage. For AI answers, a homepage redirect often destroys the source’s usefulness because the cited claim no longer has page-level support.

Keep AI crawlers unblocked at every layer

AI crawler access must be tested at robots.txt, HTTP status, CDN and WAF layers. A crawler can be allowed in robots.txt and still fail because bot protection challenges it, rate-limits it or blocks its IP range.

Crawler or control	Why it matters in a migration	Official source
Googlebot	Controls crawling for Google Search and eligibility for AI Overviews or AI Mode supporting links	Google AI features documentation
OAI-SearchBot	Used to surface websites in ChatGPT search features	OpenAI crawler documentation
GPTBot	Used for OpenAI foundation-model training crawling; it is independent from OAI-SearchBot controls	OpenAI crawler documentation
PerplexityBot	Used to surface and link websites in Perplexity search results	Perplexity crawler documentation
CDN and WAF rules	Can block legitimate bots even when robots.txt allows them	Vendor logs and official bot IP ranges

A visibility-friendly robots.txt policy might look like this:

User-agent: Googlebot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: GPTBot
Disallow: /

This example allows search visibility while blocking OpenAI training crawling. The policy decision belongs to the publisher. The migration requirement is to make search crawler access intentional, not accidental.

Do not rely only on user-agent spoofing tests. They are useful for catching obvious blocks, but server logs are stronger evidence. For priority templates, confirm that legitimate crawler requests receive:

1. 200 on the new target URL
2. 301 or 308 on the old URL
3. No JavaScript challenge
4. No geo block
5. No rate-limit response
6. No canonical pointing back to the old domain

Send fresh discovery and canonical signals

After launch, the new domain needs fresh crawl signals that reinforce the redirect map.

Run these checks in the first 72 hours:

1. Submit XML sitemaps for the new domain.
2. Keep the old sitemap temporarily, with old URLs returning permanent redirects.
3. Update internal links so users and crawlers do not depend on redirects inside the new site.
4. Use self-referencing canonicals on new URLs.
5. Update hreflang, Open Graph URLs, structured data URLs and breadcrumb URLs.
6. Return 404 or 410 for content that truly no longer exists.
7. Request recrawls for the highest-value migrated URLs in Search Console.
8. Watch server logs for old cited URLs and new target URLs.

Google says a small to medium-sized site move can take a few weeks for most pages to move in its index, while larger sites can take longer. AI citation recovery can lag beyond index migration because answer engines may refresh source selection on different schedules.

Preserve entity consistency across the web

Entity consistency tells AI systems that the new domain belongs to the same organization. If the old domain, new domain and third-party sources tell different stories, answer engines may hedge, omit the brand or cite safer competitors.

Update these owned signals first:

1. Organization schema with the new URL, logo, sameAs profiles and contact details.
2. About, contact, pricing, security and product pages.
3. Product names, category labels and value proposition language.
4. Footer links, legal pages, terms, privacy policy and media kit.
5. Article, Product, SoftwareApplication and Breadcrumb schema where relevant.
6. Search Console, analytics, ad platforms and tag management references.

Then update the external evidence graph:

1. LinkedIn, X, YouTube, GitHub and marketplace profiles.
2. G2, Capterra, Product Hunt, Crunchbase and app directories.
3. Partner listings, integration directories and reseller pages.
4. Press boilerplates, newsroom pages and launch announcements.
5. Analyst pages, affiliate comparisons and high-ranking alternatives pages.
6. Community sources that influence buyer research, including Reddit threads where appropriate.
7. Wikipedia or Wikidata only when the company legitimately meets notability requirements.

These are not vanity mentions. They are corroborating sources. For more examples, see earned AI citation sources that brands often overlook.

Track prompts, not only pages, during recovery

Prompt tracking shows whether buyers still discover the brand after the migration. Page tracking shows whether URLs are indexed. Both matter, but only prompt tracking catches the business risk of losing AI-generated recommendations.

Track these metrics by engine:

Metric	Healthy pattern	Warning sign
New-domain citation rate	Rises steadily after launch	Old domain remains the dominant cited source after 30 days
Stale URL citation rate	Declines week by week	AI answers cite redirected, dead or wrong old URLs
Brand mention rate	Returns near baseline	The brand disappears from shortlist prompts
Answer position	Brand remains in comparable rank band	Competitors replace the brand in top recommendations
Description accuracy	New positioning appears consistently	Old product names, categories or acquisition facts remain
Citation quality	Answers cite relevant pages	Answers cite homepage, generic pages or outdated third-party summaries
Engine variance	Some lag differs by engine	One engine stays stale across every prompt cluster

Model and index refreshes can also reshape visibility after the technical migration has stabilized. Treat the 90-day monitoring window as a way to separate migration defects from normal AI answer volatility. The mechanics of that volatility are covered in how model updates affect AI visibility.

Watch query fan-out for hidden migration failures

Query fan-out creates hidden migration risk. One buyer prompt can trigger related searches for reviews, alternatives, pricing, implementation, integrations, security and community sentiment. If only the homepage and top product pages migrated cleanly, AI systems may still find old evidence through those subtopics.

Example:

User prompt	Hidden source cluster	Migration risk
“Best SOC 2 automation tools for startups”	Pricing, security, reviews, alternatives, Reddit, integration docs	Old pricing page or stale review profile may influence the answer
“Is Vendor X good for enterprise compliance?”	Security docs, case studies, G2, partner pages, analyst lists	Old security URL may remain the best supporting source
“Vendor X vs Vendor Y”	Comparison pages, third-party rankings, YouTube reviews, forums	Old comparison pages may redirect poorly or conflict with new positioning

The fix is to audit prompt clusters, not only landing pages. For each priority prompt, record the owned page, third-party source and answer quote that supports the generated response. Then update or replace each stale evidence source.

A 90-day AI citation migration timeline

AI citation recovery should be managed as a 90-day program. The first phase prevents technical loss. The second repairs external evidence. The third confirms answer stability.

Timing	What to do	Recovery signal
Day -30 to -14	Baseline prompts, citations, competitors, screenshots and answer accuracy	Pre-move AI share of voice is known
Day -14 to -7	QA redirects, canonicals, robots.txt, staging noindex removal, CDN and WAF rules	No critical crawler or cited URL fails
Day -7 to -1	Freeze redirect map and update launch-critical third-party profiles	Priority entity facts match the new domain
Launch day	Enable one-hop 301 or 308 redirects and publish new sitemaps	Old cited URLs resolve to equivalent new URLs
Day 1 to 3	Inspect logs and test top cited URLs manually	Googlebot, OAI-SearchBot and PerplexityBot reach target pages
Day 4 to 14	Repair external profiles, review sites, partner pages and docs links	Old-domain references begin declining in source graph
Day 15 to 45	Track prompts daily and fix stale answer sources	New-domain citations rise by prompt cluster
Day 46 to 90	Monitor model/index refresh effects and executive reporting	AI share of voice and answer accuracy remain within tolerance

The recovery goal is not that every old URL disappears instantly. A better goal is that priority buyer prompts cite valid new-domain pages, describe the company accurately and show a declining stale-domain citation rate.

Troubleshooting AI citation loss after launch

Use the symptom to find the broken layer.

Symptom	First check	Likely repair
Old domain still appears in AI answers	Does the old cited URL redirect one hop to the equivalent new page?	Fix redirect chains and keep the old sitemap temporarily
New domain is indexed but not cited	Are high-intent pages specific, internally linked and eligible for snippets?	Strengthen page-level evidence and internal discovery
ChatGPT does not cite the site	Is OAI-SearchBot allowed in robots.txt and WAF logs?	Allow OAI-SearchBot and confirm requests from published IP ranges
Perplexity cites stale pages	Are external sources still linking to the old domain?	Update third-party source graph and priority profiles
Google AI Overviews show old facts	Are schema, About pages and third-party descriptions aligned?	Clean entity facts and request recrawls for key pages
Competitors replace the brand	Did prompt-level source support weaken after the move?	Refresh comparison, alternatives, proof and category pages
Branded prompts work but category prompts fail	Did the migration preserve homepage signals but lose category evidence?	Rebuild category, integration, pricing and use-case source clusters

If the migration includes a rebrand, merger or acquisition, treat the entity layer as a separate workstream. A domain move changes URL identity. A rebrand changes company identity. The same issue appears in M&A scenarios, where AI systems may keep old ownership, product or brand facts unless the evidence graph is corrected.

Common mistakes that damage AI visibility

The most damaging migration mistakes are usually small shortcuts made under launch pressure.

Avoid these:

1. Carrying staging noindex or robots blocks into production.
2. Allowing Googlebot while blocking OAI-SearchBot or PerplexityBot at the WAF.
3. Redirecting retired comparison, pricing or docs pages to the homepage.
4. Changing the domain, CMS, URL structure and brand positioning at the same time.
5. Leaving the old domain live with duplicate content and unclear canonicals.
6. Updating owned pages but ignoring G2, GitHub, partner pages, marketplaces and review profiles.
7. Measuring traffic only when the real loss is fewer AI citations in buyer prompts.
8. Assuming Google AI Overview recovery means ChatGPT or Perplexity also recovered.
9. Removing old sitemaps too quickly before crawlers have processed key redirects.
10. Failing to save screenshots before launch, making post-migration diagnosis subjective.

Google states that existing SEO fundamentals remain relevant for AI Overviews and AI Mode, and that supporting-link eligibility depends on being indexed and eligible for snippets. That makes technical SEO the floor. It does not replace source graph repair or prompt-level measurement.

Post-migration checklist for AI citations

Run this checklist daily for the first week, twice weekly through day 45 and weekly through day 90.

1. Redirects: Top old cited URLs return one-hop 301 or 308 responses to equivalent new URLs.
2. Status codes: New cited URLs return 200 and are not blocked, challenged or redirected again.
3. Canonicals: Every migrated page has a self-referencing canonical on the new domain.
4. Internal links: Navigation, related content, breadcrumbs, XML sitemaps and hreflang use new URLs.
5. Robots: Googlebot, OAI-SearchBot and PerplexityBot are allowed where visibility is desired.
6. WAF: Official bot IP ranges and user agents are not challenged, rate-limited or geo-blocked.
7. Structured data: Organization, Article, Product, SoftwareApplication and Breadcrumb schema match visible page content.
8. Entity facts: Brand name, product names, category, logo, contact details and sameAs links are consistent.
9. Third-party sources: High-value profiles, review pages, partner listings and comparison pages point to the new domain.
10. Prompt tracking: Target prompts show new-domain citations and accurate descriptions.
11. Competitor monitoring: Competitors have not replaced the brand in high-intent shortlists.
12. Evidence capture: Important answer changes are saved for SEO, PR and executive reporting.

A site migration is complete only when the new domain is discoverable, citable and correctly described across the answer engines that buyers actually use.

Frequently Asked Questions

Do 301 redirects preserve AI citations?

301 redirects help, but they do not automatically preserve AI citations. They tell crawlers that a URL moved permanently, but answer engines may still cite cached old URLs, stale third-party sources or indexes that have not refreshed. Use redirects as one part of a broader citation continuity plan.

How long does AI citation recovery take after a domain migration?

Expect recovery in phases. Crawl access issues can show up within days. Google says small to medium-sized site moves can take a few weeks for most pages to move in its index, and larger sites can take longer. AI answer recovery may take longer because each engine refreshes retrieval and citations differently.

Should a site allow GPTBot during a migration?

Not necessarily. OpenAI separates OAI-SearchBot, which is used for ChatGPT search visibility, from GPTBot, which is used for foundation-model training crawling. A publisher may allow OAI-SearchBot while disallowing GPTBot. The key migration step is to avoid accidentally blocking the search crawler needed for visibility.

What should be measured besides organic traffic?

Measure brand mentions, cited URLs, cited domains, answer position, competitor inclusion, stale-domain citations, incorrect descriptions and AI share of voice. Organic traffic can remain stable while AI-generated buyer shortlists stop recommending or citing the brand.

Is a rebrand different from a domain migration?

Yes. A domain migration changes URL identity. A rebrand changes entity identity. If both happen together, update redirects and entity evidence at the same time: schema, About pages, social profiles, review sites, partner pages, media mentions and prompt-level monitoring.

What is the biggest AI citation risk in a site migration?

The biggest risk is partial continuity. The main pages redirect correctly, but AI systems still find old evidence through review profiles, partner pages, old documentation, comparison content or blocked crawler paths. That is why prompt-to-source mapping matters more than a generic redirect spreadsheet.

Can a site recover AI citations without updating third-party sources?

Sometimes, but recovery is weaker and slower. If important third-party pages still link to the old domain or describe the old brand, AI systems may keep using them as corroborating evidence. Updating high-trust external sources is often the difference between indexed and confidently cited.