Website Redesign and GEO: Don't Lose Your AI Visibility

You're planning a website redesign. New design, new site architecture, maybe a CMS change. On the SEO side, you've planned your 301 redirects. But have you thought about your AI visibility?

A poorly prepared redesign can make your site disappear from ChatGPT, Perplexity and Gemini responses for weeks or even months. And unlike traditional SEO, where rankings recover gradually, AI visibility can collapse abruptly if the trust signals AI engines had built up on your old site disappear.

Why a redesign is risky for AI visibility

AI engines don't work like Google

Google recrawls your site regularly and updates its index. If you redesign with proper redirects, Google follows them and transfers link equity within weeks.

AI engines work differently:

• LLMs have a frozen memory. Base models (GPT-4, Claude, Gemini) were trained on data up to a specific date. If your old site was in the training data, your new site isn't there yet.
• RAG recrawls, but more slowly. RAG systems (Retrieval-Augmented Generation) rely on search indexes (Google for Gemini; for ChatGPT, Bing but also its own index fed by OAI-SearchBot and other providers, according to a Peec AI analysis) to enrich responses with fresh data. If your URLs change, RAG must rediscover your content.
• The structure of cited pages is fragile. If the new template loses the short answers under headings, the tables or the visible update date of your old pages, it loses what distinguishes cited pages in our calibration study.

What can break during a redesign

• URLs change — existing citations in AI responses point to 404s
• JSON-LD schemas disappear — engines lose information that helps them read your pages (dates, FAQ), even though schemas are not decisive for citation
• Content is reorganized — FAQs, guides, and articles that AI engines cited are moved or merged
• robots.txt changes — AI bots may get blocked without your knowledge
• Author pages disappear — URLs that need redirecting, and useful information for readers is lost (see E-E-A-T and AI)
• Sitemap isn't updated — AI crawlers don't discover the new URLs

Before the redesign: the baseline audit

Step 1: run a GEO audit on the old site

Before touching anything, measure your current AI visibility. This is your baseline — the reference score you'll compare the new site against after migration.

Metrics to capture:

• Overall GEO score and score per criterion (6 criteria)
• AI citation test: on which queries is your site cited by ChatGPT, Perplexity, Gemini?
• JSON-LD schema inventory: which schema types are present and on which pages?
• Most-cited pages list: identify your pages that appear in AI responses (often FAQs, guides, detailed product pages)
• robots.txt configuration: which AI bots currently have access? (see llms.txt, robots.txt and AI accessibility)

Detekia's citation test covers a sample of queries. For a high-stakes site, an AI visibility audit of your brand before the migration sets a more complete baseline: what ChatGPT, Gemini, Claude and Perplexity say about you, on which queries, and against which competitors.

Step 2: map critical AI assets

Identify the pages and elements that contribute most to your AI visibility. These are the elements you must absolutely preserve or reproduce in the new site:

• FAQ pages with FAQPage schema: often the pages most cited by AI engines
• Guides, tutorials and comparisons: "best…" comparison pages are among the most cited pages
• Pages with answer capsules: short answers (one to three sentences) right under headings often phrased as questions, tables, a visible update date. That's what distinguishes cited pages in our study
• Author pages and Organization schema: keep them for readers and site consistency, even though they don't set cited pages apart

During the redesign: non-negotiable rules

1. Exhaustive 301 redirects

This is rule number 1 in SEO, and it's even more important in GEO. Every old URL that existed must redirect to the corresponding new URL. AI engines may have indexed your old URLs in their retrieval system — a 404 means a lost citation.

Key points:

• Don't redirect everything to the homepage. Each page should redirect to its closest equivalent.
• Test redirects with a crawler (Screaming Frog) before launch
• Keep redirects active for at least 12 months

2. Reproduce all JSON-LD schemas

Before migrating, export the complete list of your JSON-LD schemas. After migration, verify that each schema is present on the equivalent page of the new site. Not because schemas get your pages cited (Google states that its AI features require no special markup), but because a lost or wrong schema, such as a reset dateModified, gives engines inaccurate information.

Priority schemas to verify:

• Organization: name, logo, address, social networks, founding date
• Person: authors, experts, executives
• FAQPage: all pages with questions/answers
• Article: with datePublished, dateModified, author
• LocalBusiness: if you have a local presence
• HowTo, Review, ItemList: depending on your content

3. Preserve robots.txt and sitemap

Verify that your new robots.txt allows the same AI bots as the old one. This is a frequent oversight: the new CMS generates a default robots.txt that may block OAI-SearchBot, Claude-SearchBot, or PerplexityBot, the bots that feed AI answers. GPTBot, on the other hand, is used for training: blocking it doesn't stop you from appearing in ChatGPT search (OpenAI documentation).

The sitemap should be submitted to Google Search Console and Bing Webmaster Tools on launch day. AI crawlers use these sitemaps to discover your new URLs.

4. Preserve citable content

Don't delete pages that AI engines cite, even if they don't fit your new site architecture. If a 20-question FAQ was cited by ChatGPT, don't delete it — move it with a redirect and preserve the content.

If you must merge pages, make sure the citable content (direct answers, data points, lists) is preserved in the merged page.

5. Maintain author pages

Author pages are often sacrificed in redesigns. If your authors had pages with a bio, credentials, and article lists, recreate them (or redirect them) on the new site: they are useful to readers and expected in regulated sectors. Don't count on them for citation, though: in our study, an identified author does not distinguish cited pages.

After the redesign: verification and monitoring

Launch day checklist

1. Test all 301 redirects (zero 404s on old URLs)
2. Verify robots.txt (AI bots allowed)
3. Submit new sitemap to Google Search Console and Bing
4. Validate all JSON-LD schemas with Google Rich Results Test (and FAQPage with the Schema Markup Validator, since Google's test no longer supports FAQ)
5. Verify author pages are live and indexable
6. Test site accessibility with JavaScript disabled (for basic crawlers)

Week 1: first check

• Run a GEO audit on the new site and compare with the baseline score
• Check Google Search Console to verify new pages are indexed
• Manually test 5-10 queries on ChatGPT and Perplexity to see if your site is still cited

Months 1-3: recovery monitoring

AI visibility can take time to fully recover after a redesign, even with perfect redirects, and the delay varies by engine and by how often they re-read your pages. This is normal: RAG systems need to recrawl your content and rebuild trust signals.

Run a GEO audit every 2 weeks for the first 3 months to track recovery. If the score doesn't bounce back from one audit to the next, check:

• Are the short answers, tables and update dates still in place?
• Is robots.txt blocking any AI bot?
• Are all redirects working?
• Is the citable content still present and accessible?

Special case: CMS migration

A CMS change (WordPress to Webflow, Shopify to custom, etc.) is the riskiest scenario for AI visibility. The new CMS generates a different HTML structure, different URLs, and often different schemas.

Additional rules for a CMS change:

• Export raw content (text, images, meta) before migration. Don't rely on the new CMS's automatic import.
• Check the HTML structure. Some CMS platforms (Wix, Squarespace) generate HTML wrapped in JavaScript. AI crawlers may not be able to read it.
• Test schemas on the new CMS. Each CMS handles schemas differently. What worked on WordPress (via Yoast) won't be automatically reproduced on Webflow.
• Beware of SPAs. If the new site is a Single Page Application (React, Vue, Angular), content may be invisible to AI crawlers. Use Server-Side Rendering (SSR) or Static Site Generation (SSG).

GEO redesign checklist

Before:

1. Baseline GEO audit (score, citations, schemas)
2. Inventory of pages cited by AI engines
3. Export of all JSON-LD schemas
4. Page-by-page 301 redirect plan

During:

1. Exhaustive 301 redirects
2. All JSON-LD schemas reproduced
3. robots.txt with AI bots allowed
4. Updated sitemap submitted to search engines
5. Author pages preserved
6. Citable content preserved

After:

1. GEO audit of the new site (comparison with baseline)
2. Manual citation test on ChatGPT/Perplexity
3. Monitoring every 2 weeks for 3 months

Conclusion

A website redesign is an opportunity to improve your AI visibility — as long as you don't lose what you already have. Your answer structure (short answers, question headings, tables, dates), FAQs, AI bot access, and redirects are the four pillars you must absolutely preserve. Run a GEO audit before the redesign, reproduce that structure identically, and monitor recovery for 3 months. That's the difference between a redesign that advances your AI visibility and one that makes it disappear.

→The advanced schemas most sites forget

→The complete GEO guide for 2026