Website & Development
GEO in a Website Redesign: The Briefing and SEO Checklist
Copy for AI
A redesign is the cheapest moment to build in AI visibility.
It’s also the costliest moment to lose it.
The verdict: put GEO in your website redesign briefing as a testable requirement, next to design and conversion. Fixing it after launch costs a second project.
GEO (Generative Engine Optimization) means being visible in the AI answers of ChatGPT, Perplexity and Google AI Overviews.
In practice, that means three things:
- AI crawlers can read your content
- your structure and data are machine-readable
- your existing citations survive the migration, through redirects and crawler access
Skip them and your agency builds a beautiful new site that AI models understand worse than the old one.
Below: what belongs in the briefing, how to stop a redesign wrecking your AI visibility, and a website redesign SEO checklist for launch. It ties into your broader redesign process and the wider approach to building a B2B website.
Why do you lose AI citations precisely during a redesign?
Because citations are tied to URLs and content, and a redesign changes both at once.
Citations live on URLs, not on your brand
Citations in AI answers don’t sit “inside your brand” as a fixed asset.
Like your SEO rankings, they’re tied to specific URLs and to content AI models once read and judged trustworthy.
During a redesign, three things often change at once:
- the URL structure
- the way content is technically loaded
- the content itself
Change that without a plan, and the models lose the trail to exactly the pages you were cited for.
Most AI crawlers don’t run JavaScript
A second risk matters less in classic SEO. Many modern sites load their most important content only through JavaScript, in the visitor’s browser.
For Google that’s usually manageable, because Googlebot executes JavaScript.
For AI crawlers it’s different. Very different.
Vercel’s analysis of AI crawler traffic (December 2024) found that none of the major AI crawlers render JavaScript. That includes OpenAI’s GPTBot, OAI-SearchBot and ChatGPT-User, Anthropic’s ClaudeBot and PerplexityBot.
Exception: Gemini, because it runs on Googlebot’s rendering infrastructure.
And these crawlers aren’t a rounding error. Vercel counted their monthly requests across its network:
That’s close to a billion AI crawler requests a month on one network, all reading raw HTML.
Client-side content is invisible to them
Content that appears only client-side simply doesn’t exist for those models.
A redesign onto a heavy JavaScript framework can make you unintentionally invisible to the engines you want to be found in.
The same Vercel analysis found that ChatGPT’s crawler spent 34.82% of its fetches on 404 pages. Broken URLs after a launch only make that worse.
The good news
This is fully manageable, provided you put it in the assignment up front.
Fixing it afterwards is expensive and slow. That’s why GEO belongs in the briefing, not in the evaluation.
What goes into the website redesign briefing for GEO?
GEO goes in as an explicit, testable requirement, on equal footing with design and conversion.
A wish (“AI-proof would be nice”) steers nothing. A requirement with an acceptance criterion does.
If you don’t yet know how to structure such an assignment, start from our approach to writing a strong website briefing or RFP.
The six requirements
These points belong in your assignment to the agency:
- Rendering: core pages deliver their full main content in the first HTML response, via server-side rendering or static generation, not only through client-side JavaScript.
- Structure: a logical, stable URL structure and a clear heading hierarchy, so that both human and machine can follow the website architecture.
- Structured data: relevant schema markup on the right pages, so that models can unambiguously interpret entities like your company, services and authors.
- Crawler access: a robots.txt that deliberately allows the AI search crawlers that earn you citations (such as OAI-SearchBot for ChatGPT search), instead of blocking them by accident.
- Migration protection: a complete redirect and URL-mapping plan for the website migration, so that existing citations and links don’t dead-end.
- Answer-first content: a content model in which every core page opens with a short, citable answer to the question it addresses.
Why now is cheap
These requirements cost almost nothing during the build, because the agency makes those choices anyway.
Building them in afterwards costs a second project. That hurts.
That’s exactly why a redesign is the right moment: you pay once for foundations that carry both your conversion and your AI visibility.
Turn each requirement into a test
An acceptance criterion is something you can check on the staging site before you sign off. For example:
| Requirement | Acceptance test |
|---|---|
| Rendering | Body text visible with JavaScript switched off |
| Structured data | Schema validator shows the expected types per template |
| Crawler access | robots.txt allows OAI-SearchBot, Googlebot and the bots you want |
| Migration | Every old URL returns a 301 to its new address |
| Answer-first | Each core page answers its question in the first two sentences |
How do you make sure AI can read your content?
Your content has to sit in the raw HTML before any JavaScript runs.
Ask for server-side rendering or static generation
Ask your agency explicitly for server-side rendering or static generation for every page you want to be found.
That covers your service pages, your knowledge articles and your most important landing pages.
The JavaScript-off test
A handy test: open a page, turn off JavaScript in your browser, and see whether the body text is still there.
If your content disappears, it also disappears for a large share of the AI engines.
Speed and GEO point the same way
This lines up nicely with other foundations of a good site. Fast-loading, server-rendered pages score better on Core Web Vitals and are more pleasant for visitors.
Here’s what that looks like on our own site:
PageSpeed Insights field data for customerimpact.be on mobile, October 2026. LCP passes at 1.9 s; CLS at 0.13 still needs work.
So you don’t have to trade GEO off against user experience. The same technical choice serves both.
Any platform can pass
We’re platform-agnostic and build in Webflow, WordPress or custom/headless.
Which route fits depends on your team, content and growth plans. But “content in the first HTML” holds on every platform.
How the main CMSs compare for AI search is in the best CMS for AI search. Read more about the trade-off at our website development service.
Rule of thumb: no body text without JavaScript, no sign-off.
Which structure and data help AI cite you?
Structure and data that let a model understand a page quickly and unambiguously.
Three things help, and all three are cheap to build in during a redesign.
1. A clean structure
A stable URL setup and a tidy heading hierarchy make it easy to see what a page is about. Use one H1, with logical H2s framed as questions, such as “How much does X cost?”
Then a model can find which snippet holds the answer.
2. Structured data
With the right structured data, such as Organization, Service or Article schema, you literally tell a machine what an entity is, who the author is and what you offer. It doesn’t have to infer that from running text.
You check it with Schema.org’s validator, as we did on our own homepage:
The Schema Markup Validator on our own English homepage, October 2026. Check that each detected type matches what the page visibly says.
Run that check on every new page template. Not only on the homepage.
3. Answer-first writing
Pages that open with a short, complete answer and only then elaborate are rewarding to cite. The citable block sits right at the top.
That’s exactly the structure you’re reading now.
Lock it in: put this content model into the briefing. Then your redesign delivers pages an AI answer can easily take over, not just prettier pages.
This is the core of GEO. It dovetails with good conversion copy: a clear answer up front helps both the reader and the machine.
Some teams also publish an llms.txt file. It’s cheap to add at launch, but no substitute for the three points above.
Bottom line: clear structure first, extras second.
Website redesign SEO checklist: how do you keep your citations through the migration?
Treat your AI visibility during the transition as carefully as your rankings.
Redirect every cited URL
The most important measure is a complete 301 redirect plan. Every old URL that was cited or linked points, after launch, to the corresponding new address.
Google’s site move guidance recommends server-side permanent redirects (301 or 308) based on an old-to-new URL map. Keep them for at least a year.
Map source: Google suggests building that map from your sitemaps, server logs, analytics and CMS.
Without it, every new URL starts from scratch, and existing references from AI sources dead-end.
This is the same discipline as in a migration without ranking loss, now applied to your citations too. Spot-check the result with our redirect checker.
Expect a delay: Google notes that for medium-sized sites it can take a few weeks or more before the new URLs show instead of the old ones.
Check the robots.txt that goes live
Check that the new site doesn’t block the crawlers that earn you citations.
A fresh build sometimes starts with a locked-down robots.txt from the test phase. If that goes live, you shut out OAI-SearchBot, GPTBot, ClaudeBot and PerplexityBot in one stroke, and even Googlebot.
A deliberate robots.txt can be very short. Here’s ours:
Our own robots.txt, October 2026: one rule for all crawlers, so every AI bot, training and search, may read the site.
Know which token does what
OpenAI’s crawler docs say OAI-SearchBot decides whether you can appear in ChatGPT search, while GPTBot only covers model training.
The same page notes it can take about 24 hours after a robots.txt update for OpenAI’s search systems to adjust.
Google-Extended isn’t a separate crawler and doesn’t affect inclusion in Google Search, AI Overviews included.
Make crawler access an explicit launch point. Test it on day one.
Inventory what’s cited before you break anything
Before launch, inventory which pages are currently cited. Then after launch you can tell whether a drop is normal or alarming.
An AI visibility audit gives you that baseline. Referral traffic from AI tools in GA4 is the second baseline:
GA4 Traffic acquisition in our own property, last 90 days, October 2026: 58 sessions from chatgpt.com. Save this number before launch and compare it after.
How to set up that report is in tracking ChatGPT traffic in GA4.
If you want to tackle this in a structured way, our AI search optimization service monitors your visibility before and after the transition.
The launch checklist in one view
Before you switch DNS, tick these off:
- Old-to-new URL map complete, 301s tested
- Body text present with JavaScript off
- robots.txt allows the crawlers you want
- Schema validates on every template
- Baseline of citations and AI referrals saved
Miss one and you may only notice weeks later. By then the trail may be cold.
The short summary
A redesign decides for years whether AI engines can read and cite your brand.
So set GEO as a testable requirement in the briefing:
- content in the first HTML via server-side rendering
- a clean structure with structured data
- answer-first pages
- a migration plan with 301 redirects and preserved crawler access
Those choices cost almost nothing during the build. They serve your conversion and user experience at the same time.
The nuance: GEO isn’t a switch you flip afterwards, but a foundation you lay during the build, or don’t.
Want to win AI citations with your redesign instead of losing them? Contact us and we’ll review your briefing first.
Free website scan
Enter your website and get an automatic scan within minutes, with concrete technical and SEO improvements. No sales pitch.
We only use your details for your scan. No spam, unsubscribe anytime.