What Should a Web Designer Do for Google AI and ChatGPT Visibility?
By Parker Gawryletz · Local Cochrane Web Design · Last updated: · 8 min read
AI search optimization for a new website starts with normal Google Search eligibility: crawlable, server-rendered HTML, helpful people-first content, and accurate structured data. There is no special AI schema. Add clear entity details, answer-first sections backed by first-party evidence, and allow AI crawlers in robots.txt. Then measure citations honestly.
Ask ten agencies about AI search optimization for a new website and you will hear ten different promises, some dressed up as secret techniques. The honest answer is simpler. Google's own documentation says its AI features draw on the same index and quality signals as normal Search.
This guide lists what a web designer should actually build in — and what to ignore. It is the checklist behind our AI search optimization for Cochrane businesses service, written so you can hold any provider to it.
The short answer
- AI visibility starts with normal Google Search eligibility, not secret markup.
- There is no special AI schema — accurate standard structured data is enough.
- Server-rendered, crawlable HTML and clear internal links let AI systems read the site.
- Consistent business details and first-party evidence make you citable.
- Measure mentions and citations honestly — there is no AI rank to track.
AI visibility starts with normal Search eligibility
A page cannot appear in Google's AI features unless it is eligible for normal Google Search first: crawlable, indexable, and helpful to people. That means the foundations of AI visibility are the same foundations as SEO — there is no separate AI index and no shortcut around Search eligibility.
| Requirement area | What to put in the build scope | How to verify |
|---|---|---|
| Technical access | Server-rendered HTML, clean URLs, a sitemap, AI crawlers allowed in robots.txt | Load a page with JavaScript off — the content is still there |
| Content structure | Question headings with direct answers, plus lists and tables for key facts | Each section makes sense read entirely on its own |
| Entity clarity | Identical business name, details, and service area on every page and profile | Search your business name — details match everywhere |
| Evidence | Real reviews, named projects, first-party numbers and case studies | Every claim traces to something you can show |
| Measurement | Analytics plus a running log of AI mentions and citations | You can say where the site was cited last month |
These are the same items a good build quote already covers. If you are scoping a project, read the SEO requirements for a new website first — AI visibility is built on top of that list, not beside it.
What Google currently says about generative AI optimization
Google's published guidance is short: there is no special optimization for AI features and no extra markup that unlocks them. Sites become eligible through normal indexing, and the long-standing advice — helpful, reliable, people-first content — still applies. Any provider promising a proprietary AI ranking trick is promising something Google says does not exist.
The useful work is concrete: crawlable pages, clear structure, real evidence, and accurate structured data. The rest of this guide turns those into build-scope language you can hand to a designer.
Crawlable server-rendered content and internal links
AI systems read your site through crawlers, and most of them handle JavaScript poorly. Your key content should arrive as server-rendered HTML, reachable through plain internal links, with a sitemap and a robots.txt file that allows the crawlers you want — including AI agents such as GPTBot if you want ChatGPT to read the site.
- Server-rendered pages — the full article exists in the HTML response, not only after scripts run.
- Crawlable links — real anchor links between related pages, so every page can be discovered.
- Robots decisions in writing — which crawlers are allowed, and why, documented in the build scope.
Entity clarity and consistent business details
AI systems answer questions about businesses by matching details across many sources. Your legal name, service area, address details, and services should read identically on your website, your Google Business Profile, and every directory. For a Cochrane business, that consistency is what lets an AI system say who you are with confidence.
Give the site one canonical about page stating the plain facts: what you do, where, and since when. Then keep one spelling of the name everywhere else. Mismatched listings do not disqualify you — they just make every system that reads them hesitate.
Answer-first sections backed by first-party evidence
Structure pages so each section opens by answering its own heading in two or three plain sentences, then supports the answer with detail. AI systems lift self-contained passages, and passages earn citations when they carry evidence the model cannot find elsewhere — your projects, your numbers, your reviews, your experience.
For a local business, first-party evidence is the whole game. Generic industry advice exists in a thousand places; a named project on a real Cochrane street, a review from an actual customer, and the prices you genuinely charge exist only on your site.
Accurate structured data—not special AI schema
Use standard schema.org structured data that matches what is visible on the page — organization details, articles, FAQs, local business information. There is no AI-specific schema type, and Google's structured data policies require markup to describe real page content. Accurate, boring markup helps machines parse your site; invented markup works against you.
Original research, case studies and transcripts
The content most worth building is content only you can publish: documented projects with real outcomes, local pricing you actually charge, honest process explanations, and notes or transcripts from real customer questions. AI systems synthesise from many sources — original material is what makes yours the source worth naming.
This is why we keep every project on our live Cochrane client builds page linked to the real website it describes. A case study a reader can verify is a case study a machine can trust.
What llms.txt can and cannot do
llms.txt is an emerging convention: a plain-text file at your site's root that summarises what the site is about and lists key pages for AI systems. It can make a site easier for LLM tools to digest. It is not a Google standard, and there is no evidence it affects rankings.
It is still worth adding, because it costs minutes and harms nothing. Treat it as a courtesy summary for machines — never as the deliverable an AI plan is priced on.
How to measure AI search visibility without a fake rank
There is no AI ranking position to track, so honest measurement looks different: log when AI tools mention or cite your business, watch branded search and direct traffic, ask new customers how they found you, and check that your key pages stay indexed. Distrust any report that shows an AI rank.
- Ask the major AI tools your customers' real questions each month, and record what gets cited.
- Track enquiries that mention finding you through an AI answer.
- Watch Search Console for the pages and queries that keep earning impressions.
This is how we report our own service ($450–$1,000 per month): citations found and logged, never a rank. The same discipline is a good hiring filter — the best Cochrane web design companies compared guide applies it to every kind of proof.
Frequently asked questions
Is AI SEO different from normal SEO?
Mostly no. Google's AI features draw on the same index and quality signals as normal Search, so the foundations are identical: crawlable pages, helpful content, accurate markup. The difference is emphasis — AI systems favour self-contained answers, clear entities, and first-party evidence, so those deserve extra attention in a new build.
Does Google require special AI schema?
No. Google's documentation describes no AI-specific schema type and no extra markup that unlocks AI features. Standard structured data that accurately describes visible page content is the requirement, exactly as it is for normal Search. Treat any pitch built around secret AI markup as a sign to keep shopping.
How do I get cited in AI Overviews?
Be eligible for normal Search first: indexed, crawlable, and genuinely helpful. Then structure pages so each section answers a real question in a self-contained passage, and back the answers with first-party evidence — real projects, reviews, and numbers. There is no guaranteed placement, and anyone promising one is guessing.
Can ChatGPT crawl my website?
Yes, if your site allows it. OpenAI's crawlers, such as GPTBot, respect robots.txt, so you can allow or block them there. The site also needs to serve crawlable HTML — content that only appears after JavaScript runs may be invisible to them. Decide your crawler policy deliberately and write it into the build scope.
Does llms.txt improve rankings?
There is no evidence llms.txt affects Google rankings, and Google has not adopted it as a standard. It is an emerging convention: a readable summary that can help AI systems understand a site's purpose and key pages. It costs little to add, but it should never be sold as a ranking deliverable.
What content do AI systems cite?
Content they can read cleanly and cannot find elsewhere: direct answers under clear headings, lists and tables of concrete facts, original case studies, first-party numbers, and consistent business details. Generic advice already exists in a thousand places — your own documented, verifiable experience is the material worth citing.
Sources
See our AI search optimization service — an honest monthly program — crawlability, entity clarity, answer-first content, and citation reporting with no invented rank.