Building an AI visible website from scratch means making a small set of decisions before you write any page: render content on the server, keep robots.txt open to the search crawlers, structure each page around a customer question, and plan pages by service and place. The bottom line: an AI visible site is mostly a well built ordinary site, so get the foundation right first, since Google says there are no additional technical requirements beyond being indexed and snippet eligible.
Related: our guide to answer engine optimization defines the term and lists eight steps.
The build order
| Phase | Decision | Why |
|---|---|---|
| 1. Platform | Choose static generation or server rendering, not client only rendering | Vercel found the major AI crawlers do not run JavaScript |
| 2. Access | robots.txt allowing search bots, a Sitemap line, no blocking firewall rules | Blocked bots cannot retrieve you |
| 3. Structure | One H1 and clear H2 sections on every page | Clear boundaries for passages |
| 4. Pages | One page per service and per place, plus pricing, about, contact, and a page per key question | Specific questions need specific pages |
| 5. Identity | Organization or LocalBusiness markup with sameAs, consistent name, address, and phone | Helps machines recognize one entity |
| 6. Discovery | Sitemap with accurate lastmod, Search Console, Bing Webmaster Tools, IndexNow | Get into both indexes quickly |
| 7. Proof | Reviews, case pages, credentials, dated numbers with sources | Corroboration and quotable evidence |
Steps
- Pick a platform that outputs full HTML. A static site generator, or a framework in server rendering mode. Plain pages on any host also work.
- Write robots.txt first. Allow the search and user fetch bots for OpenAI, Anthropic, and Perplexity, and decide separately on training bots.
- Plan the page list. Start from 20 customer questions and group them by service and place.
- Build one page template. One H1, answer paragraph first, sections, a question block, and a sources or proof section.
- Write each page answer first. Direct answer, then detail, with prices or ranges and real examples.
- Add structured data from the same fields that render the page. So the two never drift apart.
- Add real images with alt text. Photos of your work, team, or location.
- Test with curl. Confirm the words are in the raw response.
- Verify in Search Console and Bing Webmaster Tools. Submit the sitemap.
- Measure a baseline. Run your questions before launch, then monthly.
Mistakes to avoid from the start
- A single page that tries to cover every service.
- Content loaded by a script after the page opens.
- A sitemap that stamps today's date on every URL, which teaches search engines to ignore lastmod.
- Hiding text in tabs that load on click.
- Skipping Bing because you only watch Google.
What the build cannot do
A well built site makes you eligible and readable. Whether AI names you also depends on outside sources and competitors, and answers vary run to run, so plan reviews, directory listings, and monthly measurement from launch.
Questions people ask
How do I build a website that AI can read?
Use static generation or server rendering so content is in the HTML, allow the search crawlers in robots.txt, structure pages with one H1 and clear sections, and write answer first pages.
Does the platform matter for AI visibility?
Yes for rendering. Vercel found major AI crawlers do not run JavaScript, so client only rendered pages can look empty to them.
What pages should a new site have?
One page per service and place, plus pricing, about, contact, and pages for key customer questions.
What should I set up at launch?
Search Console, Bing Webmaster Tools, a sitemap, IndexNow, and a baseline measurement of customer questions.