Skip to content

AI Website SEO, Is a Site Built with AI Good for Google?

An AI website is as good for SEO as the HTML, speed and content it ends up with. Google judges pages, not the tool that built them. This guide shows what Google has actually said, which technical points decide whether an AI-built site ranks, and how to check your own in fifteen minutes.

10 min read

FactorWhy it decidesHow to check it
HTML before JavaScriptThe crawlers behind ChatGPT, Claude and Perplexity read the raw HTML and run no scriptsOpen the page source and search for your main heading
SpeedGoogle uses Core Web Vitals in its ranking systems, and slow pages lose visitors firstPageSpeed Insights or Lighthouse, mobile view first
Structured dataMakes facts like address, prices and authors unambiguousGoogle's Rich Results Test, plus a read for invented facts
Internal linksCrawlers find and weigh pages through real linksEvery page reachable in three clicks, no page without a link to it
Original contentGoogle rewards helpful content however it was producedWould a customer learn something here that ten other sites do not say?
Titles and descriptionsAI drafts tend to repeat one pattern on every pageOne unique title per page, the search term at the start
llms.txtOptional orientation for language models, not a ranking factorFive minutes of work, it cannot break anything
Being cited by answer enginesAssistants quote pages that already rank and answer clearlyAsk ChatGPT, Perplexity and Gemini, read your server logs

Yes, an AI website can rank as well as any other. Google judges the page, not the tool that built it, and says so in its own guidance. What decides is what the AI produced, whether crawlers can read the HTML, how fast it loads, how pages link and whether the content is worth finding.

We build static websites with AI coding tools every day and run our own pages through the same checks we describe here. This guide is the short version of what we have learned, with links to the primary sources so you can verify every claim yourself.

What Google says about AI-generated content

Google answered the question in February 2023, in a post on its Search Central blog titled Google Search’s guidance about AI-generated content. The central sentence is short. „Appropriate use of AI or automation is not against our guidelines.“ The same post adds that „using AI doesn’t give content any special gains“. AI earns you nothing and costs you nothing. The page is judged like any other page.

Quality over origin

Google describes its approach as a focus „on the quality of content, rather than how content is produced“. In practice that means the usual signals for helpful content still apply, experience, expertise, a clear author and a reason for the page to exist.

The more detailed page on using generative AI content, last updated on 1 October 2026, makes two practical points. Generative AI can be useful for research and for adding structure to original content. And models predict likely words, so they can be wrong. Google asks site owners to check AI output before publishing, explicitly including titles, meta descriptions, structured data and image alt text. Those are exactly the fields AI tools fill in without being asked.

Where AI content turns into spam

The line is drawn in the spam policy on scaled content abuse, introduced in March 2024. It targets many pages generated mainly to manipulate rankings rather than to help people, no matter whether automation, humans or both made them.

For an AI-built site the typical trap is the city page. Two hundred pages that read „Plumber in Leeds“, „Plumber in York“ and so on, with nothing but the town name swapped, are the textbook case. An AI writes them in ten minutes, and that speed is the problem. Five service pages with real prices and real examples beat two hundred of them.

Why the tool matters less than the output

„AI website“ covers very different things. A page from a hosted builder, an app generated in a chat, and a static site edited by an AI coding tool all carry the same label, and they reach Google in very different shape.

Hosted builders such as Wix and Framer serve finished HTML and take care of sitemaps, mobile layouts and basic metadata. Their limits are elsewhere. The copy their AI drafts is generic, because every user in your industry gets a similar draft. And you control less of the technical detail.

App builders work differently. Lovable, for example, documented for a long time that its React apps render in the browser by default and recommended prerendering for search engines. Several 2026 guides report that newer projects now render on the server. That shows the real question. It is never „was AI involved?“ but „what does the server send when a crawler asks?“ Our comparison of AI website builders goes through the tools one by one.

What decides ranking for an AI-built site

The table at the top sums it up. Here is what sits behind each line.

HTML that exists before JavaScript runs

In December 2024 Vercel published an analysis of AI crawlers on its network, The rise of the AI crawler. In one month GPTBot made 569 million fetches and Claude’s crawler 370 million. None of the crawlers from OpenAI, Anthropic, Perplexity or Meta executed JavaScript. Only Gemini, which uses Google’s infrastructure, and AppleBot rendered pages.

Google itself does render JavaScript, as its JavaScript SEO basics explain, but in a second step after crawling. For everyone else, text that only appears in the browser does not exist.

The test takes ten seconds. Open your page, view the source (not the inspector) and search for your main heading. If it is there, crawlers can read it. If you only find an empty container and a script tag, they cannot.

Speed

Google uses Core Web Vitals in its ranking systems. The thresholds are a Largest Contentful Paint under 2.5 seconds, an Interaction to Next Paint under 200 milliseconds and a Cumulative Layout Shift under 0.1. Ranking is only half of it. Visitors leave slow pages before they read a word.

AI tools like to add things that cost speed, a second animation library, three font families, a hero image at full camera resolution. Run every important page through PageSpeed Insights on the mobile setting and fix what it flags in red first.

Structured data

Structured data in JSON-LD tells search engines in machine-readable form who you are, where you are and what a page is. For a local business that means Organization or LocalBusiness with address and opening hours, Article for guides, Product or Offer for prices.

The AI-specific risk is invention. Ask a model for structured data and it may add a rating of 4.9 from 127 reviews that never happened. Markup that claims what the page does not show breaks Google’s guidelines and can cost you rich results. Check every value against the visible page.

Crawlers discover pages through links and use them to judge which pages matter. AI tools build page by page and rarely think about the whole. The result is pages nobody links to, menus that only exist after a script runs, and buttons that are not real links.

Every page needs at least one ordinary link with an href from a related page. Service pages link to the guides that explain them, guides link back to the service. Three clicks from the home page to any page is a good rule.

Real content

This is the part no tool can make up for you. Your prices, your process, photos of your work, the questions your customers actually ask, the street you work in. Google’s quality guidance and every answer engine reward pages that say something specific.

There is one trap that is particular to templates. If you buy a template and keep its sample text, that text is on every other site built from the same template. Replace every sentence of demo copy before you launch. A good template makes that easy by marking what is placeholder, but nobody can do it except you.

llms.txt, honestly rated

An llms.txt is a short markdown file in your site’s root that tells language models what the site is about and which pages matter. It was proposed in 2024 and has spread through many tools. Google is clear that you need no special AI text files to appear in its AI features, and there is little public evidence that assistants rely on the file.

Our view is in the llms.txt guide. It costs five minutes and cannot break anything, and it is no substitute for clear pages. If you want one, our free llms.txt creator writes it from your real subpages.

Being cited by answer engines

Answer engines like ChatGPT search, Perplexity and Google’s AI Overviews quote pages they can find and understand. Google’s page on AI features and your website says there are no additional requirements and no special optimizations needed beyond being indexed and eligible for a snippet.

What helps in practice is a direct answer in the first lines of a page, facts as plain text (what you offer, for whom, at what price, where), and not blocking the search crawlers of OpenAI and Perplexity in your robots.txt if you want to be found there. Answer engine optimization covers this in depth.

The checklist for an AI-built site

Work through these before launch and again after every larger change.

  1. View the source of your home page and one subpage. The main heading and body text are in the HTML.
  2. Every page has its own title of 50 to 60 characters, the search term near the start, and its own meta description.
  3. One page answers one search intent. No two pages compete for the same term.
  4. No mass-generated pages that only swap a town name or a keyword.
  5. Every sentence of template or demo copy is replaced with your own.
  6. Prices, address, opening hours and contact details stand as text on the page.
  7. Structured data matches what the page shows, without invented ratings or awards.
  8. PageSpeed Insights shows green or near green on mobile for your main pages.
  9. Every page is reachable through ordinary links, and a sitemap.xml lists them all.
  10. The robots.txt does not block the search engines and assistants you want to be found by, and no staging noindex is left behind.
  11. Images have meaningful alt text that you have read, not text the AI guessed.
  12. The site is verified in Google Search Console and the sitemap is submitted.

Where AI-built sites typically fail

We see the same handful of mistakes again and again when we check sites with our tools.

The most common is the same title on every page, usually the brand name alone. Close behind come pages whose text exists only after JavaScript, demo text left in place, and a noindex from the test environment that went live with the site. Then come images at several megabytes and structured data with facts nobody can confirm.

None of these is caused by AI as such. They happen because a tool builds fast and nobody checks afterwards. The same patterns also make a site look machine-made to visitors, as our guide on how to tell if a website is AI-generated shows.

How our templates handle SEO

Our AI website templates are static Astro sites. Every page is built into plain HTML before it reaches the server, so every crawler sees the full text whether it runs scripts or not. Each template ships with a sitemap, a robots.txt and structured data in the layout.

What makes the difference for SEO over time is the knowledge that comes with the code. Each template contains a guideline file for the AI, a knowledge folder with an SEO plan for its industry and seven Skills, one of them for SEO. When you ask your AI tool for a new service page, it follows written rules, one search intent per URL, a unique title, no mass-produced city pages, facts checked against the page. That works with Claude Code, Cursor, Codex, GitHub Copilot, Windsurf and other AI coding tools.

The honest limits. A template cannot write your real content, and its demo text has to go. There is no drag and drop editor, and you need an AI coding tool and a hosting account. What the route costs compared with builders and agencies is laid out in what an AI website costs.

Check your own site

You do not need to guess. Our free AEO check reads your website the way a language model reads it, tests whether it answers the six questions assistants look for and shows you a finished paragraph for every gap. The llms.txt creator writes the file from your real pages in about thirty seconds.

Then do the three checks that cost nothing. View the source of your main pages, run them through PageSpeed Insights, and ask two or three assistants the question your customers would ask. What comes back tells you more than any ranking factor list.

Frequently asked questions

No. Google states that appropriate use of AI or automation is not against its guidelines and that it focuses on the quality of content rather than how it was produced. What it does penalize is mass-produced pages made mainly to manipulate rankings, whether a person, a script or an AI wrote them.

Not as a category. Most hosted builders serve proper HTML and handle titles, sitemaps and mobile layouts. The weak spots are elsewhere, in generic copy that every other user of the builder also gets, in limited control over technical details and, for some app builders, in pages that only appear after JavaScript runs.

Only what is in the HTML the server sends. A study of Vercel's network in December 2024 found that the crawlers of OpenAI, Anthropic and Perplexity do not execute JavaScript. If your text is only rendered in the browser, these crawlers see an empty shell. Google's own crawler does render JavaScript.

You do not need one. Google says no special AI text files are required for AI Overviews, and there is little public evidence that assistants rely on the file. It is a short, harmless orientation aid. Write it once from your real pages, then spend the remaining time on the pages themselves.

The same as for any other website. Indexing usually takes days to a few weeks once the site is linked and in Search Console. Ranking for competitive terms takes months and depends on content and links, not on whether an AI wrote the code.

For the design and code, nobody expects it. For articles and guides, Google suggests making clear how content was created where readers would reasonably want to know. More important than a label is that every fact on the page was checked by a person who knows the business.

Keep reading