AI-Ready Builders

Version 1.0.0

Methodology

Every score on this site comes from the same function. The website and the monthly scanner import it. A number is published only after a fetch. If a check did not run, its points are omitted, not estimated.

Readable, 30 points

The homepage is fetched as a browser, then rendered in a headless browser when one is available. The score is the share of rendered words already present in the raw HTML, full marks at 80 percent, plus the H1 being in that HTML and the first rendered lines overlapping the raw text. A thin page is capped. Up to two internal links from the raw HTML are included, with the homepage carrying 70 percent of the weight when three pages are measured.

The scanner uses Playwright. The live checker uses the Cloudflare Browser Rendering binding only when that binding is configured. Without it, a page with at least 200 words and one H1 in the raw HTML scores 27 out of 30, and the result is labeled provisional. An empty client-rendered shell scores near the bottom. That gap is the point of the board.

Crawlable, 15 points

One point if robots.txt returns 200 or is absent (404). One point for each of GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Applebot-Extended, and CCBot when the matching group allows the path. Two points each for a live fetch as GPTBot, ClaudeBot, and PerplexityBot that returns a 2xx and is not a challenge page. A 404 from the live fetch is not treated as an AI block. A 401, 403, 429, or 503 is. Google-Extended is shown on the result and does not move the score.

Discoverable, 30 points

Title, description, canonical, language, a single H1, and Open Graph share 10 points. JSON-LD that parses in the raw HTML, with a recognized type such as Organization or WebSite, is 10. Structured data that appears only after JavaScript runs gets 3. A valid sitemap with at least one URL and a lastmod date is 5. llms.txt that returns text and starts with a level-1 heading is 4, and llms-full.txt is 1 more. llms.txt stays a small weight on purpose.

Fast, 25 points

Fifteen points follow the PageSpeed Insights mobile lab score. Ten points follow Chrome UX Report field data: FAST is 10, AVERAGE is 5, SLOW is 0. If there is no field data, those 10 points follow the lab score. If no PageSpeed API key is set, Fast is null and the score is out of 75. This October 2026 snapshot was produced without a key, so Fast is blank on every row.

Sampling

The builder score is the median of the sites that passed the fingerprint. A builder is ranked only with 20 or more sites. Medians within 2 points tie, and the next rank skips. Month-over-month change is shown only when both months use the same methodology version and both have at least 20 sites. The first snapshot has no change.

Candidates start from public showcase and directory pages. Each homepage is fetched again and kept only if the fingerprint matches that builder. Default subdomains are dropped: lovable.app, framer.website, webflow.io, wixsite.com, zyrosite.com, and hostingersite.com. Parked domains and adult or gambling hostnames are dropped. One site per origin. WordPress is fingerprinted loosely and is never ranked. Bolt, v0, Durable, and 10Web stay in a beta group until a fingerprint and a sample of 20 exist. An empty sample is shown as unranked with no score.

This is not a random crawl, and it is not the HTTP Archive technology report. That report is the planned monthly source once a BigQuery project is connected. Until then the board is labeled as an initial showcase sample. Hostinger's public showcase had fewer than 20 live custom domains on 21 October 2026, so the sample also includes domains from a public Hostinger Website Builder technology index. The builder page lists the skip reasons.

Affiliate relationships never change scores

The scanner does not read an affiliate flag. Programs for Framer, Webflow, Hostinger, Lovable, Wix, Durable, and 10Web are linked as plain URLs and are not joined. When a program is joined later, the link will be labeled as an affiliate link. The median will still be the median of the measured sites.

Changelog

Download

Latest snapshot (JSON) and the same rows as CSV. The files in the repository under data/ are the source. The routes serve the newest month bundled into the site.