October 2026 · measured, not estimated

Which website builders ship sites that Google and AI crawlers can actually read?

A public monthly leaderboard and a free checker for whether a website's text, metadata, and AI-crawler access are visible without executing JavaScript. The number on each row is the median of real sites, fetched and fingerprinted. Nothing on this page is a guessed score.

October 2026 snapshot · methodology 1.0.0

RankBuilderMedianGradeReadableCrawlableDiscoverableFastSampleChange
1WixWebsite builder71.5 / 75A301428—26—
2HostingerWebsite Builder and Horizons69 / 75A301524—26—
3WebflowVisual site builder58.3 / 75B301514.5—24—
4FramerDesign-to-site builder52 / 75C30913—25—
5LovableAI app builder38.7 / 75D3.91521—24—
unrankedBolt betaAI app builder——————0—
unrankedv0 betaAI UI generator——————0—
unrankedDurable betaAI website builder——————0—
unranked10Web betaAI WordPress builder——————0—
baselineWordPressBaseline, not a hosted builder61.5 / 75B301322—14—

Initial public snapshot. Candidates came from public showcase and directory pages, then each homepage was fetched and kept only if its fingerprint matched the builder.

Default builder subdomains are excluded. Parked domains and adult or gambling hostnames are excluded.

PageSpeed Insights did not run (no PAGESPEED_API_KEY). Scores are out of 75: Readable 30, Crawlable 15, Discoverable 30. Fast is blank.

A builder is ranked only when at least 20 sites were measured. Medians within 2 points are a tie. WordPress is a baseline and is never ranked.

This is not a random crawl of the web. HTTP Archive sampling is the planned monthly source once a BigQuery project is connected.

Lovable was remeasured after the fingerprint was extended to the public /~flock.js runtime together with id="root" (and the phrase "Lovable hosting"). The first pass only matched the badge, which most published apps omit.

How to read the board

A builder is ranked only when at least 20 custom-domain sites were measured in the same month. Fewer than that, and the row stays unranked, including every beta builder that does not yet have a reliable public fingerprint. WordPress sits at the bottom as a grey baseline. It is the usual comparison point, and it never receives a rank.

The score has four parts. Readable is 30 points: how much of the visible text is already in the raw HTML, before JavaScript runs. Crawlable is 15: robots.txt rules for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Applebot-Extended, and CCBot, plus a live fetch as GPTBot, ClaudeBot, and PerplexityBot. Discoverable is 30: title, description, canonical, language, a single H1, Open Graph, JSON-LD, a sitemap, and llms.txt. Fast is 25, from PageSpeed Insights. This snapshot did not call PageSpeed, so Fast is blank and the total is out of 75. The missing points are not filled in.

Grades use the percent of points that were measured. A is 85 or above, B is 70, C is 55, and D is below that. Two medians within 2 points are a tie. The change column is blank on the first snapshot, and it stays blank if the methodology version changes. Affiliate programs are not joined. A relationship would not move a score anyway. The methodology lists every test and the sample sources.

Default subdomains such as lovable.app, framer.website, webflow.io, wixsite.com, zyrosite.com, and hostingersite.com are excluded. Parked domains and adult or gambling hostnames are excluded. One origin is kept per site. The sample is a showcase sample, not a random crawl of the web.