Skip to content

Research · October 29, 2026

We read 6,692 sites on four builders. 36.6% of Webflow homepages have no canonical tag.

Every Webflow, Framer, Wix, and Squarespace site we could find by DNS in the top 1,000,000 sites, read on October 29. Each builder leaves a different gap. And 96% of Wix sites now point AI agents to their own MCP endpoint.

36.6%
of Webflow homepages have no canonical tag
Webflow adds one once you set it
96%
of Wix sites point AI agents to an MCP endpoint
In an llms.txt file Wix writes
32.7%
of Squarespace homepages have no meta description
135 of 413 sites
2.7%
block any AI bot
Across all four builders

The four builders side by side

WebflowFramerWixSquarespace
Sites studied4,4298271,023413
No canonical tag on the homepage36.6%0.2%0.4%0.2%
No meta description on the homepage3.9%0.6%18%32.7%
No <h1> in the homepage HTML4.9%11.1%30.7%23.2%
Words in the homepage HTML (median)1,0891,328414498
No sitemap at /sitemap.xml12.7%0.5%4%4.6%
Serve an llms.txt file25.4%22.9%96.7%1.7%
Block at least one AI bot3.8%0.7%0.1%3.4%
1 · Webflow

The canonical tag is a setting.

1,622 of 4,429 Webflow homepages have no canonical tag. Webflow writes one on each page once the site has a Global canonical tag URL, under Site settings, then SEO.

Without one, Google picks the main version itself. Search Console shows that as "Duplicate without user-selected canonical" when it finds copies. 12.7% of all Webflow sites also serve no sitemap at /sitemap.xml.

Webflow's guide to canonical tags

What one line in the homepage looks like once it is set

<link href="https://www.moloco.com/" rel="canonical"/>

moloco.com, a Webflow site, October 29, 2026

2 · Wix and AI agents

Wix adds an llms.txt to almost every site.

llms.txt is a proposed plain-text file, from llmstxt.org, that lists a site's main pages for AI tools. 96.7% of Wix sites serve one, and 982 of them point AI agents to the site's own MCP endpoint at /_api/mcp. MCP (Model Context Protocol) lets an AI agent read a site's live content without scraping it.

On the other builders the file is a choice: 25.4% of Webflow sites, 22.9% of Framer sites, and 1.7% of Squarespace sites serve one.

ukrainealarm.com/llms.txt, a Wix site
## AI Agent Access

This site is powered by Wix and supports the
Model Context Protocol (MCP) for agentic AI access.

- [Site MCP Endpoint](https://www.ukrainealarm.com/_api/mcp)
3 · JavaScript and staging domains

Little sign of trouble from JavaScript or staging domains.

All four builders send the page text in the HTML. Only 1.6% of homepages had fewer than 50 words before any JavaScript ran. Google says server-side rendering helps because not all bots can run JavaScript.

No homepage's canonical tag pointed to a staging address such as .webflow.io or .framer.app, and 2 sitemaps listed one. The staging copy can still be indexed on its own: Webflow's Staging indexing setting, under Site settings, then SEO, controls it.

All four buildersShare
Homepage under 50 words in the HTML1.6%
Median words in the homepage HTML965
Canonical tag points to a staging address0%
Sitemap lists a staging addressunder 0.1%

Check your own site

  1. 01

    Inspect your homepage in Search Console.

    In URL Inspection, check User-declared canonical. If it says None on Webflow, set the Global canonical tag URL under Site settings, then SEO, and publish.

  2. 02

    Open Search Console, then Pages.

    Look for "Duplicate without user-selected canonical". Each one is a page Google treated as a copy of another. Open it and check which URL Google chose.

  3. 03

    On Squarespace or Wix, open each page's SEO settings.

    Add a description where it is empty. 32.7% of Squarespace and 18% of Wix homepages have none.

How we did it

All our studies

Which sites

We looked up the DNS of every domain in the Tranco list of the top 1,000,000 sites (list ZJ8PG). 8,681 pointed at Webflow, Framer, Wix, or Squarespace servers. We kept the 6,873 whose homepage carried that builder's code, and counted domains that redirect to the same site once: 6,692 sites.

Sites behind their own proxy, such as Cloudflare, do not show the builder in DNS and are not included.

What we read

For each site, on 2026-10-29: robots.txt, the homepage, the sitemap and up to 20 of its files, and /llms.txt. Words in the HTML count the text in the page as sent, before scripts run.

We read public files only. They show what a site tells Google and AI crawlers, not what Google has indexed.

Our crawler

Requests came from CrawlCoachBot. We skipped the 39 sites whose robots.txt disallows it.

See which of your pages Google left out.

CrawlCoach's free report reads your Search Console: up to 500 pages, which ones Google left out and why, and the 3 to fix first.

Get my free report

Free for up to 500 pages. No card needed. Sign in with the Google account you use for Search Console.