Is Framer good for SEO? What it does and doesn’t do

Framer is still a small part of the web. According to W3Techs, as of 8 November 2026 Framer runs 0.2% of all websites and 0.3% of the sites whose content management system W3Techs can identify.

Framer’s SEO guide, updated on 15 September 2026, lists what comes ready to use: titles and descriptions for each page and each CMS item, automatic image optimization, editable semantic tags for headings and images, structured data added through custom code, and redirects. Framer runs the hosting and delivery; the site owner is left with content and structure.

Framer also serves a markdown version of each page at the same URL to any request that asks for it with the Accept: text/markdown header, or with ?md added to the address. The robots.txt that Framer generates allows every AI and search crawler by default.

Settings describe intent, not results. The HTML of Framer’s own pricing page, downloaded by DoGeo on 8 November 2026 without running any JavaScript, already contained the plan names, prices and limits as text. For a given domain, the free AI crawler test reads its robots.txt and compares what a browser and AI agents receive.

These ceilings rarely bind a corporate or product site. They decide the question for a site that plans to publish thousands of pages, one for each question its customers ask, because on Pro the CMS stops at 40,000 items even with every add-on; beyond that, Framer’s Enterprise plan sets custom limits.

Three Framer patterns that can hide content from search engines, according to Framer’s help center:

• Overlays: menus, modals and “read more” panels built as overlays stay out of the first HTML. A page linked only from an overlay menu becomes an orphan page, with no internal link a crawler can follow.
• Accordions: text revealed by a click can go unindexed, unless the collapsed state keeps the answer on the page with its height and opacity set to zero, the setup Framer recommends.
• Infinite scroll: extra items load with JavaScript. Googlebot renders a window about 12,000 pixels high, so Google sees the items that fit in it; crawlers that do not run JavaScript see none of them.

The fix is the same for all three patterns: put the important links in the footer or in visible text, and keep a plain HTML page that lists every page of the Framer site.

The trade-off is control. According to Framer’s help article on the language attribute, the lang and hreflang tags are managed in the backend and cannot be edited from the interface, so a site that needs a different hreflang setup cannot change it.

Framer moves between pages as a single-page application, loading the next page without a full browser refresh. Its help center explains that hreflang tags only show the right values after a full reload, which is how a crawler requests every URL.

Each AI translation consumes credits from the Framer plan; translating by hand, or adding a locale, never does, according to Framer’s guide to adding a locale.

• Redirects: listed on the Pro plan, with wildcards and capture groups to move whole folders. They only work for paths within the same domain; a whole domain is redirected at the old domain’s provider.
• Structured data: JSON-LD written by hand in custom code, with CMS fields as variables so each item gets its own. Framer’s guide recommends checking it with Google’s Rich Results Test after publishing.
• robots.txt and llms.txt: robots.txt is generated automatically; replacing it, or adding an llms.txt, means uploading static files, available on Pro and Enterprise, up to 50 files on Pro.
• Canonical URL: set automatically from the connected domain. A custom value is only needed when the site is served through a reverse proxy.

What Framer does not hand over is the code. In the words of Framer’s help center: “Framer does not offer HTML export for self-hosting.” A Framer site lives on Framer’s servers, and leaving Framer means rebuilding the site somewhere else.

OpenAI documents three agents. OAI-SearchBot decides which pages can appear in ChatGPT’s search answers; ChatGPT-User visits a page when a person asks ChatGPT something that needs it, the closest signal that a page was read in a conversation; GPTBot only collects content that may be used for training, according to OpenAI’s crawler documentation.

Framer’s own firewall does not cover this. According to Framer’s help center, updated on 1 October 2026, its firewall rules match the visitor’s country and the requested path, need the Advanced Hosting add-on or an Enterprise plan, and do not apply country conditions to AI agents. The log has to come from a CDN or proxy placed in front of the Framer site.

This is what DoGeo sets up for each client: a count of which pages each AI robot visits, and when. The log does not show what the assistant writes in its answer, nor mentions that never needed the site; for that, a fixed set of questions put to each assistant is still needed.

Moving has one known risk: losing the traffic that the old URLs already receive. The protection is a 301 redirect from every old URL to its new equivalent, live on the day of the switch; the website migration SEO checklist walks through it phase by phase.

A comparable case comes from Webflow, another hosted builder. When Kulturo, a tech recruitment agency and DoGeo client, moved from Webflow to its own code on 24 August 2026, with all its URLs redirected by 301, its Google clicks in the following three weeks were 69.2% higher than in the three weeks before, according to Search Console.

Those three weeks coincide with a record in Kulturo’s impressions in Google’s AI answers, so the figure shows that no traffic was lost, not that migrating made it rise. It is Webflow data and says nothing about Framer. The rest of the numbers are in the Kulturo case study.

Anyone still choosing a builder for a new site will find two other options measured on price, SEO, multilingual support and exit in Webflow vs WordPress for SEO.