What Framer handles for you
These are all working on plenor.io without any custom code:
Server-rendered HTML, so a crawler gets the page text on the first request
A sitemap that updates when you publish
robots.txt, canonical tags and social preview images
A title and description per page, including CMS templates
Images resized and served as AVIF or WebP from a CDN
A plain markdown version of every page, at the page URL plus ?md
That covers most of what an SEO plugin does on WordPress. What it does not cover is below.
Can ChatGPT read a Framer site?
Yes. In Plenor’s October 2026 crawl we fetched 12 AI company homepages built in Framer, using GPTBot’s user agent. All 12 returned full HTML, with 734 to 3,107 words of real text, and first byte arrived in 0.23 to 0.63 seconds. None blocked GPTBot in robots.txt. 7 of the 12 publish an llms.txt file, 5 carry structured data and 5 have exactly one H1.
On plenor.io, every URL returned identical HTML to GPTBot and Googlebot. That is the part that matters: most AI crawlers do not run JavaScript, so a page only exists for them as whatever is in that first response.
The accordion problem
This one caught us. Our homepage FAQ used a closed accordion, and for weeks the eight questions were in the HTML while the eight answers were not. The answers existed only in our schema. The page looked fine in a browser, because a browser runs the JavaScript. A crawler saw questions with nothing under them.

We found it by reading the raw HTML, and no audit tool had flagged it. The fix was to show the answers open, as cards. All eight answers now sit in the HTML.
You don’t always have to give up the accordion. On our services rows we kept it, and collapsed the body to zero height instead of hiding the layer. The text stays in the HTML, the page looks the same, and the video inside only loads when someone opens the row.
If you use an accordion or tabs in Framer, test them. Copy the last sentence of a hidden answer and search for it in the page source.
Schema: Framer has no fields for it
Framer has no built-in schema editor, so structured data goes in by hand. Three things we learned doing that:
Paste it as a page-level snippet. Site Settings, then Code, with the snippet scoped to one page. That puts the JSON-LD in the HTML. A script that injects schema after load is invisible to any crawler that does not run JavaScript.
Check it after you paste. One of our pastes wrapped six URLs in angle brackets, and another page ended up with two FAQ blocks that disagreed. The editor showed neither. View the page source, then run it through validator.schema.org.
Match the visible text. If you change an FAQ answer on the page, change it in the schema too.
Too many H1s
Framer builds each breakpoint as its own layout, so a heading can appear in the HTML more than once. A heading style applied for looks can also end up as an H1 by accident. Of the 12 homepages, 5 had exactly one H1, and the others ranged from none to 15. Our own case study template still repeats its title, so we are not above this.
Most of the time this is cosmetic. It costs nothing to check, though: search the source for <h1 and make sure the one that counts says what the page is about.
Speed: what actually slows a Framer site
Framer resizes your images. It does not touch your video. It serves the exact file you upload, and that is where we lost the most.
Our homepage had a hero video of 8.5 MB and a small card video of 102 MB, shown 453 pixels wide. Here is what happened when we fixed that, on Lighthouse mobile, three runs each, median: main content showed in 7.2 seconds instead of 10.3, and page weight dropped from 50 MB to 10 MB.

What we changed: re-encoded the hero to 2.4 MB, cut the card video to 7 MB by removing the audio track and lowering the resolution, and added a still image that shows while the video loads.
Two fixes we expected to help did nothing. Removing the hero's fade-in made no measurable difference, and neither did showing a still image instead of the video on phones. The other autoplaying videos on the page were competing for the same connection, so that is the next thing to fix. It's also why you should measure more than once: single runs of this same page ranged from 5.4 to 9.5 seconds.
Pages without video were never the problem. Our migration page scores 100 on mobile, with main content showing in 1.2 seconds. The 12 AI homepages we tested ranged from 3.2 to 18.3 seconds on the same test, one run each, with a median of 11.7, and only one came in under 4 seconds.
Where Framer falls short
Framer is not the right tool for every site, and it helps to say so plainly.
Large content libraries. The Pro plan allows 2,500 CMS items and 10 collections. Hundreds of posts are comfortable. Thousands are not, and a blog that size should stay on WordPress behind a proxy.
No plugin layer. There is no Yoast or Rank Math. Titles and descriptions are built in, but schema and anything beyond that is done by hand.
Redirects depend on the plan. They come with Pro, so budget for it if you are moving a site with existing rankings.
Teams that publish every day. If several writers publish daily with editorial workflows, a CMS built for that will suit them better.
Does moving to Framer hurt rankings?
Not if every old URL is redirected before the switch. Rankings usually wobble for one to two weeks while Google recrawls, then settle. Sites lose traffic when they move without a redirect map, and that would happen on any platform.
We cover the details in WordPress to Framer and Webflow to Framer, including what to do about plugins, blog imports and old permalinks.