None of the three platforms stops a site ranking. Each has a default that designers often leave alone: Framer sites tend to repeat their main heading for each screen size and rarely carry structured data, Webflow sites often miss the global canonical setting, and WordPress sites carry the most code. On mobile speed tests, the Framer sites were the fastest of the three. Choose the platform the team can maintain, then fix that platform's blind spot before launch.
The short answer
The platform matters less than what the person building on it finishes. On every check where one platform looked weak, some sites on that platform got it right. The differences are in the defaults, and defaults are what get shipped when a launch is rushed.
How we checked
We checked 133 live sites whose platform we confirmed from their own code. 64 run on WordPress, 38 on Webflow and 31 on Framer. Most are design and web agency websites: those from the 81 in our agency study that run on one of the three, plus agencies that advertise themselves as Webflow or Framer specialists. Framer sites are harder to find in agency searches, so 21 of the 31 come from Framer's own public gallery, which includes portfolios and startup sites. That makes the Framer group less like the other two, and we say so wherever it matters.
For each homepage we checked whether the text is in the HTML, the title, description, canonical and main headings, structured data, image descriptions, robots.txt, the sitemap, and what a missing page returns. On 13 October we ran one mobile Lighthouse test on 20 sites from each platform, chosen at random.
Side by side
| Check | WordPress (64) | Webflow (38) | Framer (31) |
|---|---|---|---|
| Main text present in the HTML | 100% | 97% | 87% |
| Exactly one main heading | 83% | 79% | 6% |
| Canonical tag on the homepage | 100% | 71% | 100% |
| Any structured data | 97% | 76% | 35% |
| Fewer than half the images described | 30% | 57% | 69% |
| Homepage HTML size, median | 279 KB | 151 KB | 456 KB |
| Sitemap found | 89% | 95% | 100% |
| Missing page returns a real 404 | 92% | 100% | 100% |
| robots.txt names any AI crawler | 17% | 24% | 0% |
| Publishes an llms.txt file | 20% | 55% | 16% |
One reading of each homepage on 12 October 2026. These describe how real sites on each platform tend to look, which mixes what the platform does by default with who chooses it. They are not lab tests of the platforms themselves.
Framer: the same heading, three times
Only 2 of the 31 Framer homepages had exactly one main heading. 21 had more than one, and in 18 of those the extra headings were exact copies, usually two or three of them. That pattern fits Framer building a separate layout for each screen size and keeping all of them in the page, with the ones that do not fit hidden.
Repeated headings are not an error that stops a page ranking. The cost is subtler: the HTML a search engine or AI crawler reads repeats content that a visitor sees once, and the page is heavier for it. The median Framer homepage was 456 KB of HTML before any images, against 151 KB for Webflow. The fix is in the design: use the heading level for one visible heading, and style the copies for other breakpoints as ordinary text.
Framer sites also carried structured data far less often: 35%, against 97% on WordPress. Framer lets you add it through custom code, so this is something to add on purpose, not something the platform prevents.
Webflow: the canonical setting nobody fills in
11 of the 38 Webflow homepages had no canonical tag. Webflow's own help centre explains why: canonical tags are added across the site once you enter a global canonical URL in the site settings, and that field is easy to leave empty. Without it, search engines have to work out for themselves which version of an address is the real one.
Webflow sites were also the least likely to describe their images: on 57% of homepages, fewer than half the images had alt text. Webflow lets you set alt text on each image and in CMS fields, so this is a finishing step rather than a limitation. Webflow sites were the most likely to publish an llms.txt file, at 55%, though there is no published evidence that the file changes what AI tools recommend.
WordPress: the most complete, and the most to carry
WordPress homepages were the most complete on the basics. Every one had a title, description and canonical, and 97% had structured data, much of it the website and breadcrumb markup that SEO plugins typically add. The trade-off is weight and maintenance: a median of 279 KB of HTML, and three of the 64 returned a success status for a page that does not exist, which Google calls a soft 404.
The WordPress risk is rarely the platform. It is the plugins, themes and page builders stacked on it, and whether anyone keeps them tidy after launch.
Mobile speed
On mobile, the Framer sites were clearly the fastest. We ran one Lighthouse mobile test on 20 homepages per platform, chosen at random, and compared the medians.
| Mobile Lighthouse, 20 homepages each | WordPress | Webflow | Framer |
|---|---|---|---|
| Performance score, median | 60 | 52 | 89 |
| Largest contentful paint, median | 4.6 s | 5.3 s | 2.8 s |
| Within Google's 2.5 second "good" mark | 1 of 20 | 1 of 20 | 8 of 20 |
| Over 4 seconds | 12 of 20 | 15 of 20 | 6 of 20 |
| Total page weight, median | 4.5 MB | 6.0 MB | 2.8 MB |
| Main-thread blocking time, median | 258 ms | 672 ms | 58 ms |
That sits oddly beside the HTML sizes above: Framer pages carry the most HTML but the least weight and script overall. It is not only the gallery sites either. The six Framer agency and studio sites in the test had a median score of 91, with four of the six at 90 or above. The slowest group was the Webflow specialists' own sites, with a median score of 47, against 59 for the Webflow sites in our general agency sample.
These are lab tests, one run per site, not real-user Core Web Vitals, and a single run can swing by 20 points. That is why we compare medians across 20 sites rather than any one result. A fast platform does not make a fast site: the quickest and slowest pages on every platform were far apart, from 1.1 to 11.6 seconds on Framer and from 1.7 to 28.9 seconds on Webflow.
"Framer agency" is a claim, not a platform
We searched for agencies that sell Framer websites and checked what their own sites were built on. Of 20 agencies presenting themselves as Framer specialists, only 6 ran their own site on Framer. Five used WordPress, three Next.js, two Webflow and four something else. Of 30 self-described Webflow agencies, 24 used Webflow.
That is not necessarily a problem, and an agency can be very good at a tool it does not use for itself. It is a useful question to ask, though, and it is one more reason to judge an agency by the sites it has shipped for clients.
So which one?
- Choose the platform your team can keep up to date. A neglected site on any platform loses more than the platform choice ever gains.
- On Framer, keep one real main heading per page, add organisation and page structured data through custom code, and write alt text for every meaningful image.
- On Webflow, fill in the global canonical URL before launch, and make alt text a required CMS field.
- On WordPress, choose one SEO plugin, keep the plugin list short, and check that missing pages return a real 404.
- On all three, check the build before launch, not after. That is the point of a launch search handoff.
Limits
- The groups are not identical. Most sites are agency websites, but 21 of the 31 Framer sites are from Framer's gallery, which includes portfolios and startups.
- We checked homepages only, once. Inner pages and templates may differ.
- Platform detection uses each site's own code. A handful of sites in each search could not be classified and are left out.
- Speed figures are lab tests, one run per site, not real-user data. A single run can vary a lot, which is why we compare medians across 20 sites.
Webflow Help Center: set canonical tags
Google: soft 404 errors
web.dev: Largest Contentful Paint
Our own checks of 133 live sites, 12 and 13 October 2026, with Lighthouse 13.4.1