The downsides of Webflow in 2026
Webflow repriced in May 2026, and the change is a good map of where its ceiling sits. The old Business plan got cheaper and smaller at the same time: the monthly price dropped, and the bandwidth allowance dropped with it, from 100GB to 50GB. Small sites now pay less. Busy sites pay the difference back in add-ons. Webflow publishes the arithmetic on its own help pages, and it is worth reading before your renewal date rather than after it.
The repricing tells you what you are actually buying
The CMS and Business plans merged into one plan called Premium, listed at $25 a month billed yearly. Former CMS customers did well out of it. Their CMS item limit went from 2,000 to 20,000 and collections went from 20 to 40. Former Business customers got the item increase too, 10,000 to 20,000, along with less of everything that gets metered: bandwidth fell from 100GB to 50GB, Webflow Cloud requests from 10 million to 2 million a month, and CPU usage from 120 to 30.
Webflow works one example through in its own FAQ. A Business site on yearly billing using 150GB of bandwidth paid $39 for the plan plus a $20 add-on, so $59. On Premium the same site pays $25 for the plan plus $40 in add-ons, so $65. The traffic did not change and the bill went up about 10%. At 500GB it lands at $199 either way. The headline number went down and the number you actually pay depends entirely on how much traffic you got last month.
The overage rule is where this bites. Go over your limit once and Webflow notifies you without charging. Go over a second consecutive month and the site is upgraded automatically at the start of the next billing month, prorated if you are on annual, and Webflow states plainly that it does not issue refunds for bandwidth overages. That is a defensible policy for a hosting company. What it means for you is that your hosting bill has a floor set by your own success, and the better a campaign performs the more it costs to keep the pages up.
Getting out is narrower than getting in
Code export exists, which is more than most platforms in this category offer. It sits on Workspace plans though, not Site plans. Core at $19 a month billed yearly is the cheapest tier that includes it, so paying for hosting alone does not entitle you to your own HTML. That is a small cost and an easy thing to miss when you are estimating what leaving would take.
What the export gives you is the static shell. CMS content does not come in the ZIP, you pull it out per collection as CSV, and the CMS functionality does not travel with it. Forms, site search and password protection stop working the moment you self host. Anything referenced from outside the Assets panel has to be found and relinked by hand.
The newer features narrow the door further. Webflow's documentation on AI code components says it directly: code components are stripped from code export, and sites containing them cannot be cloned. The React components the AI Assistant generates for you inside the designer are precisely the parts you cannot carry out. That is useful to know while you are deciding whether to build your pricing calculator that way, and much less useful to discover two years later.
The performance gap is real and the designer is not where you fix it
A fresh Webflow template scores well in Lighthouse, because it is an empty page. A production site four years in looks nothing like that. It has interactions layered on interactions, an asset library nobody has pruned, embeds from three vendors, and a script some marketer added in 2024 that no one is willing to remove. From inside the designer you can compress images and defer a few scripts. You cannot change how the page renders, what ends up in the bundle, or when the data is fetched.
Stay World Class came to us on Webflow plus Xano. We moved the site to Next.js, NestJS and Supabase, and Lighthouse went from 55.91 to 91 out of 100 with LCP 77% faster. Very little of that was clever optimisation. We simply had access to the render path, which is the thing a hosted builder holds and does not hand over. The slow part was the data. A four stage ETL mapped every legacy ID onto a PostgreSQL UUID so nothing lost its references on the way across.
None of this was a good trade in 2021, when a rebuild of that size meant months of engineering time you did not have. That is the part that changed. The ceiling has been in roughly the same place for years. What moved is how expensive it now is to build past it.
When staying on Webflow is the right answer
If your site is marketing pages that someone non technical edits every week, and your traffic sits comfortably under 50GB, Premium at $25 a month is very hard to beat. A migration will cost you more than the plan does for several years running, and you will have bought yourself a deployment process in exchange for a problem you did not have.
The same holds if your CMS is honestly just a blog, a careers page and some case studies. Twenty thousand items is a lot of room. And if you have no engineers and no intention of hiring any, moving to code hands you a maintenance burden nobody is there to carry, which is a worse outcome than any ceiling.
The signal that the trade has flipped is fairly specific. You are buying bandwidth add-ons most months. You want behaviour the designer will not express and the workarounds have started stacking. Your roadmap has items on it that begin with the phrase we cannot do that in Webflow. When you recognise two of those three, it is worth pricing the alternative. Until then the platform is doing its job and you should let it.
