SentientUI
Your site adapts to each kind of visitor — and shows you the lift against your original.
SentientUI can write versions of your page for each kind of visitor, in your own design, and moves traffic toward the ones doing better. Some visitors always stay on your original page, so when a version is not helping, you will see it.
Improving for everyone from day one. Separate versions for a particular kind of visitor unlock only once your own traffic shows they behave differently.
Free forever · No credit card · 5,000 visitors a month
Not ready? See your site's first impression — what visitors and AI agents actually see.
Works on the site you already have
- Webflow
- WordPress
- Squarespace
- Shopify
- Framer
- Plain HTML
- React
- Next.js
Who sees what
Every kind of visitor gets their own version of your page.
Not a new page, and not a redesign — the page you already have, with each section wearing the version that works for whoever is reading it. You can see exactly who gets what, and a slice of traffic always stays on the original so you can trust the number.
Inside your design
Different kinds of visitor on the same page each get their own version of a section — your fonts, your colours, your layout. Not a redesign, not a new page. If you saw two of them side by side, both would look like your site.
You see who sees what
The dashboard shows a simple grid: groups of visitors down the side, the parts of the page allowed to adapt across the top. Everyone starts in one row, working from day one. Other rows say plainly where they came from — a group you told us about, one we think we can see forming, or one your traffic has proven — so you always know which are real. Where a row has no version yet, SentientUI can write one from your page’s own content and offers, on brand, never an invented discount, with a plain-language note on why.
A slice stays on the original
Some of your visitors are always held out on the page exactly as you built it. The comparison against them is the proof — not a win-rate tile, and not the winning version marking its own homework.
Illustration · your page, by who is reading it
| Headline | Plans | Customer proof | |
|---|---|---|---|
| EveryoneWhere every visitor starts | Automate the busyworkLive | Your original | Logos above plansLive |
| Small teamsA group you told us about | Free for teams under 10Live | Monthly plan firstWritten for them · waiting for your OK | Your original |
| Buying for a companyBeing checked — not served yet | Your original | Your original | Your original |
Every row says where it came from. A group only gets its own versions once your traffic shows it behaves differently — until then it sees what everyone sees.
Three jobs, one system
It is not just for your homepage.
The same thing that improves a landing page improves a checkout, and improves the screens people see after they log in. One place to run all three.
Find the step where people leave
Lay out the journey you care about — pricing page, sign-up, first project — and see which step loses people. Then the pages before it start fixing themselves.
- Drop-off at every step, with the worst one called out
- Each step teaches the pages above it, not just the report
- Works for a trial sign-up, a demo request, or a checkout
Where people drop out
−26 in every 100 leave here
−19 in every 100 leave here
▼ biggest drop — 34 in every 100 leave here
More people are lost inside the sign-up form than anywhere before it. Now the pages leading into it start working on that.
Learn what to put at the top
Past the login screen the question stops being which headline and becomes which order. Different people read your screens in a different order — this tells you what each of them actually looks at.
- What each kind of visitor reads, and what they scroll past
- Section order ranked per audience, not one order for everyone
- Start with one screen that matters, not the whole app
What each visitor reads first
Team admin
New teammate
Same page, same four sections, opposite order. One order for everyone is a compromise nobody asked for.
Count the money, not the clicks
A £200 order and a £20 order are not the same win, and most tools score them as if they were. This one knows the difference.
- Real orders from Shopify or your own checkout
- A refund is taken back off the score, not left flattering the total
- What an average visit is actually worth, per version
Revenue per visit
£3.44
per visit · Version B
the most-clicked
version earns £2.05
B is chosen by fewer people and still makes the most money. Score on clicks and you would have shipped A.
Assistants and crawlers are read as machines, not visitors: agentic traffic is detected, kept out of your learning and your bill, and reported back as a share of who is reading you.
The same page can't be right for everyone who lands on it.
A small team wants to see the free plan. Someone buying for a whole company wants security and a comparison first. Here is what that looks like — click through three made-up visitors and watch the page rearrange. On your site, we start by improving it for everyone, and a group only gets its own version once your traffic shows it behaves differently.
Flowbase
Workflow automation for growing teams
Two-minute product tour
2:04
Improving the page for everyone first
You have probably looked at one of these.
Here is where each of them fits, including where they beat us. We would rather you knew now than found out three months in.
Switches that turn features on and off for chosen people
Honestly a different job, not a rival. They decide who gets a feature. We work on how the page around it is put together. Plenty of teams run both.
One test at a time: pick a question, wait, read the answer
Every new idea joins a queue, and while you wait, half your visitors keep seeing the version that is losing. When you have one big question and need one clean number, though, this is still the better tool.
Rules you write by hand: if the visitor is X, show them Y
Rules quietly go out of date and nobody is ever tasked with rewriting them. Ours keep updating on their own, and keep working past the login screen. What they do better: naming one specific company and showing it a page written for it.
AI writes new versions of your page, and a test picks the best one
The closest thing to us, and genuinely quick at producing lots of ideas — Coframe will write new code and images too, with its own team running it for you. The difference is what they are looking for: mostly one better page for everyone. We keep a version for each kind of visitor, change only the parts of your page you allowed, and stay inside your design.
Each kind of visitor gets their own version of the page. It keeps learning from there.
The others look for one better page for everyone. This finds a better page for each kind of visitor and keeps every version alive — no global winner that flattens the rest. And a group of visitors always stays on your original page, so the improvement is something you can check rather than something we assert.
What it actually does, all week, without you.
It starts as soon as visitors arrive. Knowing for sure that it helped takes longer, and we will tell you how much longer instead of guessing.
Write
Versions you did not have to write
Write two or three yourself, or let SentientUI draft them from your page’s own words and offers — in your fonts and colours, never an invented discount. If you would rather check first, every draft waits for your OK before anyone sees it.
Shift
Traffic moves toward what is working
As visitors sign up or buy, more of them get the versions that are doing better and weak ones quietly step back. It does not wait for a test to finish. There is no single winner everyone gets — each kind of visitor keeps the version that works on them, and you can lock one in yourself.
Check
You find out if it helped — or didn’t
Some visitors always see your original page, so there is something honest to compare against. We tell you how long a clear answer will take at your traffic, even when that is months. If a version starts doing harm, it is pulled automatically and you get an email.
Two ways in. Same product behind both.
It does not matter whether you can edit the code. Pick the one that sounds like you.
You are judged on signups or sales, but you cannot deploy. Paste one line into Webflow, Framer, WordPress or Shopify, then write your options in the dashboard by clicking the thing you want to change. No ticket, no waiting on a developer, nobody to ask whether the result counts.
One install and your app is set up from day one. Wrap the part of your pricing page or app you want to improve, say which action counts as a win, and the same system runs on your marketing site and inside your product. Your own analytics keep working alongside it.
The questions people actually ask.
Including the awkward ones. If yours is not here, the contact form at the bottom reaches a person.
Not to get started. You paste one line into Webflow, WordPress, Framer, Shopify or a plain HTML page, then click the headline or button you want to try differently. If you do have developers, they can install the React package instead and go much deeper — it is the same product either way.
A/B tests have an ending. You pick one question, wait weeks for enough traffic, read the result, and start again. This does not stop: it keeps nudging traffic toward whatever is working, on lots of small things at once, and it can treat different kinds of visitor differently. Where a classic A/B test is still better is when you have one big question and you want one clean number for it. We would rather say that than pretend.
Because we keep a group of your visitors on your original page the whole time — about one in ten — and compare. That comparison is the number we show you, and if a version is not helping, it shows that too. We will not point at the winning version and call its own conversion rate the improvement, because a version that got picked for looking good will always flatter itself.
You write them yourself — or, for a kind of visitor that has no version yet, SentientUI can write one for you. It works from your page’s own content and offers, stays on brand, and never invents a discount or makes something up. Every version sits in the dashboard where you can read why it was written and how it is doing, in plain language, and remove it if it does not sound like you.
The versions can be live on day one, and traffic starts moving toward the better ones as soon as visitors sign up or buy. On a normal website nobody changes mid-visit — early visits teach us, and returning visitors (plus new ones who look similar) get the better version. If people log in, we know who they are straight away and can skip that wait. A clear answer on whether it helped overall takes longer, and how long depends on your traffic — the dashboard tells you, in plain numbers.
No fingerprinting, nothing shared with other sites, no profiles sold anywhere. Sessions get a random ID and the signals stay anonymous. If you use a cookie banner, you can hold everything back until someone says yes.
New visitors stop being counted until the month turns over. Anyone already on your site finishes normally, and everything learned so far stays. The counter resets on the 1st.
How it works
Three things to decide, then you are done.
1
What counts as a win?
A signup, a sale, someone finishing setup. Pick the thing you would actually be pleased about. Everything else follows from it.
2
What is allowed to change?
Write two or three versions of a headline, a button, or the order of your sections — or leave a gap and let SentientUI write a version for a visitor type that has none, from your page’s own content. Nothing invented, everything yours to review.
3
Come back and look
Traffic has been moving toward the versions that work the whole time. Next to that you get one honest number — how those visitors did against the ones kept on your original — and a plain estimate of how long until that number is solid.
Getting it on your site:
Paste one line
Works on Webflow, WordPress, Framer, Shopify and plain HTML. After that you open your own site with our editor on, click the words or button you want to try differently, and type the alternatives.
Install the React package
One npm install, then wrap the part of your page you want to improve. Your original stays in place as the fallback and the control, and the versions are written from the dashboard — no redeploy for each change.
Both end up in the same dashboard, measured the same way. The integration docs cover both.
You already paid for the traffic.
Pick one page that matters and tell us what a win looks like on it. Write a couple of alternatives, or let SentientUI write them from your own page. Traffic moves toward what works, and some visitors stay on your original so you can see, honestly, whether any of it did.
Not sure you have the traffic for it? Check before you sign up — we will tell you if you don't.
Questions? Send us a message
Support, sales, or a developer question — we reply within one business day.