June Offer Every MAX plan gets a fully custom-built system Free custom system worth $1,500-$10,000 · worth $1,500-$10,000

A website that tells you who is on it

Describe the page and the AI builds it. Then a step that is not AI at all wires in the pixel, the session recordings, Google Analytics, the Meta Conversions API, your real forms and your booking page, on every single build.

So the visitor stops being a number in a chart and becomes somebody you can ring.

Get Started

The AI writes the page. Something else wires it.

Every AI builder can produce a good-looking page. The problem starts the day after, when you discover the page has no analytics on it, or the pixel is there on the home page and missing on the one the ad points at, or an edit last Thursday quietly removed it.

You can ask a model to include the tracking. It will, most of the time. Most of the time is a terrible property for the thing that tells you whether your advertising works.

So after the AI finishes, and on every single job including small edits, a set of ordinary deterministic steps runs over the output and guarantees the parts you cannot afford to have missing. No prompt, no model, no maybe.

  1. The pixel

    Every page gets the tracking script, so heatmaps, scroll depth, rage clicks and session recordings work from the first visit rather than from whenever somebody remembers to add it.

  2. Meta pixel and GA4

    If you have connected an ad account or an analytics property, the snippets go into the head of every page, in the same place, exactly once. Re-running a job updates them in place rather than adding a second copy.

  3. Your form and your booking page

    Embedded as your real, live assets rather than a lookalike, so a submission is a lead in a pipeline stage with your qualifying rules applied, not a notification email.

  4. The chat widget

    The genuine loader for your own chat, keyed to your account, baked on every build.

  5. Per-page SEO

    Title, description and social preview tags per route, in one guarded block, with the duplicates the AI produced stripped so there is exactly one of each.

  6. The performance work nobody enjoys

    Width and height written onto every local image so the layout stops jumping, the hero image loaded eagerly at high priority, everything below the fold lazy-loaded, sitemap and robots generated.

  7. Security headers

    A deliberately minimal policy: no plugin injection, no base-tag hijack, clickjacking blocked, strict transport and the standard hardening headers, chosen not to break third-party embeds on a page nobody hand-reviewed.

Why it works this way

This design came out of a real failure rather than a whiteboard. The visual editor needs a stable identifier on each block to know what you are dragging, and the model was asked in its instructions to add them. It did, inconsistently. Some sites came out with sixty. Many came out with none, and on those the editor silently did nothing at all when you tried to move something.

The lesson generalises. Anything that must be present every time does not belong in a prompt. It belongs in a step that runs afterwards, checks, and fixes. Everything on the list above is now that kind of step, which is why an edit six months from now cannot quietly take your conversion tracking offline.

The short version

  • Describe a site or a landing page and the AI builds it.
  • Point it at a page you like and it rebuilds against that design, then checks its own work against the original.
  • Every build bakes in the pixel, GA4, the Meta pixel, your form, your booking page and your chat widget.
  • Heatmaps for clicks, scroll depth and cursor movement, plus rage clicks and live visitors.
  • Session recordings, with typed input masked by default.
  • When a visitor identifies themselves, their earlier anonymous visits fold into their contact record.
  • A known lead returning to the site can start a workflow while they are still reading.
  • Conversions go to Meta server-side as well as from the browser, deduplicated.
  • Campaign, source and click ids follow the person all the way to the sale.
  • Hosted video reuses the same visitor identity, so watch time joins the same session.
  • Split-test pages, including pages hosted somewhere else.
  • Export the whole site as files or into your own GitHub.
  • The pixel also works on a site we did not build.

What the pixel actually sees

One line, one opaque id, the same shape as a Meta pixel. It works on a site built here and on a site built anywhere else, and the snippet never contains your account number because the account is resolved on our side from the id.

Three heatmaps, and they answer different questions

Clicks tell you what people tried to press, including the things that are not buttons. Scroll depth tells you how much of the page exists as far as a visitor is concerned, which is usually far less than you think and settles most internal arguments about section order in one screenshot. Movement shows where attention lingered.

Rage clicks

Somebody clicking the same thing repeatedly is a person being failed by your page in a specific spot. It is the highest-signal, lowest-effort report in the whole category: a short list of exact places where visitors got angry, most of which take ten minutes to fix.

Session recordings

A reconstruction of the page and what somebody did on it, played back. Watching five of these teaches you more about your own site than a month of dashboards, because you stop guessing why the number is low and simply watch it happen.

Who is on the site right now

A live list of current visitors. On its own that is a novelty. Combined with the identity part below, it is the difference between reading a report on Friday and ringing somebody on Tuesday while they are on your pricing page.

Ad cohorts and AI session analysis

Behaviour grouped by which ad brought people in, so you can compare how two campaigns behave once they land rather than only how they perform in the ad account. Two campaigns with the same cost per click can produce completely different sessions, and that is usually where the difference in sales comes from.

A check that it is actually installed

Give it your address and it fetches the page and confirms the script is really there. Small thing, and it prevents the classic month of wondering why the dashboard is empty when the snippet went into a staging template.

All of this has its own page, with the frustration signals, what is masked in a recording, and an honest comparison against Hotjar, Microsoft Clarity and Crazy Egg: website heatmaps and session recordings.

What is recorded, what is hidden, and what we refuse to do

Behaviour analytics is a category with a bad habit of glossing over this, so here it is plainly.

Typed input is masked

A recording is a reconstruction of the page, not a video of a screen. Input fields are masked by default, with passwords, email addresses and phone numbers masked specifically. Canvas contents are not captured. Cursor movement is sampled rather than traced continuously, which keeps the recording small and keeps you from accidentally building a keystroke logger.

The recorder is not on every site

The base pixel is small. The recording library is only downloaded at all when a project has recordings switched on, and it is served from our own domain rather than a third-party script host, so you are not adding another origin to your page for visitors who will never be recorded.

It respects accessibility settings

A visitor whose system asks for reduced motion does not get cursor tracking. That is a small courtesy that costs a fraction of one report and is the kind of thing an audit notices.

Your account id is not in the snippet

The tag on your page carries one opaque identifier and nothing else. Which account it belongs to is resolved on our side. Anyone reading the source of your site learns nothing about your account structure, which matters to agencies whose snippets sit on every client site they have ever built.

We do not de-anonymise cold visitors

There is a whole category of product that resolves an anonymous visitor to a named person or company using identity data bought elsewhere. We are not one of them, and everything in the next section works only for people who told you who they are.

A visitor is anonymous and analytics-only until they submit a form or book a call. That is the line, and it is drawn in the code rather than in a policy document.

The rest is disclosure, and it is yours. Behaviour analytics belongs in your privacy policy, and in Europe behind your consent banner. We have built this so you can do that properly. Any vendor who tells you their script is automatically lawful everywhere is selling you a future problem rather than a feature.

The moment a session becomes a person

This is the part that is genuinely not available anywhere else on this page, and it happens in one instant: somebody who has been reading your site for three weeks fills in a form.

At that moment every one of their earlier anonymous visits stops being anonymous. The pages they read, the campaign that first brought them, the referrer, how many times they came back before saying anything, all of it folds into their contact record as their history. The first visit is marked as the first touch, so you can see what originally found them rather than only the last click before they converted.

From then on it works forwards. Every visit lands on their timeline. If they come back after a week of silence it is flagged as a return, which is a different and much more interesting event than a first visit, because a quiet lead reappearing on your pricing page is a buying signal and almost nobody notices it in time.

And a visit from a known lead is a workflow trigger. So the closer gets a notification, or a message goes out, or the deal moves, while the person is still on the page rather than the following Monday. That is the whole reason to record anything.

The timeline it writes to is the one on the deal card, the identification usually comes from a form or a booking, and what happens next is built in workflows.

Telling Meta and Google what actually happened

If you buy traffic, the accuracy of your conversion reporting is not a reporting problem. It is a targeting problem, because the algorithm spending your money learns from whatever you managed to tell it.

Server-side as well as browser

A browser-only pixel loses a real share of conversions to blockers, privacy settings and people who close the tab a moment too early. Conversions here also fire from our servers, sharing an identifier with the browser event so Meta counts each one once rather than twice.

The whole path, in order

Each event is stored against the page it happened on and its position in that visitor session, so a journey across four pages can be reconstructed afterwards rather than collapsing into a pile of undated conversions.

Click ids that survive the journey

Campaign, source, medium, content, term and the Meta and Google click identifiers are captured on arrival, kept for ninety days, and attached as hidden fields to every form on the page automatically. So the ad that produced a customer is still knowable when they buy three weeks later, which is exactly when everybody else has lost it.

Google Analytics, connected properly

Connect the account, choose the property, and the tag goes onto every page on every build. Analytics is free and better than us at traffic reporting, so we do not try to replace it. It answers how many and from where. The pixel here answers who, and what they did.

Within the consent you were given

Where a conversion comes from one of our forms, the consent the visitor gave decides what is sent. Without consent to process their data, nothing identifying leaves, and depending on what they did agree to the event either drops to an anonymous count or is not sent at all. That is behaviour rather than a setting somebody can forget to switch on.

The sales video, on the same identity as the page

Put a sales video on a landing page using an ordinary video host and you have created a second, disconnected analytics product. The video tool knows somebody watched four minutes. Your site analytics knows somebody read the page. Nothing knows they were the same person, so the most interesting number on the page is unavailable.

Video hosted with us reuses the same visitor identity as the page, so watching is part of the same session as reading, and it folds into the same person the moment they identify themselves. Engagement is recorded second by second rather than as a percentage, which is what lets you see the exact point where people leave rather than an average that hides it.

It is worth saying what that unlocks, because it sounds like a technicality and is not. A lead who watched eleven minutes of your sales video and then went quiet is not the same lead as one who bounced at twenty seconds, and until the video and the CRM share an identity nobody can tell them apart. Watch time is one of the strongest buying signals most businesses collect and then throw away.

Video engagement is becoming a feature area of its own and will get a page of its own. If the video sits inside a multi-step sales sequence rather than on a single page, that is the funnel studio, which owns the opt-in, video, booking, checkout and upsell chain.

Building the thing in the first place

Everything above is about what the page does. This is how it gets made, and there are a few decisions in here that are not obvious.

The model is chosen per job, not per account

Building or cloning a whole site is one decision that shapes everything after it, so it runs on the heaviest model available. Recolouring a button runs on the fastest. Anything structural or creative goes heavy again, and anything ambiguous takes the safe middle. You are not asked to understand model names, and you can always override it on a job you care about.

Cloning checks its own work

Point it at a public page and it screenshots it, extracts the colours, fonts and section structure, and rebuilds against those. Then the useful part: it screenshots what it built and the original side by side, lists every way they differ, fixes those, and repeats. A one-shot looks-nice check waves through a pretty page that is nothing like the one you asked for.

A design gate on builds from scratch

When there is no reference to match, the result is screenshotted and scored on how it actually looks, and a polish pass fires automatically if it falls below the bar. It is a floor rather than a guarantee of taste, but it removes the genuinely ugly first drafts before you ever see them.

An SEO check with real search volumes

Titles, descriptions, heading structure, canonical tags, image alt coverage and word count, plus candidate phrases enriched with real Google search volumes rather than a made-up score out of a hundred. Page speed scores come from the measurement Google itself publishes.

Your own data, with the safe key only

Connect your own database as a content source and build pages from your rows. Only the publishable key is accepted and secret keys are rejected outright, so what we can read is exactly what your own security rules already allow the public to read. Nobody should hand a website builder a key that bypasses their own protections, and this one will not take it.

You can take it with you

Download the site as files or push it into your own GitHub repository. Check this before buying any website builder, including this one: several large ones in this market cannot hand you a working copy of your own site, and you only find out at renewal.

What this does not do

This is a crowded category with several genuinely excellent free products in it, so the honest list matters more here than usual.

  • It does not identify anonymous visitors. No names, no companies, no matching against bought identity data. A visitor is anonymous until they tell you who they are.
  • It is not a better recorder than Microsoft Clarity, which is free forever with no traffic limit and has far more ways to filter and segment what it captured. If recordings are the entire job, install Clarity.
  • No surveys or on-site feedback widgets. Hotjar has those and we do not.
  • It is not as good a builder as Webflow or Framer. Less design control, no comparable CMS, no component ecosystem.
  • No storefront, no inventory, no shopping cart. Payment here happens on an invoice or a quote against a deal.
  • A handful of templates rather than thousands, and no app marketplace.
  • It does not replace Google Analytics and does not try to. Analytics is free and better at traffic reporting; the two answer different questions.

Who this is built for

Anyone running paid traffic to a landing page

The page is where the money either works or disappears, and the usual setup gives you a conversion count and nothing else. Here you get the recording of the visit that did not convert, the scroll map that shows nobody reached the offer, the server-side event that teaches the algorithm properly, and the click id that still knows which ad produced the customer three weeks later.

The ad-account side of that is in ads reporting.

Agencies building sites for clients

A site per client in its own workspace, with the right pixel, the right form and the right booking page baked in, and none of it depending on whoever built it remembering. Clone a site you are proud of as the starting point for the next one, and hand over an export when a relationship ends rather than arguing about it.

The reporting conversation also changes shape. Instead of sending a traffic chart, you can show the client a recording of a real visitor failing to find the phone number, and then show them the same page working the following month. That is a renewal conversation rather than a report.

More in the agency use case and what we build for agencies.

Coaches, consultants and high-ticket sellers

When a sale takes weeks and one conversation, the useful signal is not traffic. It is that a specific person watched most of the sales video, came back twice, and spent four minutes on the pricing page last night. That is knowable here and is the single best reason to call somebody today rather than next week.

See how coaches use Inflowave.

Small and medium businesses, including local and service trades

Salons, clinics, studios, trades, dealerships, firms of four people. The website was built once, by somebody who has since stopped answering, and nobody has any idea whether it works. The honest first win here is not the AI. It is finding out, in an afternoon, that half your visitors never scroll far enough to see the phone number.

A business this size does not need a site rebuilt every quarter. It needs one page that works, an enquiry form that produces a person to ring rather than an email nobody opens, and to know when a customer who asked for a quote last month is back looking at prices. All three are on this page, and the third one is the one that actually makes money.

You also do not have to move your website to get most of this. If you have a site you are happy with, add the one line and keep it. Rebuild later, if ever.

See how small businesses use Inflowave and what it costs for a small team.

Setting it up

  1. 1. Connect the accounts before you build

    Ad account and analytics property first. Everything after this point bakes them in automatically, and doing it in this order means the first version of your page is already instrumented rather than the third.

  2. 2. Build the form and the booking page first

    They are what gets embedded, and they are what turns a visit into a person. A beautiful page with a placeholder form is a page that generates nothing.

  3. 3. Describe the page, or point at one

    Say who it is for, what you are offering and what you want people to do. Or give it a public page as a design reference and let it match that. Both take one job.

  4. 4. Publish, then leave it alone for a week

    Nothing useful can be learned from a page with nine visits, and the temptation to redesign on day two is strong and always wrong. Let the heatmaps and the recordings collect something real.

  5. 5. Watch five recordings before changing anything

    Five is enough. It is a genuinely uncomfortable half hour and it will change what you were about to do, usually from a redesign to a two-line fix. Check the rage clicks in the same sitting.

  6. 6. Build the return-visit automation

    This is the step nearly everybody skips and the one with the best return. A known lead coming back to the pricing page should tell somebody, now. Set it up once and it will pay for the subscription on its own.

Against the builders, and against the analytics tools

Nobody compares these two families to each other, because normally you buy one of each. That is the point. Every row below is about what you have after you publish, and each column wins some of them.

Website and behaviour analytics capability by platform
After you publishInflowaveWix / SquarespaceWebflow / FramerHotjar / Clarity
Design control and component depthPartlyPartlyYesNo
Template library and app marketplaceNoYesYesNo
Full ecommerce store, inventory and checkoutNoYesPartlyNo
Session recordings and heatmaps at allYesNoNoYes
Recordings included with no session allowance to run outYesNoNoPartly
Filters, segments and frustration signals on recordingsPartlyNoNoYes
The recording is attached to a named person in a CRMYesNoNoNo
A return visit can start an automationYesNoNoNo
Meta pixel, Conversions API and GA4 wired at build timeYesPartlyPartlyNo
Forms and booking embedded as your own live assetsYesPartlyPartlyNo
Video engagement on the same visitor identity as the pageYesNoNoNo
Deterministic SEO, performance and security passes on every buildYesPartlyPartlyNo
Export the site as files, or to your own GitHubYesNoPartlyNo
FreeNoPartlyPartlyPartly
  • Design control and component depth: Webflow and Framer are the better tools here and it is not close. If a designer is going to live in this every day, that is where they should live.
  • Template library and app marketplace: Thousands of templates and hundreds of apps against our handful. If you want to browse a gallery until something feels right, we are the wrong product.
  • Full ecommerce store, inventory and checkout: We do not do storefronts. Payment here happens on an invoice or a quote against a deal, not in a shopping cart.
  • Session recordings and heatmaps at all: The builders do not do this; you install a third-party script. Clarity does it free with no traffic limit, which is a genuinely hard offer to beat on its own terms.
  • Recordings included with no session allowance to run out: Clarity is free forever with no traffic limit. Hotjar moved to monthly session allowances under Contentsquare, and their documentation is clear that when you hit the allowance data collection stops until the reset, account-wide, with no rollover. Losing a month of data during your best month is the failure mode worth knowing about.
  • Filters, segments and frustration signals on recordings: Clarity is deeper on filtering: 40+ filters and saved segments against our project, date range and device. On the frustration signals themselves it is closer than that sounds, since we track rage clicks, dead clicks, quick backs, form abandonment and JavaScript errors, but the ability to slice them is theirs.
  • The recording is attached to a named person in a CRM: This is the row the page exists for. When somebody fills a form or books a call, their previously anonymous sessions fold into that person’s history and every later visit lands on their timeline. Hotjar and Clarity can attach an identifier you supply, but neither is a CRM, so there is nobody to ring.
  • A return visit can start an automation: A known lead landing on the site fires a workflow trigger, so the closer can be notified, or a message sent, while the person is still reading.
  • Meta pixel, Conversions API and GA4 wired at build time: The builders let you paste a pixel id or add a script. What is different here is that a deterministic post-build step bakes them into every page on every job, so the AI cannot forget, duplicate or move them, and the server-side conversion event is already wired rather than being a developer task.
  • Forms and booking embedded as your own live assets: The embed points at your real form or booking page, so a submission is a lead in a pipeline stage rather than a notification email, and the same qualification rules apply everywhere.
  • Video engagement on the same visitor identity as the page: A hosted video reuses the page pixel identity, so watching two minutes of a sales video is part of the same session as reading the pricing page, and both fold into the same person.
  • Deterministic SEO, performance and security passes on every build: Per-page titles and Open Graph tags, image dimensions written in to stop layout shift, the hero image prioritised and the rest lazy-loaded, sitemap, robots and security headers. Baked after generation on every job rather than being something you remember to check.
  • Export the site as files, or to your own GitHub: Webflow exports static code on some plans but not CMS-driven sites. Wix and Squarespace do not export a usable site at all. This matters more than people expect at renewal time.
  • Free: Clarity is genuinely free forever with no traffic limit. Everyone else has a free tier with a builder subdomain or an allowance. If free is the requirement, this page is not your answer.

Where the others are stronger. Webflow and Framer are better website builders than we are, with far more design control and a far better CMS, and any designer who works in this daily should work in one of those. Wix and Squarespace have thousands of templates, app marketplaces and real ecommerce, none of which we attempt. And Microsoft Clarity gives you heatmaps, session recordings, frustration signals and forty-odd filters for nothing, forever, with no traffic limit and no sampling; if all you want is to watch recordings of your site, install Clarity today and pay nobody, including us. Google Analytics is likewise free and better than we are at traffic reporting. There is no version of this page where those things are not true.

The thing none of them do is join the two halves. A recording is only worth watching if you can act on it, and in every other stack the person in the recording is an anonymous session id. Here, the moment somebody fills a form or books a call, their earlier anonymous visits fold into their contact record, every later visit lands on their timeline, and a return visit can start a workflow while they are still on the page. The page, the form, the video, the pixel, the conversion event and the person are one system rather than five products joined by scripts you hope are still firing.

Tool by tool

The table is the evidence. This is the verdict, one competitor at a time, with the case for choosing them stated first, because in at least one of these it is very hard to argue with.

Inflowave vs Microsoft Clarity

Buy theirs when: Clarity is free forever, has no traffic limit, does not sample, and gives you heatmaps, session recordings, rage and dead clicks, quick backs, forty-odd filters and unlimited seats. On the pure job of watching what people do on your site it is excellent and it costs nothing. If that is the requirement, install it and stop reading.

Buy ours when: It tells you what an anonymous session did. It cannot tell you that the recording you just watched is the person who requested a quote on Tuesday, and it cannot notify your closer that they are back on the pricing page right now. Worth knowing about their limits too: recordings are kept for thirty days unless you favourite them. We are not a better recorder. We are a recorder wired to a CRM.

Inflowave vs Hotjar

Buy theirs when: Deeper behaviour analytics than ours, with surveys and feedback widgets we do not have at all, and now part of Contentsquare with the enterprise analytics that brings. For a dedicated CRO team it is a more serious instrument.

Buy ours when: Their pricing moved to monthly session allowances, and their own documentation says that when you hit the allowance data collection stops until the reset, account-wide, with nothing rolling over. That means the month a campaign works is the month you stop seeing data. Beyond that it is the same gap as Clarity: the session is not a person you can ring.

Inflowave vs Webflow or Framer

Buy theirs when: Better builders, plainly. More design control, a proper CMS, a real component model and an ecosystem. If the site is the product and a designer owns it, build it there. You can also point our pixel at a Webflow site and get the tracking half of this page anyway.

Buy ours when: They stop at the page. The form on a Webflow site posts somewhere else, the conversion event is a developer ticket, the recording is a third-party script, and the person who filled the form is a row in a different tool. We are worse at the page and better at everything that happens after somebody lands on it.

Inflowave vs Wix or Squarespace

Buy theirs when: Thousands of templates, an app marketplace, real ecommerce with inventory and checkout, and a genuinely good experience for a business that needs a website rather than a lead engine. They also own the search results for AI website builders, and for a shop or a portfolio they are the right answer.

Buy ours when: The gap is what happens to an enquiry. On a Wix site it becomes an email. Here it becomes a contact in a pipeline stage with its source attached, its earlier anonymous visits folded in, and a workflow running. If your website exists to produce conversations rather than orders, that difference is the whole thing.

Taken from vendor documentation, checked 30 July 2026. They ship changes constantly, so verify any row that decides your choice.

Common questions

Where are the heatmaps and session recordings covered properly?

They have their own page, because there is a lot more to say about them than fits here: the three kinds of heatmap and which question each one answers, rage clicks and dead clicks, what is masked in a recording, and an honest comparison against Hotjar, Microsoft Clarity and Crazy Egg including the session caps that quietly stop those tools collecting during a traffic spike. Heatmaps and session recordings.

Can I use the tracking on a site you did not build?

Yes, and plenty of people do. The pixel is one line with a single opaque id, the same shape as a Meta pixel, and it works on WordPress, Shopify, Webflow, Framer, Wix or anything else that lets you add a script. You get the heatmaps, the recordings, the identity stitching and the workflow trigger without moving your website. What you do not get on another platform is the part where the tracking is guaranteed to still be there after an edit.

How is this different from Hotjar or Microsoft Clarity?

On the recording itself, Clarity is excellent and free with no traffic limit, and it has more filters than we do. If watching recordings is the whole job, install Clarity and pay nobody. The difference here is that the session is attached to a person. When somebody fills in a form, their earlier anonymous visits fold into their contact record, every later visit lands on their timeline, and a return visit can start a workflow. Clarity can hold an identifier you give it, but there is no pipeline behind it and nobody to ring.

Can you tell me which companies visited my site anonymously?

No, and it is worth being direct about it because several products in this category do exactly that. We do not resolve anonymous visitors to real people or companies. A visitor stays anonymous and analytics-only until they identify themselves by filling in a form or booking a call. If cold person-level identification is what you need, that is a different category of tool, and it carries privacy exposure in the United States that you should take advice on before buying any of them.

What is recorded, and what is hidden?

The recording is a reconstruction of the page and what the visitor did on it, not a video of their screen. Typed input is masked by default, with passwords, email addresses and phone numbers masked specifically. Canvas contents are not captured, and cursor movement is sampled rather than recorded continuously. The recorder is only downloaded at all on projects that have recordings switched on, and it is served from our own domain rather than a third-party script host.

Is this compliant?

The technical side is built for it: inputs masked, no cold de-anonymisation, tenant resolved server-side so the snippet never carries your account id, and conversion events that respect the consent your form collected. The part that is yours is disclosure. Behaviour analytics belongs in your privacy policy and, in Europe, behind your consent banner. We have made that possible to do properly; we cannot do it for you, and any vendor telling you their script is automatically legal everywhere is selling you a problem.

Does it send conversions to Meta server-side?

Yes. Conversions fire from our servers as well as the browser, sharing an identifier so Meta counts each one once rather than twice. That matters because a browser-only pixel loses a real share of conversions to blockers and privacy settings, and the algorithm buying your traffic learns from what it is told. The events are stored with the page and the sequence they happened in, so the whole path across a multi-page journey can be reconstructed afterwards. More on ads reporting.

Does it work with Google Analytics?

Yes. Connect your Google account, pick the property, and the tag is baked into every page. Analytics is free and better than we are at traffic reporting, so we do not try to replace it. It answers how many and from where; the pixel here answers who, and what they did while they were on the page.

Can I clone a site I like?

You can point it at a public page as a design reference. It screenshots the page, pulls out the colours, fonts and section structure, and rebuilds against those. Clone builds then run a comparison loop: it screenshots what it built and the original side by side, lists every way they differ, fixes those, and repeats until they match or it runs out of rounds. Use it on your own pages, an old site belonging to a client, or a layout you have permission to work from. Copying content that is not yours is on you.

Which AI model builds the site?

It is chosen per job rather than being one setting. Building a whole site or cloning one is a single decision that shapes everything afterwards, so it runs on the heaviest model. Recolour a button and it runs on the fastest one. Anything structural or creative goes heavy again, and anything ambiguous takes the middle. You can always override it and pay for a better model on a job you care about.

Do I own the site? Can I export it?

Yes. Download the whole site as files, or export it straight into your own GitHub repository. This is worth caring about more than people do at signup: several builders in this market cannot give you a usable copy of your own website, which turns every renewal into a hostage negotiation.

Can I pull content from my own database?

Yes, from your own Supabase, and only with the publishable key. Secret keys are rejected outright, so what we can read is exactly what your row-level security already allows the public to read. That is a deliberate constraint rather than an oversight: nobody should hand a website builder a key that bypasses their own security rules.

Can I A/B test pages?

Yes. Split traffic between variants, track which one converts, and read the result. There is also a mode for sites we did not build, so you can test a page hosted elsewhere. For testing a whole multi-step funnel rather than a single page, that lives in the funnel studio. See the funnel studio.

Is this as good a builder as Webflow or Framer?

No. They have more design control, a proper CMS and an ecosystem, and a designer who lives in a builder every day should live in one of those. We are a better answer when the site exists to produce conversations and you would rather it came wired to the thing that handles them. You can also keep Webflow and put our pixel on it, which is a perfectly sensible combination.

What does it cost?

The builder and the tracking are part of the platform rather than separate products, and there is no session allowance that stops collecting data in the middle of your best month. What changes between plans is seats and the number of client workspaces. Compare the plans.

Stop guessing what happens on your own website

Build the page, or keep the one you have. Either way the visits become recordings, the recordings become people, and the people land in your pipeline.

Get Started