How the page generator works
Most landing pages lose visitors for the same few reasons: the promise is vague, the next step is unclear, and the obvious objection goes unanswered. The generator looks for those gaps and proposes copy that closes them.
- 1
Paste your page URL
Any public page works — a landing page, a homepage, a pricing page, a product page. PageDuel fetches the live HTML on the server, skips your nav and footer, and pulls out the headlines, buttons, labels, and body copy that carry the argument. No snippet, no CMS access, no setup.
- 2
The AI reads the page like a visitor
Before it writes anything, the model works out what the business sells, who the page is aimed at, and what the conversion goal is. Headlines and calls to action get the most attention, because they decide whether anyone reads the rest.
- 3
You get two variants, side by side
One variant is rewritten for clarity and specificity. The other is rewritten for motivation and urgency. Both render as full previews of your real page, not a list of suggestions in a text box, and every change is listed with a one-line reason for it.
- 4
Run them as a real A/B test
Import a variant and it becomes a live experiment in PageDuel. Traffic splits, your goals track conversions, and the results report shows which version won — with revenue attached, not just clicks. That is the part no AI can tell you in advance.
What the AI actually changes
Every change is copy. The generator does not touch your layout, your design system, or your code. It also holds headings and buttons to the same word count as the original, so nothing wraps or breaks when a variant goes live.
Headlines
Category descriptions get replaced with a specific claim. “We help businesses grow” is a line every competitor could write. The rewrite names the outcome the reader came for, in the same number of words, so the hero still fits.
Subheadlines
The line under the headline is usually spent on a feature list. It gets rewritten to answer the question the headline just created: who this is for, and how it actually works.
CTA copy
“Submit”, “Learn more” and “Get started” say nothing about what happens on the next click. The rewrite names the next step — “See how it works”, “Get your free report” — which lowers the commitment the button seems to ask for.
Social proof and trust lines
Where your page already carries proof — a customer count, a rating, a “trusted by” line — the wording is sharpened so it does some work near the decision point. The AI never invents a number: if a statistic is not already on your page, it will not appear in a variant.
Value proposition framing
Features get reframed as outcomes. Weasel words — leverage, enable, utilize, solutions — get replaced with a plain description of what the product does for the person reading it.
Objection handling
Most pages leave the hard questions to the FAQ. The variants move the reassurance next to the button instead: no credit card needed, cancel anytime, free plan available. Beside a CTA, an answered objection usually beats another benefit.
Prices, statistics, brand names, and copy that is already strong are left alone. A single run usually produces somewhere between ten and twenty-five changes per variant, each with the reasoning attached, so you can judge the logic before you ship anything. If you want a wider view of what else is worth testing on the page, the CRO audit checklist covers the parts copy cannot fix.
A before and after example
Here is the shape of a typical rewrite on a generic SaaS landing page. It is an illustration of the reasoning, not a customer result — what comes back depends entirely on your page.
Before
- Headline
- The Modern Platform for Team Productivity
- Subheadline
- Powerful features, seamless integrations, and enterprise-grade security.
- Button
- Get Started
After
- Headline
- Cut Five Status Meetings to One
- Subheadline
- See what shipped, what slipped, and what is blocked.
- Button
- Start Free
Why the second version is worth testing
The before version describes a category. Any project tool could run that page, which gives a visitor no reason to pick this one. The after version names a result the reader can hold against their own week, so it filters for the right buyer instead of appealing to everyone. The subheadline answers the question the headline creates — how would I see that — rather than listing three more features. The button drops the setup that “Get Started” implies and states the risk instead.
Notice that the word counts match line for line. That is deliberate: a rewrite that breaks your hero layout never gets shipped, however good the sentence is. And notice what is missing — a promised lift. None of this is proof. It is a hypothesis with a reason behind it, which is what deserves traffic. The sample size calculator tells you how much traffic that will take.
You get two variants rather than one recommendation for the same reason. Two different arguments, tested against your original, teach you something about your audience. A single rewrite only tells you an AI has opinions. If A/B testing is new to you, the beginner guide to A/B testing covers running the first one properly.
Frequently asked questions
Is the page generator free?+
Yes. Generating variants costs nothing. Without an account you get three generations a day; signing in raises the limit. Running a variant as a live A/B test needs a PageDuel account, which starts with a 14-day free trial and no credit card.
Do I need to sign up first?+
Not to generate. Paste a URL and the two previews appear. You only create an account at the point where you want to import a variant as an experiment and put it in front of real traffic.
Does it work on any URL?+
It works on public pages that return HTML — landing pages, homepages, pricing pages, product pages, blog posts. Pages behind a login, pages that block automated requests, and pages that build all their content in the browser with JavaScript may return too little text for the AI to work with.
Does this change my live site?+
No. Nothing is written to your site and nothing is published. The generator fetches a copy of your page and applies the rewrites inside a preview frame. Your live page only changes when you install the PageDuel snippet and launch an experiment yourself.
How is this different from a copywriting tool?+
A copywriting tool hands you text in a box and stops. Here every rewrite is bound to a specific element on your real page, previewed in place, and can be turned into a running experiment that measures conversions and revenue. The output is not a better draft. It is a hypothesis you can settle.
Can I actually test the two variants?+
Yes. Import a variant and PageDuel creates the experiment with the changes already applied. Traffic splits deterministically, your goals track conversions, and the results report shows lift with real confidence calculations — plus revenue per variant once you send purchase events.
Stop guessing which version is better
Generate the variants here, then run them as a real experiment. PageDuel gives you analytics, A/B testing, and revenue attribution from one snippet.
14-day free trial · No credit card required