How to Build a Landing Page With AI That Converts

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Build a Landing Page With AI That Converts.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Build a Landing Page With AI That Converts.

Decide the one action the page is for. Then give AI your offer, audience, proof and the objections you hear, and ask for copy in a fixed order: an outcome headline, proof, what happens next, the price or a starting point, answers to objections, and one short form. Build it, track form submissions, and change one element at a time.

Two things decide whether a landing page converts, and neither is the AI tool. The first is match: the page must say, in its first line, the same thing the ad, email or post that brought the visitor promised. The second is effort: every form field, every vague sentence and every extra link costs you enquiries. AI can write a full page in seconds, but left to itself it writes "innovative solutions tailored to your needs", which matches nothing and answers no one. The brief is where conversion is really decided.

Follow me on Instagram@sagnikteaches

Write a one-page brief before any AI touches it

A brief takes 20 to 30 minutes and makes every draft better. Here is an illustrative filled-in brief for a small food testing laboratory launching a page for allergen testing.

Connect on LinkedInSagnik Bhattacharya
PAGE BRIEF: Allergen testing for small food producers

One action:     Request a quote (form), or call.
Traffic from:   Search ads for "allergen testing for food
                business", plus a trade newsletter mention.
Audience:       Owners of small bakeries, sauce and snack makers,
                caterers; 1-20 staff; usually testing for the first
                time because a stockist or customer asked.
Their problem:  A stockist wants test results before listing a
                product, and they don't know what to test or how.
Offer:          Tests for the common allergens, results in 5
                working days from receipt, plain-English report.
Starting price: From $95 per sample per allergen (our real price).
Proof:          18 years testing food; 40+ small producers a year;
                3 reviews mentioning "explained everything".
Objections:     "Which allergens do I need?" "How do I send
                samples?" "Will the stockist accept your report?"
                "Is it expensive?"
What happens:   Quote within 1 working day, sample kit posted,
                results emailed, a call to explain if wanted.
Don't say:      "guaranteed", "certified" (unless we name what),
                anything about legal compliance.

Notice the "traffic from" and "their problem" lines. They tell the AI what the visitor was thinking a second before they arrived, which is what the headline has to meet. And the "don't say" line matters for any regulated or technical service: AI loves "guaranteed compliance", which this lab can't promise.

Subscribe on YouTube@codingliquids

The prompt that drafts the page section by section

Ask for the page in sections, in order, rather than "write me a landing page". You get copy that fits a proven structure, and it's easier to edit one section at a time.

Using ONLY the brief below, write landing page copy in this order:

1. HEADLINE (under 12 words): the outcome the visitor wants, in
   their words. Then 5 alternatives.
2. SUBHEADLINE (under 25 words): what we do and for whom.
3. PROOF STRIP: 3 short, specific facts from the brief.
4. HOW IT WORKS: 3 numbered steps from enquiry to result.
5. PRICE: the starting price and what it includes.
6. QUESTIONS: answer each objection in 2-3 sentences.
7. FORM: suggest the fewest fields we need, and the button text.

Plain words, no jargon the audience wouldn't use, no claims or
numbers that aren't in the brief. Avoid: "solutions", "innovative",
"tailored", "state-of-the-art", "peace of mind".

BRIEF:
[paste]

An illustrative set of headline options the lab got back, with the owner's notes:

Illustrative AI output: "1. Allergen testing that gets your product on the shelf. 2. Your stockist wants allergen results. We'll get them to you in 5 days. 3. Allergen testing for small food producers, explained in plain English. 4. Know exactly what's in your product. 5. Fast, affordable allergen testing. 6. The allergen test your stockist asked for."

Option 2 matches the visitor's situation (a stockist asked) and gives a concrete time. Option 4 overpromises: an allergen test answers a specific question, not "exactly what's in your product". Option 5 is generic; every competitor says it. The lab chose a lightly edited option 2, "Your stockist wants allergen test results. We'll have them in 5 working days", adding "working" because the brief says so, and a stockist deadline is exactly where a vague "5 days" becomes a complaint.

What each section of the page has to do

A structure that works for most service and lead-generation pages, from top to bottom:

  1. Headline and subheadline. Meet the visitor's situation and state the outcome. On a phone, these plus the button should be visible without scrolling.
  2. A button or short form near the top. For people who already know what they want.
  3. Proof strip. Three specific facts: years, numbers of customers, a short review quote used with permission. Specific beats superlative. If you need material, turning customer reviews into marketing copy shows how to mine reviews for it.
  4. How it works, in three steps. Uncertainty about the process stops people enquiring. Three steps is enough for most services.
  5. Price or starting point. "From $95" converts better than silence for most small-business services, because it tells people whether to bother. If you can't give a figure, say what the price depends on.
  6. Answers to objections. The questions from your brief, answered plainly. This section often does more than the headline.
  7. The form again, and a phone number. Some people read everything first and decide at the bottom.

The laboratory's allergen page, from draft to live

Follow the illustrative lab through the build. Its first version of the page, before this process, was a copy of its general "food testing" service page: a headline reading "Comprehensive Food Testing Services", a long list of test types, a nine-field enquiry form (name, company, email, phone, address, product type, number of samples, tests required, message) and a full website menu.

After the brief and prompt:

  • Headline: "Your stockist wants allergen test results. We'll have them in 5 working days."
  • Subheadline: "Allergen testing for small bakeries, caterers and food makers, with a plain-English report you can send straight on."
  • Proof strip: "Testing food since 2008 · Over 40 small producers a year · 'They explained everything' (review on Google)".
  • Form: four fields (name, email, phone, "What are you making?"), with the button "Get my quote". Everything else the lab needs is collected on the quote call.
  • Questions section: five answers, including "Which allergens should I test for?" (answered with "tell us your recipe and who's asking; we'll suggest the tests and explain why") and "Will my stockist accept the report?" (answered honestly: most ask for results from an accredited lab, which this lab is for the listed tests, and it names the accreditation on the page).

The owner spent about five hours in total: 30 minutes on the brief, an hour generating and editing copy, two hours building the page in the lab's existing website builder, and 90 minutes on tracking and testing. In this illustration, the page gets about 500 visits a month from ads. The old page had produced about 6 quote requests a month from similar traffic; the new one produced 13 in its first month. Treat that kind of jump with caution until a second month confirms it, for reasons covered further down; but the direction is what you'd expect when a page stops talking to everyone and starts talking to one visitor.

Three ways to build the page

A landing page builder with AI features

Builders such as Unbounce, Leadpages, Wix and Framer generate a first draft of a page from a description, and several include AI copywriting. Unbounce includes AI copywriting from its Build plan ($99 a month, or $74 billed annually); Leadpages includes AI page creation on all plans. These tools give you forms, templates, A/B testing on some plans, and integrations with email and CRM tools. For prices, traffic limits and which builder suits which business, see the best AI landing page builders for small businesses. Whatever builder you choose, paste your own brief-based copy into it rather than accepting its generated copy wholesale.

An AI assistant writes the page, you publish it

ChatGPT, Claude and Gemini can all write a complete single-page HTML file from your brief and copy. You then need somewhere to publish it: a simple host such as Carrd (whose Pro Standard plan, $19 a year, adds custom domains and forms), or Leadpages' HTML Pub, which is designed for publishing AI-generated HTML to your own domain. This route is cheap and flexible, and it has a catch: someone has to edit HTML for every change, and forms and tracking need wiring up properly.

A page on your existing website

Often the simplest option. Most website builders let you create a page and hide it from the menu. You keep your domain, hosting and analytics in one place, and the page inherits your site's design. The downside is that site templates carry menus, footers and sidebars that pull visitors away, so hide as much as your builder allows.

Writing a price section people understand at a glance

The price section is where many service pages lose people, because "from $95" raises the next question straight away: from $95 for what? AI is useful here for turning your pricing rules into a worked example, which is often clearer than any table. Give it your real pricing and ask for "one typical example with the sum shown, for the most common customer in the brief".

The lab's pricing is $95 per sample per allergen. Its illustrative worked example on the page reads: "A bakery testing two products (a loaf and a cookie) for gluten, milk and egg sends two samples and pays for six tests: 2 × 3 × $95 = $570. We'll suggest which allergens to test when you tell us your recipes, so you don't pay for tests you don't need." That last sentence answers the real worry behind the price, which is being sold unnecessary tests.

Before you publish a sum like this, check it against your price list and ask whoever does the quoting whether it's typical. An illustrative mistake from a first draft: the AI wrote "two products for three allergens: $285", multiplying by products or allergens but not both. A visitor who later got a $570 quote would feel misled, and the page would have produced an enquiry that ends in a complaint.

Before you send traffic: speed, mobile, form and tracking

Paying for visits to a page that's slow, broken on phones or not tracked is the most expensive mistake in this whole process. Run these checks first:

  1. Speed: test the page in Google's PageSpeed Insights. Large images are the usual culprit; compress them.
  2. Phone test: open the page on your own phone. Can you read the headline and see the button without scrolling? Can you fill in the form with your thumb?
  3. Form test: submit a test enquiry. Does it arrive in an inbox someone checks, and does the visitor see a thank-you message or page?
  4. Tracking: set up a conversion in your ad account and analytics for the form submission (a thank-you page view is the easiest thing to track). The step-by-step is in setting up conversion tracking before you let AI spend your ad budget.
  5. Message match: read the ad, then the headline. Do they make the same promise in similar words?

A care agency's recruitment page for applicants on phones

A different audience changes almost everything on the page. An illustrative care agency wants carers, and its applicants mostly see the job advert on social media, on a phone, often on a break. A 12-field application form and a PDF job description lose most of them.

The brief's "their problem" line reads: "Want a care job with guaranteed hours and paid travel time; worried about being thrown in without training." The before and after:

  • Before: headline "Join Our Caring Team!", three paragraphs about the agency's values, and a link to download an application form.
  • After: headline "Care work with guaranteed hours and paid travel time", three bullet points (hours, pay for travel between visits, two weeks' paid training before your first solo visit), a 45-second faceless video of carers' hands at work, and a three-field form: name, phone, "When could you start?" The button reads "Get a call back".

The full application happens on the phone call. Asking for less on the page turns it into a door rather than a hurdle, and the recruiter can screen in a five-minute conversation. For more on answering those enquiries quickly, see replying to every enquiry in under five minutes.

A garage's seasonal offer page and the ad that points to it

Message match is easiest to see when it's missing. An illustrative car repair garage ran a search ad reading "Winter Car Check, $49 – Battery, Tyres, Lights & Wipers" that pointed to its homepage. Visitors who clicked landed on "Family-run garage since 1994", with the winter check mentioned nowhere above the fold. Most left.

The fix was a simple landing page whose headline repeated the ad: "Winter car check, $49. Battery, tyres, lights and wipers, done in 40 minutes." Under it: what's checked (a list of eight items), what happens if something needs fixing ("we'll call with a price before doing anything"), available slots this week, and a booking form with three fields. The AI's contribution was drafting the page from a two-line brief and suggesting the "call before doing anything" line from the objection "will they try to upsell me?". Every claim was checked with the workshop: 40 minutes was realistic, and the list matched what mechanics actually do.

Reading results without fooling yourself

Small businesses rarely have enough traffic for formal A/B testing, so be careful about what the numbers can tell you. A quick sum shows why. With 500 visits a month and a 2% conversion rate, you get 10 enquiries. If a change raises it to 2.6%, that's 13. The difference between 10 and 13 in one month can easily be chance; a busy week, a competitor pausing their ads, or one referral partner sending people can do it. Before crediting a change, look for the pattern to hold for at least two months, or for a big difference (say, doubling) that's hard to explain otherwise.

Behaviour data helps where conversion numbers are too small. Microsoft Clarity is free and records heatmaps and session recordings, and it includes AI-generated summaries of sessions. Watch ten recordings of visitors who didn't convert and note where they stopped. You can then give your notes to an assistant with a prompt like this:

Here are my notes from 10 recordings of visitors who left my
landing page without enquiring, plus the page copy. Group the
likely reasons they left, suggest one change per reason, and rank
the changes by how easy they are to make. Don't invent data beyond
my notes.

An illustrative result for the lab: "4 of 10 scrolled straight to the price and left: consider explaining what 'per sample per allergen' means with a worked example. 3 of 10 started the form and stopped at the phone field: consider making phone optional." Both are one-change tests. Make one, wait a month, then the next. For deeper diagnosis across your whole site, see using AI to find out why visitors don't buy.

Why AI-built pages fail to convert

  • The headline is about you, not the visitor. "Welcome to [business]" or "Your trusted partner in…" tells the visitor nothing. Rewrite it around their situation.
  • Claims nobody can check. "Industry-leading", "best in class", "trusted by thousands". Replace each with a number or delete it.
  • Everything, for everyone. An AI draft from a vague brief lists every service. One page, one offer.
  • A long form "just in case". Ask only what you need to respond. Collect the rest later.
  • No price, no process. Visitors who can't tell what it costs or what happens next usually leave rather than ask.
  • Generic stock images. A handshake photo adds nothing. A real photo of your premises, product or team (faces optional) adds trust.
  • Nobody answers the enquiries. The best page in the world fails if form submissions go to an unmonitored inbox. Test it monthly.

Once the landing page works, the same brief-first approach carries across to the rest of your site; writing your website copy with AI applies it to home, about and service pages.

Landing page questions

Should a landing page have my website's full menu?

Usually not. A landing page exists for one action, and a full menu offers a dozen ways to leave. Keep your logo (linking home) and perhaps a phone number, and drop the rest. If the page is also an important part of your main site, a slimmed-down menu is a reasonable compromise.

Is a video at the top of the page a good idea?

Only if it explains the offer faster than the headline and a few lines can, and it doesn't slow the page down. Many visitors won't press play, so the headline and form must work without it. If you add one, keep it short, put captions on it, and check the page's loading speed on a phone afterwards.

How many landing pages does a small business need?

One per distinct offer and audience, not one per ad. A garage might have one for brake checks and one for fleet accounts; a care agency one for families and one for recruitment. Several ads can point to the same page if they make the same promise. More pages only help if each has enough traffic to judge.

Further reads

Sources: Unbounce and Leadpages pricing pages (AI copy and page-generation features); Microsoft Clarity product page (free heatmaps and session recordings); Google PageSpeed Insights.

Want a landing page that earns its ad spend?

On a 1:1 call we'll write the page brief together, choose where to build it with the tools you already have, and set up tracking so you know what each enquiry costs.

Book a 1:1 call with me