Landing page teardown in 20 minutes, with an AI second opinion

When paid traffic stops converting, the first instinct is to change the ads. Often the ad is fine and the page it sends people to is the problem: the headline promises something different from the ad, the offer is vague, or the form asks for eleven fields before anyone has spoken to you.

This teardown is a fixed 20-minute review of one landing page. You score it on seven checks, get a second opinion from an AI assistant using the same sheet, and leave with a short list of changes to test one at a time. It works for any page that paid ads, emails or social posts send people to.

What you need

• The landing page URL.
• The ad, email or post that sends the most traffic to it. Have the exact wording and image open.
• A phone, so you can view the page the way most visitors will.
• A timer, and the scoring sheet below copied into a doc or spreadsheet.

Review one page at a time. If you have five landing pages, start with the one that gets the most paid traffic.

The 20 minutes

The seven checks

1. Ad-to-page match

Read the ad, then open the page and look only at what shows without scrolling. Does the page repeat the ad's main promise in similar words? If the ad says "fixed monthly fee for contractors" and the page opens with "Accounting for every business", the visitor has to work out whether they're in the right place. Check the image too: if the ad shows a person at a desk and the page shows a skyline, the link between them is gone.

2. Headline clarity

Give yourself five seconds with the headline and subheadline, then look away. Can you say what the company does, for whom, and what you get? Write down what you understood. A headline that could sit on a competitor's page unchanged scores low.

3. The offer

The offer is what the visitor gets for taking the next step: a free consultation, a fixed-price quote, a trial, a guide. Look for three things: what exactly it is, what it costs (or that it's free), and what happens after the click. "Contact us" is a button, not an offer.

4. Proof

Find every piece of evidence that the business can deliver: reviews with names, client logos, case examples with numbers, credentials, years in business, guarantees. Proof near the call to action counts for more than proof in the footer. Make sure every claim is true and current; an outdated testimonial or an award the business no longer holds is a risk, not proof.

5. Form friction

Count the form fields. For each one, ask whether you need that answer before the first conversation. Note anything that slows people down: required phone numbers, dropdowns with long lists, a CAPTCHA, a submit button that says "Submit". Fill the form in with test details and see what happens next. A thank-you page that says what will happen and when scores higher than a blank "Thanks".

6. Mobile view

On your phone, open the page from the ad link if you can. Check that the headline and the call to action both show before you scroll, the text is readable without zooming, the buttons are easy to tap, and the form works with the phone keyboard (email fields should bring up the email keyboard, phone fields the number pad). Check for pop-ups that cover the page on a small screen.

7. Page speed check

Run the URL through Google's PageSpeed Insights, which is free. Look at the mobile result first. If the page has enough real visitors, the report shows field data from Chrome users and rates the main loading metrics as good, needs improvement or poor. Note any metric rated poor and the top two suggestions in the diagnostics. Then load the page on your phone over mobile data and note whether the headline appears quickly or you stare at a blank screen.

The scoring sheet

Score each check 0, 1 or 2. Copy this into your doc and fill in the evidence column; a score without a note is hard to act on later.

LANDING PAGE TEARDOWN
Page URL: [ ]                 Traffic source reviewed: [ ]
Reviewer: [ ]                 Date: [ ]

| # | Check              | 0                                  | 1                                   | 2                                          | Score | Evidence |
|---|--------------------|------------------------------------|-------------------------------------|--------------------------------------------|-------|----------|
| 1 | Ad-to-page match   | Page doesn't repeat the ad's promise | Same topic, different wording/offer | Same promise and wording above the fold    |       |          |
| 2 | Headline clarity   | Can't say what they do after 5 sec | What, but not for whom or the result | What, for whom and the result, plainly     |       |          |
| 3 | Offer              | No offer, or "contact us"          | Offer named but vague on cost/next step | Specific offer, cost and next step stated |       |          |
| 4 | Proof              | None, or only generic claims       | Some proof, far from the call to action | Specific, named proof near the call to action |  |          |
| 5 | Form friction      | Many fields not needed yet, unclear after-submit | Some extra fields or weak button text | Only needed fields, clear button, clear next step |  |   |
| 6 | Mobile view        | Call to action below the fold or form breaks | Works with minor problems     | Headline and call to action visible, form works |   |          |
| 7 | Page speed         | Poor on main mobile metrics        | Needs improvement                   | Good, or loads fast on mobile data         |       |          |
|   | TOTAL (out of 14)  |                                    |                                     |                                            |       |          |

Top three changes to test, in order:
1.
2.
3.

The total is for comparing the same page before and after changes, or two of your own pages. It isn't a grade against other companies, and a score of 14 doesn't mean the page will convert.

The AI second opinion

Once you've scored the page yourself, paste the page copy into an AI assistant with this prompt. Copy the text from the page (headline, subheadline, body, button labels, form field labels, testimonials) in the order it appears. Remove anything confidential first; a public landing page usually has nothing sensitive, but draft pages may.

You are reviewing the copy of one landing page. Score it against the sheet below
and explain each score. You will only see text, so score checks 1 to 5 and mark
checks 6 and 7 as "not assessable from text".

The ad that sends traffic to this page says:
[paste the ad headline and text]

Who the page is for: [e.g. owners of construction firms with 5-30 staff]
What we want them to do: [e.g. book a 20-minute call]

Scoring (0, 1 or 2 per check):
1. Ad-to-page match: does the opening of the page repeat the ad's promise?
2. Headline clarity: from the headline and subheadline alone, can a reader say
   what the business does, for whom, and what they get?
3. Offer: is there a specific offer, with cost and next step stated?
4. Proof: is there specific, named proof, and is it near the call to action?
5. Form friction: list the form fields and say which ones are not needed before
   a first conversation. Comment on the button label.

For each check, quote the exact words from the page that led to your score.
Then suggest up to three rewrites of the headline and one rewrite of the offer,
using only facts that appear in the page or the ad. Do not invent prices,
guarantees, client names, results or credentials. If a fact you need is missing,
say what is missing instead of making it up.

Page copy:
[paste the page copy]

Put the assistant's scores next to yours. Where you agree, move on. Where you differ by two points, reread that section: either you missed something or the assistant misread the context. The assistant hasn't seen the design, your customers or your sales calls, so its scores are a second opinion. A person who knows the business decides what changes.

Check every suggested rewrite for invented facts before using it. Assistants tend to add confident claims ("trusted by hundreds of firms") that the business can't back up.

Worked teardown: Harrow & Finch Bookkeeping

Harrow & Finch is a fictional five-person bookkeeping firm that works mostly with small construction and trade businesses. They run a Google search ad and send it to a landing page. Enquiries from the page have been slow, and the owner wants to know why before raising the ad budget.

The ad

Bookkeeping for Contractors | Fixed Monthly Fee
Monthly books, job costing and payroll for trade businesses.
Fixed fee, no hourly billing. Book a free 20-minute call.

The page, before

[Hero image: city skyline at dusk]

HEADLINE:     Your Trusted Financial Partner
SUBHEADLINE:  Comprehensive accounting solutions for businesses of every size.
BUTTON:       Contact Us

BODY:
At Harrow & Finch we believe every business deserves great financial support.
Our experienced team handles bookkeeping, payroll, tax preparation support and
more, so you can focus on what you do best.

TESTIMONIAL (footer): "Great service!" – Happy client

FORM: First name, Last name, Company, Job title, Email, Phone (required),
Annual revenue (dropdown), Number of employees (dropdown), How did you hear
about us? (dropdown), Message
BUTTON: Submit

The scores

The AI second opinion agreed on checks 1 to 3 and 5. It scored proof 0 rather than 1, on the grounds that "Great service!" from an unnamed client gives the reader nothing to check. The owner kept 1, since the firm does have real reviews to add, but the comment was fair. One of the assistant's suggested headlines said "Trusted by 200+ contractors", a number that appeared nowhere in the page or the ad; it was struck out.

The rewrite

HEADLINE:     Bookkeeping for contractors, at a fixed monthly fee
SUBHEADLINE:  Monthly books, job costing and payroll for trade businesses.
              One price each month, no hourly billing.
OFFER:        Book a free 20-minute call. We'll look at how your books are
              kept today and tell you the monthly fee before you commit.
BUTTON:       Book my free call

PROOF (next to the form): two named client reviews, used with permission,
from a builder and an electrician.

FORM: Name, Email, Phone (optional), "What kind of work does your business do?"

The new headline repeats the ad's two main promises. The offer says what happens on the call and when the price is given. The form drops from ten fields to four, and the questions the firm asked about revenue and headcount move to the call itself.

The order of tests

Harrow & Finch didn't change everything at once. If enquiries rise after six changes, nobody knows which one did it, and if they fall, nobody knows which one to undo. They put the changes in this order, one at a time, and let each run long enough to collect a meaningful number of visits before reading the result:

1. Headline and subheadline, to match the ad.
2. Offer and button label.
3. Form cut to four fields.
4. Named reviews next to the form.
5. Replace the skyline image with a smaller photo of the team at work, which also helps page speed.

They logged each change with its start date, the enquiries per week before and after, and anything else that changed in the same period, such as ad budget or a holiday week. If your page tool supports A/B tests, run each change as a split test instead; the one-at-a-time rule still applies.

Run it again in a month

Repeat the teardown on the same page after the changes have run. Compare the totals, re-run the AI prompt on the new copy, and pick the next three changes. Keep every scoring sheet in one folder so you can see what moved over time.

Do it with Merit AC: the Google and Meta media buyers in the marketing agents audit the ad accounts that send traffic to this page, so the ads get the same regular review.