How to rewrite homepage copy

The steps, from opinion to new lines

These steps are our own Teardown method, which the Teardown tool on this site runs (read 2026-10-14). Its nine steps run here in order, with steps 4 and 5 as one item. The rewrite is step 7; the steps before it decide which lines it may touch.

1. Set the two rules for every new line (method step 1). No em dashes in your own prose, and every claim must be falsifiable.
2. Write your opinion first (step 2). One or two sentences on why the page fails, dated. The method will not start without it, and the opinion later decides which lines get rewritten.
3. Quote the page exactly, then check the markup (step 3). Copy the headline, subhead, button labels and section headers word for word, with the date. Then read the raw HTML. If the two disagree, the method says the markup wins.
4. Mark the Eight Reads, then run the swap test (steps 4 and 5). Mark each Read Holds, Soft or Fails with a quote under it. For Read 8, fetch two named competitors and mark every line either one could publish unchanged.
5. Name one primary failure (step 6). Your opinion picks it, not the worst mark. A Read that failed harder becomes a secondary, and the method lists at most two.
6. Write the after as literal copy (step 7). Rewrite the hero headline, the subhead, the primary button and the one section that fixes the primary failure. Set each new line beside the one it replaces. Rewrite only what the diagnosis condemned, so the diagnosis and the fix read as one argument.
7. Steelman the original (step 8). Write the best reason a competent team shipped the page. Check that the diagnosis still stands, and cut any line that mocks the company.
8. Assemble it, swap test included (step 9). The method's output has a section headed "Does the rewrite pass the swap test?" The Teardown tool runs that check as its last stage, after the after.

One scope rule applies before you start. The method treats replacement copy for a page you own, or a client's, as a different job: commissioned copy. In our view, use the new lines on your own homepage as a brief for the final copy, as How to critique your own homepage messaging explains.

The method sets three limits on the rewrite itself:

• Copy, not advice. The method does not count "They should lead with the outcome" as an after.
• Falsifiable. Write the negation of each new claim. If no company would ever say the negation, the line is decoration.
• Proof already public. Every new claim must be supportable from the company's own public site. An invented capability or number fails the method's quality bar.

Worked sample: Stripe's homepage

"Financial infrastructure to grow your revenue. Accept payments, handle business banking,1 and implement custom revenue models—from your first transaction to your billionth."

"Get started" · "Sign up with Google"

Read on 2026-10-14. Source: Stripe homepage

The "1" is a footnote marker on the page. In the markup, the first sentence is the headline and the rest is styled as the subhead, both inside one H1. Further down, the first section reads:

"Flexible solutions for every business model."

"Grow your business with a comprehensive set of payments and financial tools—designed to work individually or together."

Read on 2026-10-14. Source: Stripe homepage

Our opinion (step 2), given 2026-10-14. "No payments company would say the opposite of 'grow your revenue', so the headline proves nothing. The numbers that could prove it sit lower on the page." That points at Read 3, the Claim.

What the copy does well. The subhead names three concrete jobs: accept payments, handle business banking, run custom revenue models. "From your first transaction to your billionth" gives the range of businesses it serves. The proof is strong and carries units. The band headed "The backbone of global commerce" reads "$1.9T in payments volume processed in 2025" and "99.999% historical uptime for Stripe services".

The diagnosis (step 6). Read 3 fails: the negation of "grow your revenue" is a line nobody publishes. Read 8 is the secondary. Adyen's homepage says "Drive revenue with optimized payments and automated payouts." (Adyen, read 2026-10-14). Checkout.com's says its engine is "designed to increase conversions and capture more revenue." (Checkout.com, read 2026-10-14). Either could run Stripe's headline unchanged.

The fair reading (step 8). In our view, a broad headline lets one page speak to a startup and an enterprise at once. The subhead's range, from a first transaction to a billionth, fits that reading. The diagnosis still stands, because the page already holds numbers that could carry the headline.

The after (step 7). Two of the four parts change: the headline and the first section. The subhead and the button stay, because the diagnosis did not condemn them.

• Headline. Before: "Financial infrastructure to grow your revenue." After: "Stripe processed $1.9T in payments in 2025. You can start in as little as 10 minutes." Both facts are on the page, in the stats band and in "Get up and running with Stripe in as little as 10 minutes."
• Subhead. After: unchanged, "Accept payments, handle business banking,1 and implement custom revenue models—from your first transaction to your billionth." It already names three jobs a reader can check.
• Primary button. After: unchanged, "Get started". The Ask was not the failure.
• First section header. Before: "Flexible solutions for every business model." After: "Payments, billing, card issuing and stablecoins, separately or together."
• First section intro. Before: the "Grow your business" line quoted above. After: "Accept payments online and in person, run any billing model, issue cards and move money in stablecoins. Each product works on its own or with the rest." Every item comes from the section's own product headings.

The swap test on the rewrite. Our first draft of the headline was "Financial infrastructure with 99.999% historical uptime." It was falsifiable and on Stripe's page. It still failed, because Adyen's homepage claims "99.999% historical platform uptime." The $1.9T figure survives the swap. Adyen gives its own volume as "€1.4T" processed annually. Checkout.com's hero gives no volume figure, but lower down its page gives "$ 300 BN" as "The total volume of ecommerce payments we processed in 2025 for enterprise merchants around the world." Neither rival could publish Stripe's number.

Common mistakes

• Rewriting before diagnosing. Write the opinion and name one failure first.
• Rewriting every line. Keep what held, and say why you kept it.
• Skipping the swap test on the new lines. A true, falsifiable line can still be one a rival publishes too, as the uptime draft above shows.

Questions people ask

Can AI rewrite my homepage copy?

Yes, but in our view a rewrite only helps once something has named what is wrong. The Teardown tool needs your opinion on why the page fails before it starts. It then quotes the page, marks the Eight Reads and names one primary failure. It returns replacement copy, line by line, and its last stage runs the swap test on the rewrite.

How do you write a homepage headline?

In our method, a headline passes two tests. First, write its negation: if no company would say the opposite, the headline is decoration. Second, put two named rivals' logos on it. If either could publish it unchanged, it is not yet yours. Build the line from proof already on your own site, such as a number with a unit.

Sources

• Stripe homepage, https://stripe.com/, read 2026-10-14
• Adyen homepage, https://www.adyen.com/, read 2026-10-14
• Checkout.com homepage, https://www.checkout.com/, read 2026-10-14
• Two options for creating a great homepage headline, Gobbledy, https://gobbledy.substack.com/p/two-options-for-creating-a-great, read 2026-10-14
• Our teardown method skill, as run by the Teardown tool (ours, read 2026-10-14)