The redesign brief says the hero feels weak. The proposed fix is a sharper headline, a dashboard mockup, and a brighter button. Before any of that, put the existing sentence on trial.
First, locate the rung you are standing on
Landing-page copy often jumps between three kinds of claim without noticing:
- Capability: the product accepts a CSV, reads a mailbox, or produces a draft.
- Workflow: the operator can review missing items before sending a status update.
- Outcome: the business saves time, avoids errors, or improves a commercial result.
A capability can often be inspected in the product. A workflow claim needs the sequence and its limits. An outcome needs evidence that connects the product to that outcome. Do not climb to the next rung because it sounds better.
For US advertising, the FTC’s advertising-substantiation policy says objective express and implied claims require a reasonable basis before an objective claim is disseminated. That policy has a US scope and does not replace legal review for your market, but the operating lesson travels well: proof belongs before the claim goes live, not in a future case study.
Put the evidence beside the words it can support
Illustrative product, not a customer case or measured result. Imagine a tool that compares an expected-document CSV with messages in a shared inbox. On sample data, the founder can inspect an import screen, a list of unmatched files, and a review queue. No authorised evidence exists for saved time, fewer mistakes, or customer outcomes.
| Candidate line | Inspectable support | Risk if read broadly | Decision |
|---|---|---|---|
| “Import the expected-file list from CSV.” | The real import screen, supported columns, preview and errors. | Visitors may assume every spreadsheet format works. | Keep; add format limits. |
| “See which client files are still missing.” | Sample input beside the resulting missing-item list. | Matching may attach a message to the wrong client. | Keep; show the review state. |
| “Eliminate document chasing.” | Nothing supports elimination or a business outcome. | The line implies complete automation and a verified result. | Delete. |
Write the sentence that marketing usually hides
A useful limitation is not a legal fog bank at the bottom of the page. It sits next to the promise and changes how the visitor interprets it.
Illustrative wirecopy: “See which expected client files have not been matched to the shared inbox. Review uncertain matches before they change a client record.”
The second sentence does real work. It says automation has a boundary, names the risk, and explains where the operator remains in control. It does not weaken the product; it identifies the product that actually exists.
Check implied claims too. A row of customer logos implies a relationship and permission. A security badge implies a defined assessment. “Works with any spreadsheet” implies a compatibility range. “No setup” implies there are no prerequisites. If the evidence does not carry the implication, remove the device or narrow the words.
The CTA is another claim
A button promises what happens after the click. Choose it from the unresolved risk, not from a list of fashionable labels.
| Visitor still needs to know | Honest next action, if it exists | Do not use when |
|---|---|---|
| How the workflow handles a real-looking case | View the sample workflow | The destination is only a signup wall. |
| Whether a source file is supported | Check the CSV format | No format checker or precise specification exists. |
| Whether setup fits a complex environment | Request a setup review | The request is actually a disguised sales call with different terms. |
| Whether the product can be tried safely | Start with sample data | The product immediately asks for live customer data. |
If paid search is used, Google Ads guidance observed on 2026-07-28 recommends alignment between the ad, search theme, landing page, and promised next action. That is platform guidance, not evidence that a particular message will convert.
Red-team the sentence from both sides
The hurried buyer reads
- Who is this for?
- What happens to my data?
- What remains manual?
- What will the click commit me to?
The sceptical buyer asks
- Can I inspect the named capability?
- Does the example use sample or customer data?
- Which important case is unsupported?
- Is the outcome stated as fact or hope?
Ignore animation, extra testimonial slots, and pricing-card polish until the central sentence survives those readings. Design can improve comprehension. It cannot turn an unsupported outcome into a defensible one.
Put one defensible sentence back on the page
Write five lines: buyer moment; exact claim; evidence available now; harm or disappointment if the claim is read too broadly; narrowest useful message. Then choose keep, narrow, move, or delete.
Landing pages bloat when each doubt earns another section. Narrowing the hero is often cleaner: the screen, limitation and CTA can then tell the same story.
Do not redesign around a sentence you cannot support. Keep the claim at the highest rung your evidence reaches, put the limit beside it, and make the button name the real next step.