Symptom: The hero promises a number you cannot support.
Cause: Copy was written to sound persuasive before any evidence existed.
Fix: Add a real measured proof item, or drop the number and state the outcome qualitatively.
Landing Page Copy Generator
Build the page's message architecture before the prose, then draft editable landing-page copy that is grounded in the facts you supply. Every claim carries an evidence status, and anything you cannot support yet becomes an unmistakable placeholder instead of a persuasive invention.
- Message architecture first: audience, problem, outcome, value, mechanism, differentiator, proof, objections, CTA, each with its evidence status.
- Deterministic and inspectable (no AI). A number, compliance term, guarantee, superlative or urgency claim is only written as prose when it is backed by a supplied fact or a solid proof item.
- Editable draft, claims-requiring-verification list, scoreless page audit, hero framing variants, experiment handoff and an SEO pack. No login.
No login. Autosave in your browser. Share link compresses the plan into the URL (not encryption). Export JSON, CSV or Markdown.
No login and no server storage. A share link compresses your plan into the URL, that is not encryption. Export JSON for a lossless private handoff. Analytics never receive your copy or product detail.
1. Source context
Supply the facts. Mark each risk-bearing statement with its evidence status; the generator will not present an unsupported claim as fact.
Only real, supplied consequences are used. None are invented.
How it works (real steps; the tool never fabricates them)
2. Policy and reassurance facts
FAQ answers, the card statement and CTA reassurance come only from what you supply here. Unknown policy stays unknown.
3. Proof items
Real proof only. A proof item marked as an assumption is labelled everywhere it appears.
4. Objections you actually hear
The tool will not invent objections. Leave the answer blank to mark it answer required.
Add your facts and evidence status, then click Generate. The tool will show the message architecture, a draft, and every claim that still needs verification.
How it works
- Set the page goal and purchase motion (self-serve, sales-assisted or hybrid), audience type, framework and depth. Every choice changes organisation and framing, never the facts.
- Supply the source facts: what the product is, the primary audience, the problem and its evidence source, the cost of inaction, the desired outcome, the mechanism, the differentiator, proof items and the objections you actually hear. Mark each risk-bearing statement as a stated fact, customer evidence, an assumption to validate, or no evidence yet.
- Click Generate. You get a message architecture (what the page must communicate, with an evidence status per message), an editable draft, a claims-requiring-verification list, a scoreless page audit, hero variants as framing hypotheses, and an SEO pack.
Why message architecture comes before prose
Most AI writers jump straight to a headline. That hides the real problem: whether the page is communicating the right things, in the right order, with the right evidence. This tool separates the two. The architecture lists the messages the page must carry and the evidence behind each one. Only then is copy assembled. If a message is an assumption, you see it labelled as one; if it has no evidence, you see a placeholder. That is how a team avoids shipping a page whose most confident sentence is also its least supported.
Evidence status: fact, evidence, assumption or placeholder
The generator will not launder an assumption into a published fact. Each risk-bearing statement is marked, and the marking is visible in both the architecture and the draft.
| Status | How it is treated |
|---|---|
| Stated as fact | The team is asserting this. Quantified, compliance, guarantee and superlative claims still need a supporting proof item before they are written as prose. |
| Customer / research evidence | Backed by interviews, tickets or research. Treated as solid, and the source is shown next to the message. |
| Assumption to validate | Rendered with an “assumption to validate” marker and added to the verification list. Never silently upgraded to a fact. |
| No evidence yet | Rendered as an unmistakable placeholder, not prose. The generator will not invent the missing detail. |
Claim-to-proof, not copy-to-applause
A claim is only as good as what stands behind it. “Save 8 hours a week” is not a great headline; it is an unsupported promise until a measured result exists. This tool tracks which claims have proof and which do not, and refuses to write the second kind as prose. If you want proof-led copy, provide a real quote, case study, metric or certification. If you do not have one yet, the honest move is to publish without the claim, and the tool makes that the easy path rather than the hidden one.
Where this tool ends and the other owners begin
A copy generator should not quietly absorb research, strategy or statistics. These boundaries keep each CraftUp tool honest about its job.
| Owner | Owns | This tool does |
|---|---|---|
| Value Proposition Canvas | Owns customer jobs, pains, gains and value-fit evidence. | Consumes the confirmed value story. It does not run the research. |
| ICP Segment Picker | Owns which segment is worth targeting. | Aims copy at one supplied audience. It does not choose the market. |
| Pricing Page Critique | Owns pricing-page evaluation and pricing-strategy questions. | May reuse copy and positioning. It does not audit or set pricing. |
| Experiment Hypothesis Generator | Owns the falsifiable hypothesis contract. | Produces variant hypotheses and hands them off. |
| A/B Test Plan Generator | Owns sample size, duration and the statistical contract. | Never calculates statistics. |
| Product Launch / GTM and Product Strategy | Own the broader go-to-market and strategic choices. | Produces one communication artifact, not the strategy. |
Pro tips
- Write the message architecture before the prose: audience, problem, outcome, value, mechanism, proof, objections, CTA. If a message has no evidence, mark it as an assumption instead of writing it as fact.
- Use the customer's words for the problem. Interview transcripts and support tickets are better sources than a clever phrase you invented in a meeting.
- One primary audience per page. If you serve two segments, build two pages or one page that forces a single ICP above the fold.
- Every risk-bearing claim, a number, a compliance term, a guarantee, a superlative, urgency, needs a supplied fact or a real proof item. Otherwise the generator writes a placeholder, not prose.
- A feature is not a benefit and does not prove one. State what the feature is intended to change, and label it a hypothesis until you have evidence.
- If you have no proof yet, say so. Collecting proof is better than styling a fake testimonial so it reads as real.
- Match the CTA to how the product is actually bought. A sales-assisted product should not pretend to be a one-click free trial.
- FAQ answers must be facts: pricing, terms, integrations, security and support. If you do not know, mark the answer required instead of inventing it.
- Test framing, not synonyms: outcome-led vs problem-led vs proof-led are different hypotheses. Synonym swaps are not experiments.
- Keep agitation restrained. State the real consequence if you have it; never invent financial harm or urgency to create pressure.
Common mistakes (symptom, cause, fix)
Symptom: A testimonial or logo appears that no one can trace.
Cause: A placeholder was styled to look like real proof.
Fix: Replace it with a real, sourced quote or leave the unmistakable placeholder until proof exists.
Symptom: The FAQ answers refunds, trials or security that the team never agreed on.
Cause: The generator was expected to invent policy detail.
Fix: Supply the real policy facts, or leave the answer marked 'answer required'.
Symptom: The page claims to be the best, fastest or #1 option.
Cause: A superiority claim was added for impact.
Fix: Replace it with a supported differentiator or an honest note on who the product is for.
Symptom: The CTA does not match the buying motion.
Cause: Every page was given the same 'Start free' button.
Fix: Match the CTA to the goal and motion: demo for sales-assisted, trial for self-serve, waitlist for pre-launch.
Symptom: Objections the buyer actually raises are missing.
Cause: Objections were invented from a template instead of collected from real calls.
Fix: List the real objections from sales calls and support, and answer each one with facts.
Symptom: The page uses urgency that is not real.
Cause: Scarcity was added to push a decision.
Fix: Remove it. Only genuine, supplied urgency may appear.
Symptom: Copy is published with assumptions presented as facts.
Cause: The message architecture and its evidence status were skipped.
Fix: Resolve the claims-requiring-verification list before the page goes live.
Privacy and reuse
There is no login and nothing is sent to a server. Your plan autosaves in your browser under a versioned key. A share link compresses the plan into the URL for collaboration; compression is not encryption, so treat a share link as sensitive and export JSON for a lossless private handoff. Malformed storage or an oversized share link never destroys your current work. Analytics receive only structural values (tool, goal, motion, framework, counts, export format), never your copy, proof or product detail.
FAQ
Is this an AI writing tool?
No. It is a deterministic message-architecture and copy system. You supply the facts and evidence status; it assembles a message architecture, drafts editable copy, and refuses to present unsupported claims as fact. That is deliberate: you can inspect exactly why every sentence exists, and no line is invented by a model.
What is a message architecture?
It is the structured list of what the page must communicate before any prose is written: audience and context, problem, outcome, value proposition, mechanism, differentiator, proof, objections and call to action. Each message carries an evidence status, so the page never launders an assumption into a published claim.
How does it stop me from publishing a claim I cannot support?
Any risky claim, a number, a compliance term, a guarantee, a superlative or urgency, is only written as prose when it is backed by a supplied fact or a solid proof item. Otherwise it becomes an unmistakable bracketed placeholder and appears in the claims-requiring-verification list. Assumptions are always labelled.
What does the evidence status mean?
Each message is marked as a stated fact, customer or research evidence, an assumption to validate, or no evidence yet. Facts render plainly, assumptions render with an 'assumption to validate' marker, and messages with no evidence render as placeholders rather than being invented.
Does changing the framework or tone change the facts?
No. Framework changes the order and structure of the sections; tone changes framing and connectors. Neither adds a claim, a number, a guarantee or a superlative. The underlying facts and their evidence status stay identical.
Can I edit the generated copy?
Yes. Every section is editable in place. Your manual edits are kept when you regenerate, and if you change the source context afterwards the affected sections are flagged so you can review them instead of silently losing changes.
Can I copy across from the Value Proposition Canvas or another tool?
Yes, as structured facts rather than as generated claims. Bring your confirmed audience, problem, outcome, mechanism and proof into the inputs, or import a previous JSON export. The tool will not invent the research it did not receive.
Do I need to log in, and is my copy private?
No login and no server storage. Your inputs stay in your browser and autosave locally. A share link compresses the plan into the URL, which is not encryption, so treat it as sensitive and export JSON for a lossless private handoff. Analytics never receive your copy or product detail.
Learn more with CraftUp
Courses, guides and glossary entries for the work behind the copy.
From message to evidence
Use CraftUp tools and courses to strengthen the evidence behind the page, from discovery to pricing to experiments.
Freshness
Last updated: 2026-09-24
- 2026-09-24: Rebuilt as an evidence-grounded message-architecture and conversion-copy system. Risky claims without evidence render as placeholders instead of prose.
- 2026-09-24: Added per-message evidence status, a claim-to-proof model, editable copy blocks, source-change flagging, and experiment handoff that does not invent statistics.
- 2026-09-24: New storage/share schema (v2) with v1 migration, size guard and privacy warning; behavioural checker replaces the old count-based assertions.