Conversion Copy Audit

Made by VendBuddy (vendbuddy.io). Guide: https://vendbuddy.io/skills#conversion-copy-audit (This line is optional attribution. Keep it or delete it.)

Goal

For one page and one goal (signup, purchase, booking, reply), produce a scored audit and ready-to-paste replacement copy. Read-only until the user approves the edits.

Step 1: Get the real page

Fetch the live URL, not the source repo: curl -sL <url>. Extract title, meta description, H1 to H3, button text with its target, and any FAQ. Note the date. If you can, also get the baseline numbers: visits to the page and the number who took the goal action.

Step 2: Define the reader and the one next step

Write one line each: who is reading, what they want, what single action this page should earn. If the page has several competing actions, that is finding number one.

Step 3: Score (0 to 2 each; 0 absent, 1 weak, 2 strong)

1. Headline: says what you get and for whom, in plain words, within 5 seconds.
2. Subhead: the how or the proof, in one line.
3. Single primary action: one obvious button, specific text ("Start my free trial", not "Submit").
4. Proof: real numbers, named customers or quotes with permission, logos, screenshots. Never invent any.
5. Objections answered: price, effort, switching cost, "will it work for me".
6. Risk reversal: guarantee, free tier, cancel-anytime, only if true.
7. Specificity: concrete outcomes and numbers instead of adjectives.
8. Friction: form fields, steps, load speed, mobile layout, surprise costs.
9. Voice: sounds like a person, no filler ("seamless", "cutting-edge", "unlock").
10. Scannability: headings tell the story alone; short paragraphs.

Total out of 20. Under 10: rewrite the top of the page. 10 to 15: fix the top three weak items. Over 15: test, do not rewrite.

Step 4: Rewrite

For each item scored 0 or 1, give: current text, replacement text, and the reason in one line. Use the reader's own words where you have them (reviews, support emails, survey answers); quote them exactly and say where they came from. Write two headline variants so the user can A/B test.

Step 5: Measure

Give each edit a metric, a baseline, and a read date at least 14 days out. Change one thing at a time on low-traffic pages; otherwise you cannot tell what worked.

Rules

• Never fabricate testimonials, numbers, logos, scarcity, or guarantees.
• No fake urgency (countdowns that reset, "only 3 left" when untrue).
• Run the ai-writing-humanizer skill's checklist on the rewrite if it is installed.
• Do not change prices or legal terms; flag them as decisions for the owner.

Output format

A table (item, score, evidence), then the replacement copy, then the measurement plan. Keep the whole thing under two pages.