ad_landing_message_match v1.0.0

name: "ad_landing_message_match" description: "Audit message match between an ad and its landing page: extract the promise, words, offer, proof, and call to action from both sides, score eight continuity checks, flag hard fails, and decide whether to fix the ad, fix the page, or split the page. Triggers on message match, ad to landing page mismatch, scent, landing page experience, ad relevance, paid traffic bouncing, or an ad and page that say different things." metadata: { "includeInPrompt": true }

Ad Landing Message Match

Purpose

A click is a yes to a specific promise. If the page that follows says something else, in different words, for a different offer, the visitor has to start over, and most will not. This skill audits one ad and page pair at a time by hand, with no analytics account or vendor tool required. It extracts what each side actually says, scores the continuity between them, catches the mismatches that should stop spend on their own, and gives a clear ruling on which side is wrong: the ad, the page, or the pairing itself.

Workflow

1. Capture both sides verbatim. Copy the ad exactly as a visitor sees it: headline, body text, image or video still, offer terms, price, and call to action button. Copy the landing page first viewport exactly: headline, subheadline, hero image, offer block, first proof item, and primary call to action. Work from screenshots or live text, never from memory or from what the team meant to publish. Note the destination URL and, for search ads, the keyword or query that triggered the ad.

1. Build the pair sheet. Fill one row per element in references/checklist.md: promise, audience, problem, product or service name, offer and price, proof claim, call to action, and key wording. Put the ad text in one column and the page text in the next, quoted word for word. A row you cannot fill from the page first viewport counts as missing, not as present further down. Google advises putting important information toward the top so it is immediately visible, and information scent research makes the same point: on arrival, visitors judge a page from its title, images, and what is visible above the fold.

1. Run the five second test. Show someone who has not seen the pair, or simulate it cold, the page first viewport for five seconds. Ask three questions: what is being offered, who is it for, and what happens if you click the button. Compare the answers to the ad. Any answer that contradicts the ad, or comes back as a guess, is evidence of weak scent. Record the answers on the pair sheet.

1. Score the eight checks. Score each check in references/scoring-rubric.md as 0 broken, 1 partial, or 2 matched: headline continuity, offer parity, audience and problem match, product specificity, proof parity, call to action parity, shared wording, and first viewport confirmation. Run the scorer on the audit file to total it: python3 scripts/score_match.py audit.json. See scripts/audit.example.json for the file shape.

1. Check hard fails before trusting the score. A good total does not rescue a hard fail. Stop and fix first if any of these hold: the price, discount, trial length, or offer terms differ between ad and page; the page leads to a different product, plan, or destination than the ad sold; the ad makes an objective claim the page and the business cannot substantiate; the call to action on the page asks for a different action than the ad promised; or material conditions appear only in fine print on the page and nowhere in the ad. Advertising claims must be truthful and substantiated before the ad runs, and that duty covers implied claims and the overall impression, not only the literal words.

1. Rule on which side is wrong. Use the ruling rules in references/scoring-rubric.md. In short: if the product cannot deliver what the ad promises, or the claim has no evidence, fix or pause the ad. If the product delivers it but the page hides it, renames it, or leads with something generic, fix the page. If one page serves several ads with different promises, the pairing is wrong: split the page by promise, or narrow the ad set to the promise the page can keep. Never fix a mismatch by making both sides vaguer. That equalises the words and removes the reason to click.

1. Rewrite the losing side to mirror, not paraphrase. The page headline should use the ad promise in recognisable form, ideally the same key phrase, the same offer terms, and the same call to action wording. Mirroring is not copying every word: keep the page headline honest and readable, but a visitor should be able to point at the page and say, that is the thing the ad offered. Apply the same rule in reverse when the ad is the side being fixed: write the ad from the page that actually converts and can prove its claims.

1. Re-score and record. Re-run steps 4 and 5 on the fixed pair and record the before and after scores, the hard fails found, which side changed, and the exact new headline and call to action. For search campaigns, also note the Quality Score components if they are available: Google builds that diagnostic from expected clickthrough rate, ad relevance, and landing page experience, and it advises keeping messaging consistent from ad to landing page. Treat that diagnostic as a signal to investigate, not as the goal of the audit.

Output Contract

Deliver, in this order:

1. Pair sheet: ad text and page first viewport text quoted verbatim, side by side, per element, with missing elements marked missing.
2. Scorecard: the eight checks with a 0, 1, or 2 each, the evidence quote behind every score, the total out of 16, and the band verdict from the rubric.
3. Hard fail list: every hard fail found, or an explicit none found, with the fixing owner.
4. Ruling: fix the ad, fix the page, or split the pairing, with the reason in one or two sentences tied to evidence from the pair sheet.
5. Rewrite: the new headline, offer block, and call to action for the side being fixed, plus the re-score after the change when the change has been made.

Format for the scorecard:

| Check | Score | Ad evidence | Page evidence | |---|---|---|---| | Headline continuity | | | |

Operating Rules

1. Quote, do not summarise. A score based on paraphrased copy hides the exact word changes that break match.
2. Judge the first viewport first. Proof that exists three screens down does not fix a broken first impression, it only mitigates it, and the partial score should say so.
3. Never invent a lift number. No universal conversion gain from fixing match is promised in the audit. Report what changed in the pair, and let measured campaign data speak later.
4. A claim with no evidence is a hard fail on the ad side, even if the page repeats it faithfully. Matching an unsubstantiated claim makes two wrong surfaces, not one right one.
5. One pair, one ruling. Do not average scores across a campaign to excuse a single broken pair that is spending money.
6. Price and terms must match character for character in substance: amount, currency, billing period, trial length, and conditions. Close enough on money is a mismatch.
7. If the page cannot be changed, say so and rule on the ad instead. Do not ship a report whose only fix is impossible for the team reading it.
8. Keep SKILL.md tight. The element checklist lives in references/checklist.md, scoring bands and ruling rules in references/scoring-rubric.md, and full worked pairs in references/worked-pairs.md.