# Ad To Landing Page Match Ad Creative > Ad To Landing Page Match - Skill: `mardab96/ad-to-landing-page-match-ad-creative` (Agent Skill) - Install (CLI): `npx skillmds@latest add mardab96/ad-to-landing-page-match-ad-creative` - Raw SKILL.md: https://api.skillmd.com/api/skills/mardab96/ad-to-landing-page-match-ad-creative/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: mardab96 (https://skillmd.com/u/mardab96) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/mardab96/ad-to-landing-page-match-ad-creative --- # Ad To Landing Page Match Shared quality bar: `../references/output-standard.md`. All numbers cited here live in `../references/thresholds.md`. One ad, one destination, one verdict on whether the page keeps the ad's promise in the reader's first screen. ## Use this skill when An ad is live or about to be, and the page behind it has not been read next to it. Typical triggers: a page was changed by somebody who never saw the ads pointing at it, an offer changed in the ad and not on the page, or a funnel diagnosis has already located the loss at the session to conversion step and named this destination. It starts from a live URL. Poor conversion on its own is not a trigger for this skill. `losing-ad-diagnosis-ad-creative` owns the session to conversion step and decides whether the loss is even on the page: it reads impressions, link clicks, landing sessions and conversions for one ad over one window and locates the first step that falls behind, which this skill cannot do because it reads no performance counts at all. Send an ad that converts badly there first, then come here with the step it named. Coming here first risks rewriting a page for an ad that was losing people at the hook. It does not review ad wording for claim exposure either, which is `claim-safety-review-ad-creative`. ## Required input - The full ad: headline, primary text or descriptions, visible on-asset text, and the call to action button label. - The exact destination URL including its query string, since parameters can change what the page shows. - A screenshot or capture of the destination as it renders on mobile and on desktop, above the fold, or permission to read the live page. - The offer as the advertiser understands it, including any condition the ad does not state. - The conversion action the page is meant to produce. ## Analysis workflow 1. Extract the three elements from the ad that `match.promise_pairs` names: the main claim, the offer, and the call to action. Write each as one sentence before opening the page, so the page cannot quietly redefine what the ad promised. 2. Open the destination with the same URL the ad uses, parameters included, and record what actually loads: redirects, consent walls, interstitials and variant tests all change what the arriving reader sees. 3. Locate each of the three elements on the page and record where it appears: first heading, first paragraph, further down, or absent. 4. Apply `match.above_fold_rule` to the main claim. Present further down the page counts as absent for this check, because a reader who clicked is confirming a promise, not reading an article. 5. Check the offer word for word. A different amount, a different trial length or a condition present on the page and not in the ad counts as a mismatch rather than a nuance. 6. Check the call to action verb and destination. Whatever the ad's button says, the page's first action must be the same thing, and two different forms are still a mismatch. 7. Read the mobile rendering separately. Above the fold is a different amount of page on a phone, and a match confirmed on desktop is not a match confirmed. 8. Score against `match.promise_pairs`: three of three is a match, two of three is a partial match reported as a finding, one or none is a break. Report the count, name the missing element, and never wave a partial through. ## Decision rules - Three of three present and the main claim clears `match.above_fold_rule`: `ship`, confidence `high` on this skill's own definition of `high`, set out in the output format below rather than borrowed from the test scale. - Two of three present: `rewrite` the page section carrying the missing element, and name which of the three is missing in the row. - Main claim present only below the fold: `rewrite`, because `match.above_fold_rule` treats it as absent. - Offer differs between ad and page: `approval_needed` on which of the two is correct, since this is a commercial decision and not a copy fix. - Destination redirects, gates or fails to load: `investigate` before any copy work, and mark the row `needs_data`. - Poor conversion with a clean three of three match: `investigate` outside the creative. The answer is not on this page. ## Output format Open with one line: does the page keep the ad's promise above the fold, yes or no, with the count out of three. | Element | As the ad states it | As the page states it | Position | Match | Evidence label | Confidence | Decision | |---|---|---|---|---|---|---|---| What `high` means in this skill, stated here because the comparison family's `high` rests on `test.min_runtime` and `test.volume_floor`, which are conditions of a test and cannot be met by a page check at all. Borrowing them would make every row on this skill `low` for reasons that have nothing to do with how well the page was read. Here, `high` means three things and only these three: the page was opened at the ad's own destination URL with its parameters, it was read on the device family the ad actually serves to, and all three elements of `match.promise_pairs` were located on the page or positively ruled out. Miss any one, a remembered version of the page, a desktop-only read on an ad delivering to phones, or an element nobody looked for, and the row is `medium`. Two missing is `low`, as is any row where the page did not load as the arriving reader would see it. Then the specific page edits required, then `What this could not see`, `Missing data`, `Approval gates`. ## Practical example Illustrative made-up campaign, not a real advertiser. Ad headline promises a free 14 day trial with no card. Ad button says Start free trial. Destination is a product page whose first heading names the product category and whose first action is a Request a demo form. The trial is described in a paragraph 3 screens down and states that a card is required. Reading: "No. 0 of 3 above the fold." All three elements of `match.promise_pairs` fail, and the count says so. The main claim fails `match.above_fold_rule`, since the trial appears 3 screens down and therefore counts as absent. The call to action does not match: the ad sells a self-serve trial and the page's first action is a sales conversation. The offer conflicts outright on the card requirement, so this row is `approval_needed`, and the question of which version is true is not one this review can answer. The claim that the page underperforms because of its design is refused as unverified: the mobile capture was not supplied, so mobile is `needs_data`. That same gap sets the confidence. The page was read live at the ad's own URL and all three elements were located, but not on the device family this ad mostly serves to, so the rows are `medium` on this skill's definition of `high`, not `high`. ## Guardrails - Never confirm a match from the ad and a remembered version of the page. Read the live URL, with its parameters. - Never treat a claim below the fold as delivered. - Never resolve a conflicting offer yourself. Route it to the human who owns it. - Do not edit the page or change the ad's destination.