AI Blocks › Audit a landing page
Find why visitors leave your page, then prove your fix worked
Give it page captures and your funnel numbers. Get a ranked test list and a clear verdict on the first fix.
Find where your page loses buyers and prove the fix is a free AI workflow for the job "Audit a landing page". It guides your AI through several skills in order, passing each step's output to the next.
Free · No account · Made by AI Blocks · Thinking: high
What it does
A store with steady traffic and few orders is tempting to redesign. But changing the page on a guess often changes nothing, or hurts, and without a proper test you cannot tell which. Many teams also call a winner too early, on a handful of orders, and ship a change that only looked good by luck. The first step reads your heatmaps and recordings and lists the friction by money at stake. The second sorts every issue into one of four buckets: tracking faults, obvious fixes to make now, causes to investigate, and test ideas. The third ranks the tests by evidence and likely gain and picks the one or two to run. Then it pauses: you run the top test on live traffic until it has enough orders and export the results. The fourth checks whether the traffic was enough to trust. The fifth checks the split, turns the…
You give it
- Heatmap, scroll map and recording notes for the page, or screenshots of them
- Your funnel numbers, from visits, add to cart, checkout and orders, with traffic sources if you have them
- Customer reviews or survey comments (optional)
- After the pause, the exported results of the test you ran
- Customer words (optional)
You get
- A friction list ranked by page and money at stake, with a friction score out of 100
- Every page issue sorted into fix now, investigate or test, with its proof
- A ranked list of tests with the one or two to run first and the reasons the rest wait
- A test readiness score out of 100, the visitors needed and whether it can finish in about four weeks
- A call on the finished test (ship, do not ship, re-test or inconclusive), a result as a money range and a log entry
Use it when
- Your store has steady traffic but few orders, and you want evidence before you change the page.
- You ran a test and want to know whether you can trust the result.
- Someone on the team wants to redesign a page and you want a ranked reason to change one thing first.
Steps
- Heatmap and recording reader: Paste your heatmap and recording notes. Get the friction found, a test idea for each, and a friction score.
- Page issue sorter: Paste a list of site problems. Get each one sorted into fix now, investigate, test or tracking, with a ranked order.
- Test idea ranker: Paste your research findings. Get each test written as a hypothesis and ranked by evidence and profit.
- A/B test traffic checker: Paste visits, orders and revenue. Get the test size, weeks needed and a readiness score out of 100.
- Store test verdict maker: Give it test numbers and dates. Get valid or not, the lift as a range, a ship call and a trust score.
How to use it
- Download the handoff pack. It's a zip with HANDOFF.md and every skill file this workflow uses.
- Give the folder to Claude (in Claude Code, or as files in a Claude Project) and say "Follow HANDOFF.md".
- Give it the heatmaps, recording notes and funnel numbers for one key page.
- Make the obvious fixes it lists before you test anything.
- Run the top test until the traffic check says it has enough orders, export the results and give them back.