How it works
One URL in. A worst-first fix list out.
One URL in. We pick the checklist for that page — homepage, landing, product, pricing, contact or policy — and return a worst-first fix list. Shots below are real output; the public sample happens to be a pricing URL.
01
Paste a URL
Homepage or landing page first. Product, pricing, contact and policy work too — one page, not the whole site.
02
We read it as a crawler does
Raw HTML only. If your FAQ needs JavaScript to appear, we say so.
03
Every question tested
A checklist for that page type. Each question comes back clear, buried or missing — with the evidence.
04
Take the fix and go
A rewrite, a FAQ block, a brief for your coding agent, or a PDF.
01 · The audit
Every buyer question, graded three ways
Each page type has its own checklist. Every question is tested against your raw HTML and comes back clear, buried, or missing.
Buried is the one that matters — on your page, and still unusable.
- What is your refund policy?missing
- What is this product, and who is it for?buried
- Can I cancel anytime?buried
- Does pricing scale per user or per usage?buried
- Is there a setup fee or hidden cost?buried
- Who else uses this?buried
- What does each plan actually cost?clear
- What's included in each plan — what do I give up on a cheaper tier?clear
8 of the 12 questions checked on this page. Real output from a public sample scan — read the whole report.
What is this product, and who is it for?
CriticalThe most an AI can pull off your page today
“Checkly is a code-based uptime and synthetic monitoring platform (APIs, websites, browser checks) for developers, SREs, QA engineers, and engineering…”
Say this instead
Checkly is a monitoring-as-code platform that lets you Detect, Communicate, and Resolve issues via Uptime Monitoring, Heartbeat Monitoring, Synthetic Monitoring, and AI-powered testing, plus Status Pages, Alerts, and AI Root Cause Analysis. It's built for Developers (checks in TypeScript, deployed w…
Real output from a public sample scan — read the whole report.
02 · The fix
The words to use, in your page's own voice
Every gap gets a rewrite plus two alternatives, built from material already on your page.
Where your page never said something, you get a blank with the gaps marked — not an invented refund window.
03 · The handoff
Or hand the whole thing to your coding agent
Download the findings as a markdown brief and point Claude Code, Cursor or Codex at it.
It tells the agent to change your page, never your product, and to hand back anything it can't source as a question.
# Page fix brief — Pricing page
- **Page:** checklyhq.com/pricing/
- **Score:** 67/100 · 69/100 completeness
## Read this before you change anything
2. “What an AI extracts today” is a paraphrase, not a quote.
3. Never publish a `[bracketed]` placeholder.
5. Change the page, not the product.
## Where each answer belongs
- What is this product, and who is it for?: belongs in the hero/intro section, needs a short explicit sentence naming th…
## Questions only the site owner can answer
- Do you offer refunds, and if so under what conditions (time window, prorated, none)?
Real output from a public sample scan — read the whole report.
How it behaves
A wrong score is worse than no score.
Half the report is measured rather than judged — page structure, schema and the render gap involve no model at all.
It refuses to guess
A page too thin to judge returns no score at all, rather than a confident zero.
Buried is its own finding
The common problem is not a missing answer — it is one nobody can extract. Most tools score that as a pass.
It marks its own uncertainty
Drafts for things your page never said are labelled templates, blanks left visible. Unsourced specifics are flagged.
It picks one thing to fix first
Every finding is true, which is not the same as useful. The worst one goes at the top, in plain prose.
Run it on your own page.
One URL, about thirty seconds. Three free scans, no card.