SEO and visibility
Know if your site is ready.
Sends a set of specialists over the site at the same time: four for search (technical, on-page, performance and architecture), one for speed and accessibility when the web-vitals skill is available, and one for conversion that weighs message match, the single action, proof and friction. A judge tests every finding against its evidence, and the run ends in GO or NO-GO, with any blocker that survived at the top.
- One GO or NO-GO verdict
- A judge tests each finding
- Blockers come first
At a glance
- Group
- SEO and visibility
- Command
- /site-audit
- Uses
- Your site, offers and buyer profile
- Writes
- A verdict, a scorecard, a baseline
- Sends
- Nothing. Drafts only.
01When to use it
Use it before you launch or spend.
- A launch, a redesign going live or a big ad spend is close, and you want one verdict instead of four separate reports.
- A client or a prospect asks whether the site is ready, and SEO, performance and conversion should be judged side by side.
- Fixes have shipped, and a re-run should confirm the blockers cleared.
Not for
- Search problems on their own. Use /seo-audit
- Just the page an ad points to. Use /ads-landing
- Speed and Core Web Vitals alone. Use /web-vitals
02What it needs
Start with the addresses to check.
- The site, or the pages in scope. With no addresses on disk, those are all it asks you for.
- Your offers, your buyer profile and your brand voice. Conversion is never scored before the first two are loaded.
- Optional: connectors for measured performance and analytics. Without them, it works from the pages it fetches.
- An earlier site audit, if one exists, so the new run can show what changed.
03What it writes
A verdict file, a baseline and log rows.
- final/site-audit.md
- GO or NO-GO, the blockers, then one scorecard that merges every lens, grouped by lens and ranked by severity, and the quick wins.
- data/baseline.json
- The point the next run compares against. The raw signals from the fetched pages, robots file and sitemap are kept in data/ too.
- Ledger row
- One site-readiness row, stating its source and its confidence.
- Signals rows
- One row per input that was absent, out of date or in conflict with another source, naming what was wrong and which source it went with. If nothing was wrong, it says so.
04Limits
What it never does.
- Return GO while a cited critical blocker is still open.
- Let a thin or uncited finding reach the verdict. The judge cuts those first.
- Publish, send, or alter anything on the live site.
- Show a guess as a measured number. Each figure is labeled measured or inferred.
- Read another client on an agency setup. It stays inside the active client and marks the files confidential.
What stays with you.
- Fixing what it finds, blockers first, then asking for the re-run.
- The go-live decision itself. The verdict reaches you as a draft.
05Sample output
Part of a NO-GO report.
An excerpt for a made-up furniture store, checked before its redesign goes live. In a real report, every finding cites a URL, a file or a ledger row.
| Finding | Lens | Evidence | Severity |
|---|---|---|---|
| Robots file blocks every product page | Technical | example.com/robots.txt: Disallow: /products/ | Critical |
| Two buttons compete for the one action | Conversion | example.com/sofas: Shop now beside Book a showroom visit | High |
| Hero image is 3.1 MB | Performance | example.com: sofa-hero.jpg | Medium |
- Robots file blocks every product page Critical
- LensTechnical
- Evidenceexample.com/robots.txt: Disallow: /products/
- Two buttons compete for the one action High
- LensConversion
- Evidenceexample.com/sofas: Shop now beside Book a showroom visit
- Hero image is 3.1 MB Medium
- LensPerformance
- Evidenceexample.com: sofa-hero.jpg
Verdict and next step
- NO-GO: one cited blocker survived the judge.
- Blocker: the robots rule that hides /products/ from search engines.
- Quick win: delete that Disallow line, then run the audit again.
07Questions
Before you run it.
What makes the verdict NO-GO?
A single critical finding, backed by cited evidence, that the judge does not refute. The rest of the findings are ranked under the blockers, but one surviving blocker is enough.
Why put a judge on the findings?
A finding that sounds right is not yet proven. The judge tests each one against what it cites and cuts the thin or uncited ones, so the verdict rests on evidence instead of a guess.
Does it need analytics or a speed tool hooked up?
No. The default run works from the fetched pages, the robots file and the sitemap. Connectors for measured performance and analytics are optional, and the report labels every number measured or inferred either way.
Have us build it
We can build it for you.
Start with the free plan. A person names the first fix.