Code & apps
Syntax, leaked keys, dangerous calls, loose comparisons, empty catches and copy-paste repetition.
Code checklistIndependent QA for AI output
The quality gate for AI-generated work.
ES6 QA AI checks the code, copy, documents and images your AI produces — offline, in your browser, against a published standard. When something needs judgement rather than a rule, a named specialist takes it from there.
- No account to start
- Nothing leaves your browser
- Named human sign-off
Report · ASY-2481
Example82of 100
Ships with fixes
Marketing article · 1,240 words
2 findings above tolerance
Accuracy & sourcing74
AI phrasing signals88
Structure & readability91
Two statistics carry no source
Paragraph 3 and 6 make numeric claims without attribution.
Six phrases from the AI tell-list
"delve into", "ever-evolving", "it's important to note".
Checked offline · nothing uploadedHuman review available
Passes through the assay
- JavaScript
- TypeScript
- Python
- HTML
- CSS
- SQL
- JSON
- Markdown
- CSV
- Blog copy
- Landing pages
- PNG
- JPEG
- WebP
Every roll labelled, every batch logged.
01 — Coverage
Six kinds of output, one consistent standard.
Automated checks handle what can be measured. Everything that needs taste, context or a professional credential goes to a person who signs their name to it.
Written content
AI phrasing tells, hedging, unsourced numbers, repetition, adverb density and reading level.
Signals guideDocuments
Heading order, placeholders, broken anchors, ragged CSV rows and inconsistent date formats.
Document workflowImages
Resolution, contrast, sharpness, compression and embedded provenance — including C2PA markers.
Provenance guideWebsites
Pasted markup is checked for missing alt text, absent metadata, heading gaps and bad links.
Check a pageVideo & audio
Frame-level review isn't a rule set — these go straight to a reviewer with the right credits.
Request a reviewer02 — Process
From machine check to human signature.
- 01
Run the checks
Paste or drop a file into the console. Every check executes locally, in the browser, in under a second.
- 02
Read the findings
Each finding is specific, ranked by severity and explains what a fix looks like. No aggregate mystery score.
- 03
Escalate what needs judgement
Pick the skill you need. Your request is matched to a reviewer who works in that discipline, with your report attached.
- 04
Keep the sign-off
Reports and reviews are stored in your dashboard so the audit trail lives with the deliverable, not in a chat thread.
03 — The standard
Built to be argued with.
A verification service is only worth what its rules are worth. So the rules are written down, versioned and shown next to every finding they produce — including the ones that exist because we were wrong before.
Read the full methodNothing is uploaded to check it
The verification engine is plain browser code. Your draft never leaves the machine it was written on.
Every finding cites a rule
The standard is public. You can disagree with a rule and see exactly why a line was flagged.
Humans are named, not pooled
Reviewers are listed with their discipline, rate and turnaround. You know who signed the work.
No claim of authorship detection
We don't pretend to know which model wrote what. We check whether the output is fit to publish.
04 — Guides
How to review the work AI handed you.
Engineering
A practical checklist for reviewing AI-generated code
The eleven failure patterns that show up in assistant-written code, and how to catch each one before review.
ReadMediaImage provenance, C2PA and what metadata can actually prove
What Content Credentials do and don't tell you, and the offline signals worth checking on any image.
ReadEditorialThe signals in AI-written copy that readers actually notice
Which tells matter for comprehension and trust, and which are noise that editors should stop chasing.
ReadStart with one file
Find out what your AI output is hiding.
The console is free, needs no account and uploads nothing. See the findings first, then decide whether you want a specialist to sign it off.