Editorial tool / 15 minutes

AI Content Audit

Make a publish-or-hold decision from visible evidence, human judgment, and reader usefulness. This tool does not detect AI authorship.

Local and private: Do not paste your article. This scorecard stores only ratings, the claim-risk answer, a last-updated time, and its format version on this device.

Basic usage actions are measured to improve this tool. Your ratings and draft are never sent.

Mark the proof sheet

Award points only for evidence visible in the draft or its working files.

0 of 7 categories scored. Claim-risk answer still required.

01 Original value ×5
0
Repeats common summaries
3
Adds a useful framework or example
5
Contains firsthand work, original analysis, or a genuinely useful synthesis

Scores 1, 2, and 4 sit between these published anchors.

02 Claim support ×4
0
Claims have no traceable sources
3
Most consequential claims are linked
5
Every material claim traces to an appropriate primary or authoritative source

Scores 1, 2, and 4 sit between these published anchors.

03 Specificity ×3
0
Could fit any topic or brand
3
Includes some concrete names, dates, or steps
5
Uses precise details throughout and explains why they matter

Scores 1, 2, and 4 sit between these published anchors.

04 Reader outcome ×3
0
No clear job for the reader
3
Gives advice but leaves gaps
5
A defined reader can complete a useful task after reading

Scores 1, 2, and 4 sit between these published anchors.

05 Human judgment ×2
0
Avoids decisions and trade-offs
3
Makes a few recommendations
5
Clearly states choices, limits, trade-offs, and reasons

Scores 1, 2, and 4 sit between these published anchors.

06 Voice and rhythm ×2
0
Uniform, generic, padded
3
Mostly natural with a few dead zones
5
Sounds consistent, intentional, and appropriate to the publication

Scores 1, 2, and 4 sit between these published anchors.

07 Transparency ×1
0
Ownership and process are unclear
3
Byline and sources are present
5
Byline, relevant expertise, methods, and material AI use are clear where useful

Scores 1, 2, and 4 sit between these published anchors.

Consequential claim gate

Does this draft contain an unsupported medical, legal, financial, safety, or reputation-sensitive claim?

Decision rules

The score is advice. The gate is final.

A high total cannot cancel a zero in claim support. Unsupported medical, legal, financial, safety, or reputation-sensitive claims force a hold no matter what the arithmetic says.

ScoreDecisionRequired action
85–100Publish after proofreadFix only factual or mechanical errors; do not sand away the voice.
70–84Hold for one focused editRepair the two lowest-scoring categories, then rescore.
50–69Rewrite around original evidenceKeep useful research, but rebuild the structure and argument.
Below 50Do not publishReturn to reporting, testing, or source collection before editing prose.

Use the result

A score should change the draft

  1. Rate what the draft proves now, not what you meant to add later.
  2. Repair the two lowest raw ratings first; weighting does not decide the repair order.
  3. Resolve the hard gate, rescore, then give a named human the final publishing call.

Keep editing

Read the full 15-minute audit Run a human editorial pass Repair voice without faking humanity Compare slop with useful human work