Human-Content Audit Report
Prepared for Alex Sample — freelance B2B writer (illustrative) · alexsamplewrites.example · August 2026
Overall verdict
Verified Human · High confidence
9 of 10 audited pieces show strong consensus for human authorship across every method. 1 piece triggered a single-engine flag we assess as a likely false positive — details and fixes below.
How this audit was run
Each piece was checked by two independent commercial AI-detection engines, plus provenance and metadata analysis (document creation history, CMS byline consistency, publication timeline). We only report a verdict when independent methods agree — a single detector score is never treated as evidence. Engine names are withheld in the public sample; your report identifies them.
Piece-by-piece results
| Piece | Engine A (AI probability) | Engine B (AI probability) | Provenance | Consensus |
|---|---|---|---|---|
| Blog — "How to price retainer work" | 2% | 1% | Clean ✓ | Human · high |
| Case study — SaaS onboarding | 4% | 3% | Clean ✓ | Human · high |
| Newsletter — March issue | 1% | 2% | Clean ✓ | Human · high |
| Whitepaper — logistics trends | 6% | 8% | Clean ✓ | Human · high |
| About page | 3% | 1% | Clean ✓ | Human · high |
| Blog — "12 tools every writer needs" | 55% | 9% | Clean ✓ | Review ⚠ |
| Guest post — client comms | 5% | 4% | Clean ✓ | Human · high |
| Landing page — services | 7% | 6% | Clean ✓ | Human · high |
| Blog — "Freelance taxes, simply" | 2% | 3% | Clean ✓ | Human · high |
| LinkedIn article — pricing myths | 4% | 2% | Clean ✓ | Human · high |
The flagged piece — and why it's likely a false positive
"12 tools every writer needs" scored 55% on Engine A but only 9% on Engine B, with clean provenance. This disagreement pattern is typical of a false positive on listicle-format content: short uniform sentences and repeated structures are exactly what detectors mistake for AI. Because only one of three methods flagged it, consensus does not support an AI verdict.
Recommended actions: keep the Google Docs version history for this piece linked in your records; consider adding one personal anecdote per section (raises human signal on every engine); re-test after edits. If a client ever questions this piece, share the cross-engine disagreement — it makes the false-positive case for you.
Provenance & metadata summary
Publication timestamps are consistent with an organic multi-year writing history. CMS bylines are stable. Document metadata on supplied drafts shows iterative editing sessions over multiple days — a strong human-authorship signal. No C2PA content credentials were present, which is normal today for text content and does not affect the verdict.
What this report is — and isn't
No AI detector produces proof, and anyone who claims otherwise is selling something. This audit reports consensus signals across independent methods, with confidence levels, so you can have the AI conversation with evidence instead of a single score.
Why we never rely on one engine: independent peer-reviewed testing of 14 detectors found false-positive rates on human writing ranging from 0% to 50% depending on the tool, and a Stanford study found seven detectors misclassified an average of over 61% of TOEFL essays by non-native English speakers as AI. OpenAI withdrew its own detector in 2023 "due to its low rate of accuracy." Vendors publish much lower figures about their own products; those are self-reported, and we don't treat them as equivalent.
Sources: Weber-Wulff et al., International Journal for Educational Integrity, 2023 · Liang et al., Patterns, 2023 · OpenAI · Vanderbilt University
Want this for your own portfolio?
Free while in early access. 48-hour turnaround. Results stay private to you.
Get my free audit