Post-draft diagnostic for photography content. Scans for machine tells, assesses authenticity risk, checks voice continuity against the photographer's voice standard. Fires before committing content changes. For during-writing standards, see like-a-human.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Post-draft diagnostic for photography content. Scans for machine tells, assesses authenticity risk, checks voice continuity against the photographer's voice standard. Fires before committing content changes. For during-writing standards, see like-a-human.
voice-audit (this file) reviews the output — scanning for drift, flagging machine tells, assessing authenticity.
When drift is detected, the response is restoration, not rewriting. A few targeted edits — not a wholesale rewrite that introduces new machine patterns.
Machine Tell Scan
Run this scan against any content written or edited with AI assistance.
Structural tells
Uniform sentence length throughout — mid-length sentences with no spikes or drops
Paragraph template loops: thesis line → 2-3 supports → tidy restatement, repeated section after section
Visible symmetry: every product description the same length, every gallery text the same number of sentences
The sandwich: intro promises what it will say, body says it, conclusion summarizes what it said
Transition tells
"Moreover," "Furthermore," "Additionally," "It's worth noting," "In conclusion," "In essence," "Let's dive in"
Any transition smoother than a gear grind — this voice uses: "And then." "But —" "Look closer." "What you don't see:" "The real subject is..."
Word-level tells
Hedging stacks: "It's important to note that," "One might argue," "It seems like," "Perhaps"
Art-world clichés: captures the essence, evokes a sense of, invites the viewer, explores the interplay, juxtaposition of, speaks to the human condition, transcends the ordinary, a meditation on, pays homage to, celebrates the beauty of, a testament to, breathtaking, mesmerizing
Vague emotional swaps: "contemplative mood" for grief, "tranquil scene" for solitude, "the passage of time" for decay, "a visual journey" for struggle
Rhythm tells
Even sentence length from start to finish
Fragment stacking every time emphasis is needed (occasional = human; constant = machine)
False variation: complex/simple alternation in a mechanical pattern
Substance tells
Everything important, nothing specific: "This beautiful photograph captures a stunning moment in nature"
Interchangeable descriptions that could be swapped onto any photographer's product page
Correct but bloodless: subject identified, technique noted, but no one who took the shot would recognize the description
No rough edges: every paragraph resolves cleanly
Generic beauty language where specific visual observation belongs
Therapeutic/cognitive verbs used abstractly. Grep: \b(process|calibrate|reframe|optimize|integrate|install)\b. Photography carve-out: keep the literal craft sense (frame/reframe a shot, process a negative/RAW, calibrate a monitor); flag the self-help use ("process the emotion" → "what it does to you").
Stock demographic listicle — four-part "Some collectors want…" buyer stack. Name the one viewer the piece serves, or cut to a single concrete clause. (Personas belong in audience-profiles, not recited as a listicle.)
Composite first-person attestation — "I shot this at dawn" / "I was there" with no place, month, f-stop, or what-happened-just-before. Earn the claim with the one photographer-only detail, or report the image without it; the vague half-measure is worst.
Seam detection
One or more paragraphs shift to a more abstract, polished, or catalog-like register
Paragraphs whose sentences all open the same way
A section that feels "inserted" rather than grown from the artist's perspective
Voice Continuity Check
Compare the draft against the like-a-human voice standard. The baseline voice has these markers:
Must be present
Photographer's eye: Descriptions that reference what the light was doing, where the eye lands, what was happening off-frame
At least one concrete, specific detail per piece: Location, time of day, weather, technical choice that shaped the image
Restraint: More unsaid than said. The image carries the weight; the words serve it
Visual precision over emotional generality: "The fog sat about three feet off the water" vs. "a misty morning"
Reverence without preaching: Creation honored as His. Not every piece names God, but none treats the subject as mere decoration
Must be absent
AI-overrepresented vocabulary (see Machine Tell Scan)
Catalog-speak ("This exquisite piece will complement any living space")
Road-map paragraphs that announce what's coming
Drift indicators
If three or more of the "must be present" markers are missing, or two or more of the "must be absent" items appear, the draft has drifted. Flag specific locations and suggest minimal restoration edits.
Precision Check
Visual specificity present? Light quality, time of day, weather, physical details about the scene. If the description could apply to any photograph of similar subject matter, it hasn't been sharpened.
The competitor test. Could this description be copy-pasted onto another photographer's site and still work? If yes, add a pinning detail or flag for rewrite.
Technical details earned? When technique is mentioned, does it illuminate the image ("wide open, so the background dissolves") or just display credentials ("shot at f/2.8, 1/250s, ISO 400")?
Emotional words match emotional reality? When the image depicts solitude, is "solitude" on the page — or has it been swapped for "peaceful serenity"? The sharp word is the right word.
Product descriptions practical? For prints: Does the buyer know the size, medium, and what the medium does to this specific image? "On metal, the blacks go absolute" is useful. "Printed on premium materials" is catalog-speak.
Authenticity Risk Assessment
After completing all checks, assign an overall rating:
Authenticity Risk: Low / Medium / High
Evaluate based on:
Machine tell density
Voice continuity (photographer's eye present or absent)
Precision density (specific visual details or generic claims)
Art-cliché density
Emotional register accuracy
Promotional drift
Catalog-speak contamination
High risk if:
More than 5 machine tells flagged
Three or more voice markers missing
No concrete visual details in the entire piece
Every description is the same length (visible symmetry)
The piece could survive on a stock photography site unchanged
No rough edges anywhere
Art-world clichés appear more than twice
Product descriptions read like a department store catalog
Output: Flag specific locations. Suggest 3-5 minimal restoration edits. Do not rewrite. Restore the voice.
