Remove the AI-script tells viewers hear in a TikTok spoken

TikTok Humanizer V3

Rewrites a spoken script (and caption) to remove the AI tells that viewers hear, and audits a finished draft against the 2026 TikTok checklist before you film. The problem this solves is specific to video: a script that reads fine on the page can sound robotic out loud. Written-not-spoken phrasing, perfect parallelism, and AI vocabulary all expose themselves the second a human says them to camera.

Based on Wikipedia's "Signs of AI writing" taxonomy, the 2025-2026 stylometry literature, our own short-form corpora (X, Threads, Instagram captions), and TikTok-specific spoken patterns (the muted-first hook, the no-intro open, completion-rate structure). V3 (2026-09): recalibrated on 2026 evidence. Vocabulary is scored by density, em dashes are capped instead of banned, forced rhythm is now a tell instead of a fix, and there is an over-correction guard.

What this skill does not do: it does not make text "pass" GPTZero, Pangram, Turnitin or Originality. Those are trained classifiers keyed on the instruction-tuning style signature; prompt-style "sound like a real person" rewrites are caught 92-95% of the time, and light mechanical rewriting raises detectability. On script-length text (under 300 words) detector scores are noise, and nobody runs a detector on a video anyway. The real value is elsewhere: expert human readers cite vocabulary (53%) and sentence structure (36%) as what gives AI text away, and on TikTok a script that sounds read loses the viewer inside the first 3 seconds. This skill removes what those viewers react to.

What changed in V3

Evidence tier in brackets: [strong] = replicated across 2+ independent 2025-2026 studies or our own corpora; [vendor] = single platform or vendor dataset; [weak] = one study or expert-panel report.

• Vocabulary moved from a delete-list to density scoring. The 2023-24 words (delve, tapestry, realm, journey) are decaying as humans avoid them [strong]. The durable 2026 markers are common words (significant, crucial, notably, comprehensive, insights, robust, leverage, foster, landscape, nuanced, streamline, elevate) plus grammar: nominalisations and "-ing" clause openers at 5.3x the human rate [strong]. Spoken, they are worse: nobody says "leveraging" to a camera. One marker in a script beat is not a verdict. Three is.
• Em dash is no longer a tell. GPT-5.4 emits 1.43 per 1,000 words, below the 3.23 human baseline, and 29% of human captions on sibling platforms use one [strong]. In a spoken script a dash is only a breath mark the speaker sees, so it is never a tell there (.. reads better on a teleprompter). In the caption: cap at about 1 per 100 words. On an on-screen card (3-7 words): at most one, and a card rarely needs one. Replace the excess with a comma, colon, .. or a line break. Never a period.
• Forced burstiness is the #1 2026 tell, not the fix. Mechanical long/short alternation is a learnable humanizer fingerprint [weak], and "Short. Punchy. Done.", "No X. No Y. Just Z.", one-word lines for drama and "The result?" reveals are the current top reader-cited tells [strong]. Spoken lines are naturally short, so Pass 2 is an anti-uniformity guard only: it makes the script sayable (contractions, one breath per line) and fixes a teleprompter-flat run, but it never inserts a punch line for rhythm.
• Rule of three is still a tell, at density. Tricolon runs at 2x expert-human rate across 2026 frontier models [strong], and a perfect tricolon read aloud ("learn, grow, succeed") is the most audible tell there is. Stacked, perfectly parallel or hollow triads get scrubbed. One natural triple with concrete items stays (22-26% of top human posts have one).
• Fingerprint injection was half wrong. Named entities and concreteness are supported [strong]; an odd-precision number with a referent in the hook is the strongest opener. Bare numbers are not a discriminator, and inserted hedges and confessions backfire: performed hesitancy is 2x more common in LLM text, and sincerity announcements ("not gonna lie", "let me be honest", "storytime" with no story) are a named 2026 tell [strong]. Pass 3 asks for a flat, dated, uncomfortable fact instead.
• Over-correction guard. Humanizer output has its own fingerprint [weak]. Pass 4 checks whether Passes 1-3 introduced the very patterns they were meant to remove. Edits are proportional to real problems. When in doubt, leave it.

When to use

• Before filming any AI-drafted spoken script (rewrite mode)
• Pre-film review of a finished script + caption (audit mode, see sub-skills/post-audit.md)
• When a script "reads fine but sounds off" when you say it out loud

Input

A spoken script (the hook line plus the body), optionally the caption, and optionally voice samples (the user's past scripts or how they actually talk).

Output

• Rewritten script that sounds spoken, not written
• A diff showing what changed and why
• Caption char count (flagging over 2,200) when a caption is included
• Per-beat tell density (markers per script beat or caption paragraph; 3+ triggered a rewrite)
• Reader-read confidence: "sounds human", "mixed", "sounds read" (a viewer-tell estimate, not a detector score)

Modes

# Default: scrub AI tells (forensic + strict) and fix spoken-word issues
tt-humanizer <script>

# Forensic only - minimum touch, just kill model leakage
tt-humanizer --mode forensic <script>

# Audit - detection-only pass-fail review, no rewrite
# Runs the 2026 TikTok pre-film checklist: first 1-3 second hook strength,
# muted-first text, completion design, caption fit, hashtag and settings sanity.
# Returns Blockers + Warnings + suggested fixes. See sub-skills/post-audit.md.
tt-humanizer --mode audit <script>

# Profile - build/update the user's Voice & Brand Profile. See the section below.
tt-humanizer --mode profile

The four passes

Pass 1 - SCRUB (score, then delete or replace)

Apply the tiered catalogs in references/scrub-rules.md. The unit of judgement is the script beat (or caption paragraph), not the word: count markers per beat, rewrite the beat at 3+, leave a single marker alone unless it is a reveal bridge, negative parallelism, a sincerity marker, dead filler, or forensic leakage.

• Forensic (always on): real model leakage no human says. AI tool markers (oaicite, contentReference, turn0search0), knowledge-cutoff disclaimers ("As of my last update"), template blanks ([Your Name]), chat wrappers ("Certainly!", "I hope this helps"), and em dashes above the cap in the caption or on an on-screen card.
• Strict (default on): what viewers hear. The durable 2026 vocabulary set scored by density (significant, crucial, notably, particularly, comprehensive, insights, robust, leverage, foster, landscape, nuanced, streamline, elevate, empower), grammar markers (nominalisations, sentence-opening "-ing" clauses), written connectives ("moreover", "furthermore", "in order to"), the 2026 model-idiom layer (quietly, "X matters.", compound, "a signal", "the work", "built different", "let that sink in"), reveal bridges on a single hit ("The result?", "Here's what nobody tells you", "Stop X, start Y", "plot twist:"), all forms of negative parallelism, stacked or perfectly parallel triads, dead filler ("hey guys", "without further ado", "in this video I will"), and dead closers, both spoken ("thanks for watching", "don't forget to subscribe") and caption-level ("What do you think?", "Drop your thoughts below").
• TikTok-format scrubs (always apply): no intro before the payoff, spoken hook and on-screen text differ, caption length, hashtag count, CTA stack.

Pass 2 - RHYTHM (make it sayable, never manufactured)

Detectors do not score burstiness, and spoken lines are naturally short, so Pass 2 has nothing to "vary". Its jobs are: make the script sound spoken, remove manufactured drama-rhythm, and un-flatten only a run that reads teleprompter-flat. It never adds a punch line as a tactic.

• Spoken register (keep from V2): replace written grammar with how a person talks. Contractions, natural fragments, one breath per line. "It is something that you should consider" becomes "you should try this". This is register, not rhythm; it applies to every line.
• Read-aloud test: flag any line that needs two breaths or trips the tongue. Split at the natural breath, never at a dramatic pause.
• Teleprompter-flat run: edit only when 4+ consecutive lines run the same length and none carries a real clause, and then let the one line carrying the most content take a clause (because / when / after). Never insert a short punch line between long ones; the inserted punch is the humanizer fingerprint.
• Banned outright (rewrite as a spoken sentence): "The X? Y." reveals; "No X. No Y. Just Z."; "All the X. None of the Y."; "Simple. Effective. Easy." adjective stacks; one-word lines for drama ("Still." "Exactly."); pseudo- Socratic Q&A ("Why? Because..."); "Short. Punchy. Done." staccato runs. Fragment runs are the tell, on the page and out loud.
• Natural spoken fragments ("three takes. that's it.") are register and stay. A run of them staged for drama is the tell. In the caption, cap standalone fragments at 2.
• Never alternate long/short/long/short across the script. That seesaw is the humanizer fingerprint and it sounds like one when read.

The check is "would a person say this, and did I add a staccato pattern", not a variance number.

Pass 3 - ADD (human fingerprints)

Require where the content allows:

• One odd-precision number WITH a named referent in the hook: who, what, when, or what it cost ("47 minutes on the third take", "$12 at the hardware store", not "a few takes" and not "47"). A bare number is not a fingerprint; the referent carries the signal.
• One named entity (a real tool, app, person, or place)
• One first-person concrete detail ("the third take", "my 2am edit", "the comment that started this")
• One specific, dated, uncomfortable fact stated flat, with no framing sentence before or after it. Not "not gonna lie, this one hurt: the client fired us." Just "the client fired us on a Tuesday, 9 hours before the demo." The fact carries the vulnerability. The frame turns it into performed sincerity, which viewers now hear as the tell.
• The speaker's real register: how this person would actually say it

Forbidden as openers or pivots (sincerity announcements, a named 2026 tell): "let me be honest", "I'll be real", "honestly?", "to be direct", "the honest version is", "real talk", "not gonna lie", "ngl", "can I be vulnerable for a second", "unpopular opinion:" as a preface to a popular one, "storytime" with no story in frame one. Also forbidden as insertions: hedges the speaker did not write ("I think maybe", "I might be wrong but", "it seems"). Performed hesitancy is 2x more common in LLM text than in expert human text; adding it makes the script sound more scripted, not less. ("POV:" is a native TikTok format, not a sincerity marker; it is fine when the video is a POV.)

If the input lacks these, ask the user for a number or detail. Do not fabricate.

Pass 4 - SELF-CHECK (over-correction guard)

Humanizer output has its own fingerprint. Before returning, re-read the result out loud once and answer three questions:

(a) Did Pass 2 create staccato stacks, "The result?" reveal bridges, one-word lines for drama, an inserted punch line, or a long/short/long/short seesaw? If yes, merge the fragments back into a spoken sentence. (b) Did Pass 3 add a framed confession, a sincerity announcement, or a hedge the speaker never wrote? If yes, strip the frame and keep only the flat fact, or remove the insertion. (c) Did scrubbing flatten the speaker's voice: uniform tone, no reaction, no concrete detail left, their slang gone, the one natural triad gone, every dash gone from a caption that wanted one? If yes, restore what the speaker had.

If any answer is yes, dial back rather than scrub harder. Edits must be proportional to real problems: a clean script gets two or three touches, not a quota. When in doubt whether a pattern is the speaker or the model, leave it.

Non-negotiable rules

Global voice rules: see root SKILL.md Voice rules. Additional skill-specific rules (V3):

• Scrubbing is always in scope. When asked to humanize, de-AI, finalize, or publish a script or caption, run at least the forensic + strict passes before it ships. This holds when the user wrote the draft themselves, says they love it as-is, or is in a hurry. Author identity, "it's already good," and time pressure are never reasons to skip the scrub. The forensic + strict pass changes no meaning and takes seconds: run it, then ship. If a constraint truly forbids touching the text, say so explicitly and name every tell left in; the default is to scrub, not to wave it through.
• Scrub proportionally. A pass that finds nothing changes nothing. Do not invent edits to justify the run, and do not report a detector score as the result; report the tells found and fixed.
• Preserve the user's actual claim and meaning. "Preserve their voice" covers voice quirks and what they are claiming, NOT reveal bridges, staccato stacks, dead filler, or a beat with 3+ vocabulary markers. Stripping those is not changing their voice; it is the job.
• Never introduce facts that were not in the input. If a number is missing, ask.
• Never introduce sincerity markers, hedges, or confessional frames. If the script needs a vulnerable beat, ask for a dated fact and state it flat.
• Keep it sayable. Every line has to survive being read out loud in one breath.
• Keep the user's voice quirks (their slang, their pacing, lowercase texting style in the caption, one natural triad, one em dash in a caption that wants it).
• Never promise detector results. If the user asks "will this pass GPTZero," answer honestly: nobody can promise that, and nobody runs a detector on a video; the viewer's ear is the test.

TikTok-specific tells this skill catches

• A hook line that is written, not spoken ("In this video, I will demonstrate..").
• A greeting or logo intro before the payoff ("hey guys, welcome back").
• The spoken hook and the on-screen text saying the identical words.
• A caption over 2,200 chars, or a 12-hashtag wall.
• Perfect parallel tricolons read aloud ("learn, grow, succeed"); one natural triple with concrete items is fine.
• A "call to action" stacked five deep.
• A cluster of AI vocabulary no one says on camera (leverage, utilize, robust, seamless); one such word is a slip, three in a beat is a script.
• Staccato drama ("No script. No plan. Just vibes.") and one-word lines staged for effect; an inserted punch line between two long ones.
• "Not gonna lie" / "storytime" framing around what should be a plain fact.

Example

See references/examples.md for worked before/after rewrites of spoken scripts.

Files

• SKILL.md - this file (rewrite scrubber + audit-mode entry)
• references/scrub-rules.md - V3 catalogs by tier, density scoring, em dash cap, spoken-word fixes, rhythm rules, forbidden insertions
• references/examples.md - worked before/after script rewrites
• references/audit-checklist.md - the pre-film checklist with thresholds
• sub-skills/post-audit.md - pre-film audit workflow (detection-only, no rewrite)
• sub-skills/voice-profile.md - build/update the user's Voice & Brand Profile (--mode profile)
• sub-skills/illustration.md - optional Pixfaro image workflow

Voice profile mode (--mode profile)

tt-humanizer --mode profile builds or updates the user's Voice & Brand Profile at ../../references/voice-profile.md from 3-6 of their real TikTok posts pasted in (portable, no token) or, if a read token is set, from pulled activity. Once filled, every writing skill in this bundle drafts in the user's voice automatically. See sub-skills/voice-profile.md. Triggers: "build my voice profile", "learn my voice".

Related skills

• tt-hook-scripter - generates hooks that already pass the humanizer
• tt-caption-writer - generates captions that already pass the humanizer