---
name: social-tiktok
title: "TikTok Publishing Playbook"
description: "Draft and revise TikTok scripts, captions, and publishing packages in an identifiable author's voice. Activate for TikTok posts, product demos, launches, customer proof, and creator collaborations; choose formats by audience intent and separate verified platform guidance from editorial judgment."
license: Apache-2.0
compatibility: "Text and media editing; browser and TikTok account access for current publishing controls."
domains: content
rules:
  - content(tiktok)
  - match(\btik\s+tok\b)
  - content(fyp)
---

## Overview

Write a post a particular person could stand behind, with a useful result for the intended viewer. Apply supplied facts and proof first, the author's documented voice next, then platform guidance. Use `content-voice` for generic editing; the rules here govern spoken delivery and TikTok packaging.

Treat prescriptions marked (directional) as craft choices. Drafting doesn't authorize uploading, messaging, or ad operations. Evidence dates identify publication or, for undated documents, verification month. Open reference/evidence.md when citing benchmarks, explaining distribution or AI policy, or advising a monetizing creator.

## Mental model

Make the post understandable without prior familiarity with the account. For You supplied 72.7% of video views in the sampled population; that does not mean every upload starts from zero (measured, Metricool, n=2,314,756 posts, 2026-05).

TikTok's historical explainer describes interactions, video information including captions/sounds/hashtags, and account settings; it uses completion of a longer video as a strong-signal example. It says follower count and prior hits are not direct factors (official, TikTok recommendation explainer, 2020-06). Don't turn that old description into current weights, a guaranteed initial audience, or a fixed distribution test.

## Format and anatomy

### Choose the viewer's job

Default to video when motion carries the demonstration. Video led photos/carousels in the cited comparison; this is a baseline, not a universal winner (measured, Buffer Formats, n=45M+ cross-platform posts, 2026-03).

| Situation | Format and length choice |
|---|---|
| Brand awareness or a compact demonstration | Test the short band: 15–30 seconds led engagement in the brand sample (measured, Socialinsider, n=6M brand videos, 2026-07). |
| Search-intent explanation or a substantial story | Test the extended band: 120–180 seconds led median views in that sample, not necessarily search conversions (measured, Socialinsider, n=6M brand videos, 2026-07). |
| Reference material the viewer needs to inspect | Consider Photo Mode/carousel; make each frame readable and the sequence useful (directional). |
| A written observation without useful footage | Consider a text post if the account offers it; preserve the author's wording and avoid pretending motion adds proof (directional). |
| Product inspection and live questions | Consider LIVE with Shop only after checking account eligibility and purchase controls. TikTok's UK case report describes Lux Laces using a debut Shop LIVE (official, Lux Laces case report, 2026-08; platform-reported case, not a comparative format test). |
| A response to an existing post or comment | Consider reply-video, Duet or Stitch if available and permitted; contribute a demonstration or substantive response (directional). |
| A timely follow-up | Consider Stories if available; put durable proof in a standalone post too (directional). |
| Related explanations or paid depth | Check Playlists or Series availability and terms before planning grouping or paid access; these are different publishing decisions (directional). |
| Material exceeding the extended band | Check the actual upload limit before scripting long-form. Keep only depth the viewer needs (directional). |
| Creator seeking program earnings | Verify current regional Creator Rewards terms before choosing duration; this evidence set does not verify current thresholds. Brand GTM work follows the buyer's job, not an assumed rewards floor. |

Don't assert unverified upload limits, caption caps, or universal feature access. Confirm these in the publishing account, including market and account type.

### Build the opening and script

TikTok's ad playbook emphasizes the opening 3–6 seconds and a hook → selling points → CTA structure; this is advertising guidance, not an organic retention cutoff (official, SMB Playbook, undated; verified 2026-09).

Use the first frame to show the object of the claim. Skip greetings and logo cards in the opening. Put a readable topic label beside it. Continue with the action that demonstrates the claim, explain its boundary, then give a useful next step or stop. Read the script aloud while watching the asset; cut sentences that merely narrate visible movement (directional).

Use the example hooks as optional approaches: visible fault, buyer question, founder decision/artifact, and observable result. Choose what fits the source and author's register, without rotation quotas; don't copy fictional experiences.

### Package the post

Use full-screen 9:16, at least 720p, and purposeful sound as the ad-playbook production baseline, not an organic upload requirement (official, SMB Playbook, undated; verified 2026-09). Preview on a phone with the posting UI visible; keep proof and text clear of controls and caption overlays. Correct subtitles and describe visual-only evidence in speech or text (directional).

Front-load the subject in the caption using language the buyer uses. Keep it to a compact thought unless conditions need more space; omit script repetition. Place descriptive hashtags after the prose without a tag stack. Hashtag-bearing posts performed better observationally, but no optimal count was established (measured, Metricool, n=2,314,756 posts, 2026-05). Spoken keywords can clarify the topic; don't promise speech-indexing benefits (directional).

Export clean source footage and recut it for this opening and CTA. Choose a cover showing the evidence; write alt text for photo assets (directional). TikTok recommends its Commercial Music Library for product/business promotion; other sounds require the necessary rights and its Music Usage Confirmation. A trending sound's availability isn't commercial clearance or permission to export it elsewhere (official, Commercial Music Help, undated; indexed text checked 2026-09; rights-check application is directional).

## Launch and proof posts

For campaign planning, record audience, buying situation, promise, sourced proof, desired action/destination, stage, disclosures, and account market/type. Use author samples when available; missing samples don't block a restrained factual script. Distinguish owned, sponsored, affiliate, and paid-amplified content. Choosing no CTA is valid. If outcome proof is absent, prepare a demonstration brief rather than inventing customer success.

Use the short or extended bands above as testing options, never compulsory duration (directional):

| Post shape | Anatomy | Length choice |
|---|---|---|
| Launch day | Show the task/result; identify the released product and availability; demonstrate; give the verified access path. | Short |
| Demo or artifact | Show input, perform the relevant operation, inspect output, explain the failure boundary. | Short for a single operation; extended for a walkthrough |
| Customer proof | Show the authorized source record; state [metric and unit], [baseline] → [result], [cohort/sample], [measurement dates], and relevant test conditions beside the number in speech or readable overlay. Identify self-reported results and remaining limitations; if these bounds are missing, show a demo without an outcome claim. Placeholders are nonpublishable until verified. | Short if proof fits; extended for context |
| Founder decision/build update | Name the decision, show its artifact, explain the cost and unresolved tradeoff. | Short |
| Objection answer | Quote a real objection with permission or paraphrase it; test it visibly; state who the result applies to. | Short; extended when the test needs it |
| Recap/lessons | Revisit the promise using observed evidence; name what remains unresolved and the next action. | Extended only when earned |

Apply the strip-test: with the product name removed, the post should still show or teach something useful (directional).

Confirm the destination before finalizing the CTA: an available profile link, verified Shop control, or staffed requested-DM path. Don't promise a typed URL is clickable; test the complete path. TikTok requires its content-disclosure setting for promotion of your own business as well as third-party branded content; identify that relationship in the writing and select the appropriate disclosure category (official, Commercial Guidelines, 2025-08; rechecked 2026-09). Verify account/market-specific controls before publication.

Seed through relevant creator collaborations and useful participation in existing conversations. Give employees distinct observations and disclose their relationship; agree compensation for additional creator work. Don't script fake customer reactions or coordinate engagement trades. TikTok's rules restrict spam, fake engagement and deceptive incentives (official, Integrity and Authenticity, 2025-08).

Prefer founders for decisions they personally made and brand handles for repeatable product explanations; no verified matched TikTok study establishes a universal founder advantage (directional). Give collaborating creators proof and boundaries while preserving their voice; audience/brand tension was reported in creator research (measured, CreatorIQ, n=5,095 creators, 2026-08).

Illustrative launch-week schedule, not a platform timing prescription; offsets are planning examples:

| Offset | Post and reply plan | Qualified-interest check |
|---|---|---|
| T−7 | Teaser showing the unresolved task; collect real objections. | Comments naming the buying situation |
| T−3 | Founder decision or preview; answer scope questions. | Relevant profile visits and access questions where observable |
| T0 | Launch demo; assign a reply owner for the first hours. | Destination actions; substantive product questions |
| T+1 | Objection-answer or comment response; acknowledge limits. | Follow-up questions showing evaluation |
| T+3 | Authorized proof or deeper demo; consider a staffed LIVE. | Saves, relevant shares, product clicks where available |
| T+7 | Recap with actual evidence; answer remaining blockers. | Qualified inquiries and attributable downstream outcomes |

Skip slots without new evidence; don't manufacture a daily quota (directional). Open reference/launch-detail.md when preparing creator briefs, Shop/LIVE, Spark Ads, paid amplification, or launch measurement. It contains scoped case evidence and current business-product options.

## Voice on this platform

Write spoken clauses the author can say comfortably. Preserve their contractions, local idiom and natural fragments; captions may be lowercase only when the author writes that way. Have a fluent local editor review idiom in spoken context (directional; inference from MIDI, 2026-06).

Match subtitle punctuation to speech and use overlay text for orientation. Don't manufacture hesitations, typos, fake corrections, or choppy jump-cut speech. Lo-fi means a workable production choice, not simulated carelessness (directional).

Review script, overlays, caption and replies for information rather than supposed AI markers. Replace an unsupported “this changed everything” with what changed in the footage. Keep genuine questions, contrasts, three-part explanations, and useful punctuation; remove them when they pad the message or distort the speaker's register. Don't prescribe slang, contractions, sentence-length variation, or rough production to simulate authenticity (directional).

Time the explanation against the actual action. Cut a sentence that only repeats a visible movement, but retain a prerequisite or exception needed to understand the result. Edit subtitles for names, numbers, negation, and meaningful sounds; a word-for-word transcript still needs readable segmentation. Put necessary visual evidence into narration or accompanying text, and don't rewrite captions into a different claim just to shorten them (directional).

Don't pair generic synthetic narration with unrelated stock footage as a substitute for the author's evidence. Use a trend's phrasing only when it fits the actual situation; captions and replies should sound like the same person speaking to that viewer (directional).

Bad: “This changed everything.” Better: “The seam still catches when I pull it sideways.” Use the latter only when the footage supports it (directional).

## Cadence and engagement

Choose a cadence the author can sustain with fresh evidence and reply coverage. A brand should budget demonstrations and customer permissions; a creator should budget original episodes and, if relevant, verified program requirements. Increase output only while typical-post quality and qualified response remain useful (directional). Buffer warns that platform engagement definitions are not directly comparable (measured, Buffer Engagement, n=52M+ cross-platform posts, 2026-03). Benchmark within TikTok using consistent metric definitions (directional).

Use audience analytics first. Buffer found evenings from 6–11 p.m. generally strongest, Saturday strongest overall, and Sunday at 9 a.m. the strongest individual slot; test these rather than asserting a universal best time (measured, Buffer Timing, n=7.1M posts, 2026-07).

Assign early and out-of-hours reply coverage when needed; remove fraud/spam and distinguish criticism from abuse. Reply personally to substantive comments; a question is optional, never an obligatory closer (directional). Question posts correlated with more comments, while requests for likes correlated with fewer interactions; this does not establish a follow-request penalty (measured, Metricool, n=2,314,756 posts, 2026-05).

Review at day 10: the study observed 96% of reach and nearly 98% of interactions within that window, not a hard expiry (measured, Metricool, n=2,314,756 posts, 2026-05). Compare available retention and traffic-source data with qualified comments, profile/product actions and downstream outcomes. Keep organic and paid results separate (directional).

## What gets suppressed

Reused material without creative edits, including another person's watermark/logo, and low-quality or minimally edited GIF-only clips can be For You-ineligible under guidelines effective September 13, 2025 (official, Integrity and Authenticity, 2025-08). Re-verify the regional policy before citing it. Don't invent universal QR-code, short-clip, or delete/repost penalties.

Keep AI disclosure and provenance intact. TikTok's general guidelines require labeling realistic AI-generated or significantly edited people/scenes; generic TTS not imitating a recognizable person and minor aesthetic edits are listed exceptions (official, Integrity and Authenticity, 2025-08). A cloned testimonial or fabricated product result doesn't become legitimate through a label. US Shop separately requires disclosure for fully generated or significantly AI-edited content and prohibits fabricated product effects; its no-demotion assurance is Shop-scoped (official, US Shop AIGC Rules, 2026-07).

TikTok reports layered AI labeling and AI-spam detection testing in sensitive topics (official, TikTok AI Update, 2026-07). Treat those as scoped announcements, not proof of a universal AI penalty or immunity. Open reference/evidence.md when an AI-assisted asset needs policy review.

## Examples

These are fictional drafting scenarios (directional), not customer claims. Verify every depicted action before use. Bracketed values require supplied evidence; any unfilled example is not publishable.

### Developer tool: visible fault

Hook/script: “The retry duplicated this job. Watch the job ID when I replay the request. With deduplication enabled, the second request returns the existing job. This recording covers a replay after completion; a simultaneous retry still needs its own test.”

First-frame/overlay: show the duplicate rows; “Job retry demo.” Caption: “Job deduplication on a completed request. Concurrent retries aren't covered here. #BackendDevelopment”

Sound: recorded explanation, no music. Closer: “The runnable example is linked in the profile.” Use only after verifying that path. Cover: the duplicate rows with private data masked. Follow-up: answer concurrency questions without claiming an unrun test passed.

Why it works: The visible-fault hook names what the recording proves.
The limitation follows the provenance rule and bounds the access CTA.

### B2B SaaS: buyer question

Hook/script: “[Verified buyer question about assigning an unowned approval]? I assign the owner here, then the request appears in their queue. The routing works in this sandbox; nobody has measured response time from this demo.” Use a real question, quoted with permission or clearly paraphrased.

First-frame/overlay: show the unassigned request; “Approval routing demo.” Caption: “Approval routing with an assigned owner. Sandbox data; this doesn't establish faster decisions. #Procurement”

Sound: own narration. Closer: “The setup guide is linked in our profile.” Verify that path before use. Cover: the owner field. Follow-up: explain permissions and distinguish routing from adoption.

Why it works: The buyer-question opening answers a documented concern visibly.
The caption prevents a product action becoming an invented business outcome.

### Consumer product: founder decision and artifact

Hook/script: “I kept [documented pocket design] in this prototype because [supplied founder rationale]. Here's the tradeoff: the mug fits, but the lid rubs the zip. I'll close the bag with it upright so you can see the pressure here. The size guide has the internal dimensions.” Use first person only for the founder's own supplied decision.

First-frame/overlay: show the prototype and design record; “Pocket design tradeoff.” Caption: “Mug fit in the side pocket. Check your mug against the internal dimensions. #BagDetails”

Sound: own speech with the zip audible. Closer: “Check the size guide before ordering.” Point to the verified product destination. Cover: contact point. Follow-up: answer sizing questions using supplied dimensions. Disclose the seller relationship.

Why it works: The decision-led opening connects the founder's rationale to an inspectable artifact.
The honest limitation avoids the miracle-fix pattern and supports a buying decision.

### Local service: observable result

Hook/script: “This joint stays still in the repair check. Here's the movement before repair and the same check afterward. The repair was [documented repair]; this is [inspection date], so it doesn't establish long-term durability.”

First-frame/overlay: show the repaired joint under the check; “Chair joint repair.” Caption: “Chair joint repair in [service area]. Result shown at inspection; other joints weren't part of this job. #FurnitureRepair”

Sound: workshop recording and explanation. End after the inspection limitation; no CTA is needed. Cover: the joint, with customer permission. Follow-up: request consent before moving identifiable job details into a private inquiry.

Why it works: The result-led opening anchors before/after footage to the same check.
The placeholders enforce provenance; the bounded result avoids a durability promise.

## Checklist

- [ ] Hook and first frame show the same idea; length and format fit the viewer's job.
- [ ] Every specific is supplied or marked inside a nonpublishable example; no invented experience or results.
- [ ] Meaning and author-voice review covers speech, overlays, caption and replies without fabricated roughness or style quotas.
- [ ] Asset brief, cover, corrected subtitles, photo alt text where supported, and UI safe-zone preview are ready.
- [ ] Sound permissions, customer permissions, commercial disclosures and applicable AI labels are checked for this market.
- [ ] Caption and descriptive hashtags are ready; CTA destination is verified or no CTA is deliberate.
- [ ] Playlist/Series assignment is considered only if supported; pin a useful clarification comment only if that control exists, otherwise answer normally.
- [ ] Reply owner, follow-up candidates, moderation coverage and the study-based review window above are assigned.

## References

- [Metricool study](https://metricool.com/press-release-tiktok-study-2026/), 2026-05-12; [CTA analysis](https://metricool.com/tiktok-trends/), 2026-05-13.
- [Buffer Formats](https://buffer.com/resources/data-best-content-format-social-media/), 2026-03-19; [Engagement](https://buffer.com/resources/state-of-social-media-engagement-2026/), 2026-03-05; [Timing](https://buffer.com/resources/best-time-to-post-on-tiktok/), 2026-07-20.
- [Socialinsider lengths](https://www.socialinsider.io/blog/how-long-are-tiktok-videos/), 2026-07-10.
- [TikTok recommendation explainer](https://newsroom.tiktok.com/how-tiktok-recommends-videos-for-you?lang=en), 2020-06-18; [SMB Playbook](https://ads.tiktok.com/business/library/Global_SMB_Creative_Playbook.pdf), undated, verified 2026-09.
- [Integrity and Authenticity](https://www.tiktok.com/community-guidelines/en/integrity-authenticity), 2025-08-14; [US Shop AIGC Rules](https://seller-us.tiktok.com/university/essay?knowledge_id=491489038501663&lang=en), 2026-07-31; [TikTok AI Update](https://newsroom.tiktok.com/helping-people-spot-and-understand-ai-generated-content-on-tiktok-nz?lang=en-NZ), 2026-07-10.
- [Commercial Guidelines](https://www.tiktok.com/community-guidelines/en/regulated-commercial-activities), 2025-08-14, effective 2025-09-13; rechecked 2026-09; [Commercial Music Help](https://support.tiktok.com/en/business-and-creator/creator-and-business-accounts/commercial-use-of-music-on-tiktok), undated, indexed text checked 2026-09; [TikTok accessibility](https://newsroom.tiktok.com/creating-an-accessible-and-inclusive-tiktok?lang=en-150), checked 2026-09.
- [Lux Laces](https://newsroom.tiktok.com/tiktokshoplivefashionandcollectables?lang=en-GB), 2026-08-04; [CreatorIQ](https://www.creatoriq.com/press/releases/creatoriq-state-of-creators-report-2026?hs_amp=true), 2026-08-11.
- [Aborn](https://emilyaborn.com/ai-writing-tells-what-they-cost-you/), 2026-07-10; [Cox](https://huntingthemuse.net/library/how-to-tell-if-writing-is-ai), 2026-06; [Gichigi](https://tahigichigi.substack.com/p/12-red-flags-of-ai-writing-and-how), 2026-02-18; [MIDI](https://arxiv.org/abs/2606.02147), 2026-06-01.

Re-validate when: regional policies or program terms change; publishing controls are renamed; new benchmark reports appear; recommendation documentation changes.
Validated: 2026-09