git:20260905.0fd23a5 to git:20260905.9c336b2

---

name: social-threads

title: "Threads (Meta) Publishing Playbook"

description: "Write and revise Meta Threads posts, replies, and launch copy in the author's documented voice. Use for posting on Threads, a Threads campaign, or a threads.com brief; covers format choice, honest proof, conversation, and conversion paths. Excludes multi-post threads on other platforms."

license: Apache-2.0

compatibility: "Network access for source checks; a Threads account for composer and publishing checks."

domains: content

rules:

- match(\bmeta\s+threads\b)

- match(\bthreads\s+(app|post|meta|campaign)\b)

- match(\b(posts?|posting|publish|reply)\s+(on|to|for)\s+threads\b)

- match(\bthreads\.(com|net)\b)

- match(@threads\b)

---

## Overview

Write for Meta Threads when the brief names the app or says “post on Threads.” Interpret “write a thread” using the named destination; don't default ambiguous requests to another platform.

Use supplied facts and proof first, the author's documented voice next, and platform register as fallback. Apply content-voice without copying its generic rules into the draft. Drafting doesn't authorize posting, replies, or DMs. Open [reference/platform-and-evidence.md](reference/platform-and-evidence.md) when planning regional timing, using newer surfaces, checking disclosure scope, or interpreting research.

## Mental model and mechanics

Use [threads.com](https://www.threads.com/) as the web entry point; Meta migrated the domain (official, Meta Web, 2025-04). Write for the relevant community rather than treating platform membership as a reachable prospect list (directional).

| Constraint or surface | Writing decision |

|---|---|

| Standard post: up to 500 characters; text attachment: up to 10,000 characters (official, Meta Text, 2025-09) | Count the finished root including its link. Keep the core point in the root; use an attachment for necessary depth. |

| Keyword search began rolling out in September 2023 (official, Meta Introduction, 2023-09) | Name the actual subject in readable language; avoid keyword stuffing (directional). |

| Tagged-topic posts generally receive more views (official, Meta Personalization, 2025-07) | Choose a relevant topic when it fits; this is an editorial choice, not a verified maximum-tag rule (directional). |

| Communities Hub, Community Progress, expanded Champion recognition, and Local Communities exist (official, Meta Communities, 2026-06) | Read the community before drafting. Recognition doesn't establish endorsement or buying intent (directional). |

| Your Algo introduced private topic preferences with an initial market-limited rollout (official, Meta Communities, 2026-06) | Keep topic relevance clear; don't promise fixed engagement weights or universal distribution (directional). |

| Insights includes views, interactions, follower detail, discovery sources, and trends (official, Meta Personalization, 2025-07) | Use available owner metrics; don't assume competitors expose the same data (directional). |

Treat engagement benchmarks as observations. Video and images led Buffer's Threads format medians, but results overlapped substantially (measured, Buffer Engagement, n=2025 Threads subset of Buffer-published posts, 2026-03). Choose media for the information it carries.

## Rules and format decisions

Keep a source note for each factual claim and permission for customer material. Obtain missing proof before drafting factual copy; unresolved example placeholders make a draft unpublishable. Never invent experience to supply personality.

Lead with the observation, decision, visible result, or genuine question. Supply enough context for the intended reader to answer; don't hide the subject behind an empty curiosity hook. Develop the idea in connected prose, then stop or give the desired action (directional).

Length bands below are qualitative drafting choices, not performance optima (directional). “Brief” means the observation plus necessary context; “developed” adds evidence and its limitation within the standard-post cap; “extended” needs an attachment or independently useful follow-up.

| Goal | Choose | Anatomy and length band |

|---|---|---|

| Introduce a useful idea or announcement | Root post | Concrete first line, context, optional action; brief or developed (directional). |

| Discuss someone else's artifact | Quote post | Credit the source and supply your own interpretation; brief. Don't use quotation as a pretext to target the author (directional). |

| Show an interface, object, or result | Image or image sequence | Brief or developed caption explaining what to inspect. Crop for legibility, redact private data, and give each image a purpose (directional). |

| Show a task changing on screen | Native video | Brief caption plus demonstration. Open on the relevant action; show input, observable output, and any consequential limitation. Caption speech and explain results without relying on audio (directional). |

| Share an observation or scene | Text | Brief if the detail stands alone; developed if the scene needs context. Technical subjects are welcome when the intended reader can follow them (directional). |

| Explain something requiring depth | Text attachment | Developed root summarizing the claim and why to open it; extended supporting text (directional). |

| Add a distinct update or answer a recurring question | Self-reply | Make each addition useful independently; don't split sentences merely to manufacture a chain. Add it when the information is ready, without a timer or promised resurfacing (directional). |

For polls, GIFs, spoilers, ephemeral posts, and messaging, open the reference's availability table before committing the publishing package. These aren't verified universal features in this evidence set.

## Launch and proof posts

Collect the audience and buying situation, promise, available proof with sources, desired action, destination link, campaign stage, and disclosure obligations. Record the response owner and approval boundaries. “No CTA” is a valid choice.

Use the format bands above for these post shapes (directional):

| Stage or shape | Native anatomy | Band and proof requirement |

|---|---|---|

| Launch day | Buyer problem, newly available capability, concrete example, availability boundary, action | Developed; link to the actual offer. Remove the product name temporarily: the post should still show or teach something. |

| Demo or artifact | Task, inspectable output, constraint, optional trial path | Brief caption with media; use the real product and identify staged input. |

| Customer outcome | Before and after, conditions, source, unresolved limitation | Developed; give the measurement window and denominator for any rate. Obtain quote and asset permission. |

| Founder decision | Decision, reason, cost or tradeoff, present status | Developed; first-person only for the author's own documented experience. |

| Objection answer | Actual buyer concern, direct answer, evidence, fit boundary | Brief or developed; offer a relevant next step without pressuring a public commenter. |

| Recap or lessons | What happened, what remains uncertain, next useful artifact | Developed or extended; distinguish interest from adoption and attributed sales. |

Put a link in the root when visiting is the intended action. Use a follow-up reply for supporting material when that keeps the root readable; publish that reply with the root when the CTA depends on it. Use a profile destination only after checking that the named link is present. Don't move links merely to satisfy an assumed suppression rule (directional).

Keep source, medium, and campaign parameters consistent; use utm_content to distinguish post or placement (official, Google URL Builder, 2026-09; undated guidance). Preserve campaign attribution downstream. Supplement tagged visits with optional self-reported discovery, retain raw answers, and allow unknown attribution (directional).

State founder, employee, affiliate, or sponsor relationships plainly. Put paid relationships where the reader will see them and check applicable disclosure requirements; the verified material doesn't establish a universal Threads sponsorship-tool rule (directional). Use required AI disclosure as specified below.

Contribute useful answers in relevant communities before asking for attention. Read custom feeds if available; don't treat them as guaranteed promotional inventory. Evaluate Champions by their actual contributions. Employee participation should be voluntary, individually written, and openly affiliated. Exclude engagement rings, bought reactions, copied praise, and mass tagging from the plan; don't portray these exclusions as measured enforcement thresholds (directional).

Use this illustrative launch rhythm; T is launch day and the offsets are planning choices, not ranking windows (directional):

| Day | Shape | Reply plan and measurement |

|---|---|---|

| T-7 through T-4 (directional) | Problem observation and useful community participation | Collect buyer language; distinguish substantive questions from general agreement. |

| T-3 through T-1 (directional) | Teaser with an inspectable artifact | Answer availability questions; verify the destination and prepare response ownership. |

| T+0 (directional) | Launch announcement and demo | Staff the first hours while discussion is active; track qualified replies, available Insights, and tagged visits. |

| T+1 through T+3 (directional) | Objection answer or bounded proof | Address recurring questions; record trial or service-fit interest and downstream activation. |

| T+4 through T+7 (directional) | Recap or new customer application | Return to unresolved conversations; report source-attributed outcomes with their observation window. |

Publish founder decisions from the founder's experience; let a brand handle speak through an accountable operator without pretending to be a customer. No founder-versus-brand reach advantage is established here (directional).

## Voice on this platform

Read the author's posts and the target conversation for sentence shape, warmth, and humor. Preserve natural contractions and fragments; use conventional punctuation unless the author already writes otherwise. Match the degree of informality without inserting typos, false edits, forced lowercase, or missing words (directional; Aborn).

| Context | Register decision |

|---|---|

| Warm or silly exchange | Extend the observed joke without attaching a sales pitch; retain the author's actual humor (directional). |

| Earnest or vulnerable exchange | Answer gently and directly. Don't introduce slang or comic exaggeration that trivializes it (directional). |

| Technical disagreement | State the claim and its conditions. Challenge the decision rather than assigning motives or insulting a group (directional). |

| Product announcement | Use the operator's normal vocabulary; replace corporate ceremony with an observable use case (directional). |

A defensible counter-intuitive claim includes evidence and scope. Polarization built from contempt or unsupported generalization doesn't improve that claim. Omit canned “Hot take:” prefixes; don't call them algorithmically penalized (directional).

Review the actual contribution instead of checking for an “AI sound.” Replace unsupported surprise or “nobody talks about” framing with the specific observation. Cut a contrast, three-part list, fragment, or closing question only when it adds no meaning or clashes with the author. These forms and em-dashes are not authorship evidence; formal prose can be natural for this speaker. Use prose for an ordinary reply and numbering when navigation genuinely needs it (directional).

Don't paste an Instagram caption stripped of hashtags. A Threads root must identify its own subject; a reply must address its parent, and a quote post must separate the quoted claim from the author's interpretation. Reconstruct context from the source, not presumed shared knowledge. Keep “I tried” attached to the person who tried it; describe what the demo actually shows when that is all the evidence establishes (directional).

Use supplied decisions and their costs for specificity (directional; Lees). Have a fluent editor review local idiom in context; this is an application of language research, not a measured Threads uplift (directional; MIDI). Neither these heuristics nor detector scores establish authorship; manuscript detector findings don't validate short-post detection (directional; Detector Study).

## Cadence and engagement

Start timing experiments on weekday mornings at 6–11 a.m. local; Thursday at 9 a.m. was the strongest slot in this sample (measured, Buffer Timing, n=2.5 million posts, 2026-02). Replace this starting point with account evidence. Open the reference for regional results; don't impose a worldwide evening or weekend rule.

Top Threads performers posted more often, but the report establishes no universal optimum and warns that higher frequency can reduce reach per post (measured, Buffer Engagement, n=Buffer-published social posts, 2026-03). Schedule evergreen material only when the publishing tool supports it, leaving live time for replies. Pre-approve themes and escalation boundaries (directional; Sendible).

## What gets suppressed

Don't assert fixed penalties for root links, Instagram cross-posting, technical content, or self-reply timing. The verified evidence doesn't support those rules. Sendible's weaker linked and polished-image posts describe its own account experiment, not a platform enforcement mechanism (directional; Sendible).

Meta announced a phase-in of personalized political recommendations; the announcement doesn't establish completed rollout or blanket throttling (official, Meta Civic, 2025-01). Choose political topics according to the brief and actual audience.

Meta requires its disclosure tool for organic digitally created or altered photorealistic video and realistic-sounding audio, and describes AI info labels (official, Meta AI Info, 2026-02, updated 2026-06). Preserve provenance metadata and check the finished media. That source doesn't establish a blanket AI-text label or caption reach penalty. Don't omit disclosure to pursue engagement.

## Examples

All briefs below are illustrative, not reports of actual businesses or outcomes. Treat their facts as supplied solely for the exercise. Bracketed fields require source-backed replacement before publication.

### Developer tool: demo

Illustrative brief: a runnable demo includes a dry-run preview from a demo database showing an index removal. Dependencies remain a manual check. The founder writes plainly.

> The dry run shows an index being removed before the migration runs. I've attached the preview from our demo database. You still need to check what depends on that index. The runnable example is here: [demo link]

Why it works: The demo rule ties the caption to visible proof.

The manual dependency check bounds the promise.

### Business software: customer proof

Illustrative brief: a customer approved reporting shorter invoice-preparation time and sharing workflow notes. The pilot covered domestic invoices only; exact measurements are still missing.

> [Customer] recorded invoice preparation falling from [before] to [after] during [window] with our billing tool. The pilot covered domestic invoices only. Their approved workflow notes are here: [source link]

Why it works: The proof rule requires conditions and source provenance.

The unresolved fields keep this draft unpublishable.

### Pottery creator: warm observation

Illustrative brief: a potter supplied a photograph of a bowl that wobbles when empty and stays steady holding garlic. It isn't for sale; their voice is dry and affectionate.

> My wobbly bowl has been promoted to garlic storage. It stays put with a bulb in it, which is more than I could say when it was empty. This one is staying in my kitchen.

Why it works: Register matching preserves the potter's supplied joke.

The consumer example uses an honest limitation and deliberately has no CTA.

### Local bicycle service: announcement

Illustrative brief: a repair shop offers advance photo assessment for commuting bikes. Concealed damage needs an in-person inspection; the supplied booking page explains which photos to send.

Weak draft:

> We're pleased to announce our new photo assessment service.

Revised:

> You can send us photos of your commuter bike before booking a repair. We'll say what we can assess from the pictures and what needs a closer look in the shop. The booking page explains which photos to send: [booking link]

Why it works: The launch rule makes the first line useful to a buyer.

The inspection boundary prevents a remote-diagnosis promise.

## Checklist

- [ ] First line carries the point; the post develops a central idea in the author's voice.

- [ ] Format and finished character count fit the verified constraints, including the actual URL.

- [ ] Every specific has provenance; customer permissions are recorded; no example placeholders remain.

- [ ] Meaning and author-voice review completed; no invented experience, manufactured errors, or punctuation quotas.

- [ ] CTA or deliberate no-CTA choice matches the goal; destination works; any required link reply is included in the publishing package.

- [ ] Topic choice fits; any newer feature was checked in the target account.

- [ ] Affiliation, sponsorship, and applicable AI disclosure checked; media provenance preserved.

- [ ] Asset brief, caption, image alt text, and video accessibility supplied where relevant.

- [ ] Reply owner and follow-up plan supplied; self-replies add information without artificial delay.

- [ ] Publishing is a downstream step; after authorized publication, verify assets, disclosure, and any promised reply link.

- [ ] Measurement distinguishes views and conversation from qualified interest and attributed outcomes.

## References

- [Meta Web](https://about.fb.com/news/2025/04/new-features-threads-web-experience/), 2025-04-24.

- [Meta Introduction](https://about.fb.com/news/2023/07/introducing-threads-new-app-text-sharing/), keyword-search update 2023-09-07.

- [Meta Text](https://about.fb.com/news/2025/09/attach-text-threads-posts-share-longer-perspectives/), 2025-09-04.

- [Meta Personalization](https://about.fb.com/news/2025/03/new-threads-features-more-personalized-experience-you-control/), updated 2025-07-22.

- [Meta Communities](https://about.fb.com/news/2026/06/meta-launching-new-features-500-million-monthly-threads-users/), 2026-06-16.

- [Meta Civic](https://about.fb.com/news/2025/01/meta-more-speech-fewer-mistakes/), 2025-01-07.

- [Meta AI Info](https://about.fb.com/news/2026/02/meta-prepares-for-2026-us-midterms/), 2026-02-19; updated 2026-06-01.

- [Buffer Engagement](https://buffer.com/resources/state-of-social-media-engagement-2026/), 2026-03-05.

- [Buffer Replies](https://buffer.com/resources/threads-comments-engagement/), 2026-02-24.

- [Buffer Timing](https://buffer.com/resources/the-best-time-to-post-on-threads/), 2026-02-04.

- [Sendible](https://www.sendible.com/insights/how-to-grow-on-threads-and-build-a-human-brand-presence?hs_amp=true), 2026-05-19.

- [Google URL Builder](https://support.google.com/analytics/answer/10917952?hl=en), undated; validation 2026-09.

- [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.

- [Lees](https://www.housingwire.com/articles/taste-guts-receipts-ai-writing/), 2026-08-07; [MIDI](https://arxiv.org/abs/2606.02147), 2026-06-01; [Detector Study](https://arxiv.org/abs/2608.26710), 2026-08-27.

- [Respondology](https://www.prnewswire.com/news-releases/47-of-consumers-hold-brands-responsible-for-toxic-and-spammy-comments-respondology-report-finds-302736755.html), 2026-04-09.

- [Detailed mechanics and evidence](reference/platform-and-evidence.md), validation 2026-09.

Re-validate when: feature names or routes change; limits or policies update; Meta publishes ranking details; vendors release new methods or reports.

Validated: 2026-09