How the Threads feed mixes followed and recommended
Meta describes the Threads feed as an ML-ranked blend of followed and unfollowed accounts. What changes when you write for the half that has no context.
Meta publishes a ranking page for Threads, which is more than most platforms offer. Its description of the Threads feed is short and specific: a machine-learned mix of content from accounts you follow and accounts you don't.
That single sentence carries an operational consequence most Threads advice skips. Every post you publish is evaluated twice, by two audiences with nothing in common, and the version that works for one is frequently the version that fails for the other.
Two audiences, one post
Your followers arrive with everything. They know the account, they have seen last week's post, they recognise the running joke, they know which product you sell and why you keep mentioning the warehouse. Context is free.
The recommended half arrives with nothing. No memory of the account, no reason to extend goodwill, no idea whether "the warehouse thing" is a bit or a business. They get one post, out of order, between two posts from people they actually chose to follow, and they decide in about a second.
Most brand accounts write exclusively for the first group, because the first group is who they picture. It is the group that replies. It is the group whose names appear in notifications. And it is the smaller and slower-growing of the two, because it can only grow via the second one.
What the recommended half doesn't have
Four things, and each maps to a specific failure in a post.
No subject. A post that opens with a pronoun referring to a previous post is invisible to a stranger. "This is why we stopped doing it" is a complete sentence and zero information. Name the noun in the first line, every time, even when it feels redundant to the people who already know.
No history. Running bits, recurring characters, in-jokes and long-form arguments all require a one-clause gloss for the newcomer. Not an explanation, a gloss. "Week four of rebuilding the shop van" does the work in six words and costs the regulars nothing.
No goodwill. Followers forgive a weak post. Strangers do not evaluate the post against your average, they evaluate it against the two posts either side of it in a feed they curated themselves. Your median post is competing with other accounts' best.
No reason to check your profile. The post has to be self-sufficient. If the payoff requires a click to the bio, a pinned post, or a previous thread, the payoff does not exist for most of the recommended audience.
Rewriting for the stranger without ruining it for the follower
This is the part that trips people up, because the naive fix is to over-explain and the over-explained version bores your existing audience into silence, which costs you the engagement that got you recommended in the first place.
The reconciliation is that context should be carried by concrete nouns rather than by preamble.
| Written for followers | Written for both |
|---|---|
| "Update on the thing from Tuesday" | "The bakery oven finally hit temperature" |
| "We tried it again and it worked" | "Third attempt at the resin pour, and the bubbles are gone" |
| "As promised, here's the result" | "Here's what a 40-hour ferment actually looks like" |
| "Big news" | "We're opening a second unit in March" |
The right-hand column is not longer. It is more specific, and specificity is what carries context without spending words on it. A follower reading "third attempt at the resin pour" loses nothing; a stranger reading it gains the entire setup.
Three more habits that serve both halves:
- Put the thing being discussed in the first eight words. Not a hook in the marketing sense, a subject in the grammatical sense.
- End on something answerable. Threads is conversational and replies are the currency. A stranger who cannot reply will not follow, and a question they can answer from a standing start is the cheapest possible entry point.
- Let the visual state the subject too. If a post might be encountered muted or half-scrolled, the frame should say what this is about. A generated title card or a legible first frame does more for the recommended half than any caption change.
Video on a two-surface feed
Video makes the split sharper, because a video costs the stranger more attention up front than a text post does.
Design so the first frame carries the subject. Not a logo, not a fade-in, not a slow establishing shot. The single most useful edit on a Threads video is deleting the first second, and the second most useful is putting a legible line of text on whatever frame is now first. Burned-in captions handle the rest.
Keep continuity where the followers benefit and legibility where the strangers do. A recurring generated character or a fixed visual style is worth a great deal on a feed where a stranger might see you three times across a month without following; character consistency is what makes those three encounters register as the same account rather than three unrelated posts. The production case for this is in Threads for brands.
Avoid anything that reads as reposted from elsewhere. Meta has been explicit about this on its other surfaces even if it has not published the same rules for Threads specifically.
The originality rules Meta publishes elsewhere
Worth reading, with the attribution kept straight. These are Instagram's and Facebook's published rules, not Threads' rules. Threads' own ranking page does not state them. They are the clearest available signal of how the company thinks about recommendations across the family.
Instagram's recommendations and originality guidance says accounts that repeatedly — 10 or more times in 30 days — post other people's content without a material enhancement become ineligible for recommendations; eligible recommended content is described as having no visible watermarks; and where identical copies are detected, only the original version is recommended.
Facebook's March 2026 original creators announcement says duplicative posts, minor edits, simple reactions and stitched clips without substantial addition will be deprioritised.
Read together, the pattern is that the recommended surface is the one Meta protects. Followed content is content you asked for; recommended content is content Meta is vouching for, and it is stricter about what it vouches for. If the same posture applies to Threads, the practical implication is that the cross-post you dropped in from another app is competing for the followed half only.
That does not make cross-posting worthless. It makes it a floor rather than a strategy, which is the same conclusion platform-native versus repurposed content reaches from a different direction. Export clean assets from your own timeline instead of pulling them back off another platform; nothing exported carries a Versely watermark on any plan, so the only foreign watermark in your pipeline is one you put there.
Measuring the split
You cannot see the ratio directly, but you can infer it, and the inference is more useful than the exact number.
- Follower growth per post is the cleanest proxy for recommended reach. Engagement can come from either half; new follows overwhelmingly come from the half that did not already follow you.
- Reach far above your follower count means the recommended half found the post. Track which posts do this and look for what they share. In most accounts the answer is dull: they were the self-contained ones.
- Reach at or below your follower count on an otherwise good post usually means it required context. That is a writing diagnosis, not a ranking mystery.
- Reply-to-reach ratio tells you whether the strangers who saw it engaged or bounced. A post with wide reach and thin replies reached people it did not land with.
Log these against your own median rather than an external benchmark, the same discipline described in brand video metrics that matter. Twenty posts of baseline before you draw a conclusion.
FAQ
Does Meta say what share of the Threads feed is recommended?
No. The published ranking page describes the feed as a mix of followed and unfollowed content without stating a ratio, and the ratio would presumably vary by user and session anyway. Design for both halves rather than trying to pin a number.
If a post needs context, should I just not post it?
Post it, and add the gloss. The cost of one concrete clause is close to zero and it converts an invisible post into a legible one. The posts genuinely not worth publishing are the ones where the entire payoff is membership in the audience, and those belong in replies rather than in the feed.
Do I need different posts for followers and strangers?
No, and running two tracks is how brands end up posting twice as much for the same result. The single-post fix is specificity: a concrete subject in the first line serves both audiences at once. Two versions of the same post is also close to the duplicate-content shape Meta penalises on its other surfaces.
Does this apply to Instagram and Facebook too?
The two-surface structure does, and Instagram publishes more about how it ranks than Threads does. The difference is that Instagram and Facebook attach monetization eligibility to their originality rules, so the cost of failing the recommended surface is higher there. Our overview of platform algorithms for brands covers where the mechanics diverge.