How the Threads algorithm works
What drives reach on Threads: freshness, replies, how long a post lives. A breakdown across 46,000 posts with measured response.
Meta doesn't publish its ranking rules, so the only honest thing to offer is observation. Here's ours: we collected 46,816 posts from 34,464 authors and looked at what happens to response.
About the data: this is an anonymized analysis of public posts. We compare medians and feature frequency within the sample; correlation does not prove that the feature itself caused higher response.
A post lives a long time
The most surprising thing in the numbers. The usual assumption is that a post collects reactions in the first few hours and then dies. On Threads it doesn't work that way.
| Post age | Likes (median) | Replies |
|---|---|---|
| under 2 hours | 1 | 1 |
| 2–6 hours | 3 | 2 |
| 6–12 hours | 5 | 3 |
| over 12 hours | 8 | 6 |
Response keeps growing even a day later. So the feed doesn't show only fresh posts: a post can take off the next day.
Practical takeaway: don't delete a post if it didn't move in the first hour. And don't publish the next one right away — it will compete with the previous one for the same attention.
Most posts get nothing
A quarter of posts collect not a single like. A third — not a single reply.
| Share of posts | |
|---|---|
| Zero likes | 25% |
| Zero replies | 30% |
| 50 likes or more | 18% |
| 500 likes or more | 3% |
Worth understanding before you get upset about your own numbers. A post with ten likes isn't a failure — it's the middle of the platform.
What lifts response
We compared posts by feature and looked at the median number of replies.
| Feature | With | Without |
|---|---|---|
| Question at the end | 6 | 2 |
| Broken into paragraphs | 4 | 2 |
| A number in the text | 3 | 2 |
| Emoji | 3 | 2 |
A question at the end gives three times more replies. It's the strongest of the simple moves — provided the question is genuine and about a specific thing in the post. A routine "how about you?" is visible from a mile away and doesn't work.
Paragraphs double the response. The reason isn't the algorithm but the human: a solid block of text gets scrolled past in the feed.
Replies matter more than likes
Judging by how the feed behaves, discussed posts get shown more widely than merely liked ones. The logic is clear: a reply is a more expensive action than a like, and the platform gains more by promoting what sparks conversation.
Hence the practical takeaway: a post that ends with a question works better not because the algorithm likes questions, but because people answer them.
Consistency
An account that posts and comments every day gets shown more actively than one that appears once a week. But a sharp spike — a hundred actions in an hour — the platform reads as automation and limits your reach.
We tested this on ourselves: our data collector made about 6,500 requests to the feed in a day, and the account got limited — the "For you" feed wouldn't load for three days.
A steady pace is safer than bursts, and it wins on response too.
What the algorithm dislikes
- Strings of plain hashtags. Posts containing the
#symbol showed weaker response in our sample. That does not apply to the built-in topic tag: Meta says posts with tagged topics generally receive more views. Choose one precise topic instead of a list of hashtags. - Outbound links. The platform has no interest in sending readers away, so posts with links get shown less. Better to put the link in a comment or in your profile.
- Sameness. Posts identical in form, one after another, stop being shown even to followers.