This guide explains how the X (Twitter) algorithm decides who sees your posts, translated into actionable knowledge for content creators. For the full technical reference, see reference/algorithm.md.
Every time someone opens their X feed, a pipeline processes thousands of candidate posts to select the ~50 they'll see. Your post competes for those slots.
Your post is published
↓
Shown to a small slice of your followers (~10-20%)
↓
Phoenix AI predicts: will each viewer reply, bookmark, dwell, repost, or disengage?
↓
Based on predictions + early actual engagement → expand or suppress
↓
If expanding: test with a small out-of-network audience
↓
If that works too: broader distribution → "viral"
The key insight: the algorithm doesn't decide if your post is "good." It predicts whether specific users will engage with it. Your job is to create content that triggers high-value engagement signals.
Not all engagement is created equal. The algorithm weights different actions dramatically differently:
▲ Algorithm Value
│
Author replies to comments ─────────── 75-150× ████████████████████████████████
Threaded conversations ──────────────── 13-27× ████████████████
Reposts / Quote tweets ──────────────── ~20× ██████████████
Bookmarks ───────────────────────────── ~10× ██████████
Dwell time (time spent reading) ─────── High █████████
Video completion % ──────────────────── High █████████
Profile visits ──────────────────────── Medium ██████
Likes ───────────────────────────────── 1× █
│
▼
The takeaway: A post with 10 genuine replies is worth more to the algorithm than a post with 270 likes. Optimize for conversations, not likes.
The first 30-90 minutes after posting determine your post's fate. Here's why:
- Initial test: Your post is shown to a small sample of your followers
- Signal collection: Phoenix measures reply rate, dwell time, bookmark rate, negative signals
- The decision: Based on these signals, the algorithm either expands distribution or lets the post die
What this means practically:
- Post when your audience is active (not midnight)
- Be available to reply to comments in the first hour (author replies = 75-150× multiplier)
- Never post and disappear
- Front-load your best content in the hook (determines whether the test audience engages)
Your followers are the launchpad. To reach people who don't follow you:
- Strong initial signal from your followers (replies, bookmarks, dwell)
- Embedding match — Phoenix uses AI to determine if your post's topic matches non-follower interest clusters
- No negative signals — no blocks, mutes, or "not interested" from the test audience
- Expansion testing — small batches of non-followers see it, and if they engage, more batches follow
- Compound effect — each round of engagement triggers the next expansion
For small accounts: This means your followers are your most important asset, even if you only have 50. Their engagement in the first hour is the gateway to broader reach.
Impact: 30-50% reach reduction
X wants to keep users on the platform. External links signal that you're sending people away.
Fix: Put links in a reply to your own post, or use "link in bio."
Impact: Reduced distribution + trust penalty
"Like if you agree", "Follow for more", "RT to spread the word" — the algorithm classifies these as low-quality patterns.
Fix: Ask genuine questions that invite real discussion.
Impact: Asymmetric — one "not interested" outweighs many likes
Blocks, mutes, reports, and "not interested" clicks do far more damage than equivalent positive signals do good. The algorithm is conservative — it would rather not show your post than risk a bad experience.
Fix: Stay on-topic for your niche. Don't be gratuitously provocative. Deliver on what your hook promises.
Impact: Wasted multiplier opportunity
Author replies to comments carry a 75-150× weight. If you post and don't reply to comments, you're leaving the single biggest algorithmic boost on the table.
Fix: Block 30-60 minutes after each post to reply to every comment.
Impact: Broad suppression
More than 10 posts/day, identical content, automated engagement (likes/follows), unnaturally regular posting intervals — all trigger spam classifiers.
Fix: Post quality content at human-realistic intervals. 3-5 posts/week is optimal for most accounts.
Based on the scoring model, posts that perform best share these characteristics:
- Insight-dense content people read slowly
- Surprising information that makes people pause
- Well-formatted with line breaks (easier to read = longer reading time)
- Screenshots, tables, or code that require examination
- Genuine questions with specific enough framing to answer
- Debatable takes that invite "yes, but..." responses
- Personal experiences that prompt "same here" or "I did it differently"
- Ask for recommendations or preferences
- Actionable reference material (tips, commands, configurations)
- Curated lists of resources
- Step-by-step tutorials
- Benchmark data people want to revisit
- On-topic for your established niche
- Hook matches the content (no bait-and-switch)
- Value delivered, not just promised
- Authentic voice (not corporate/AI-generic)
X is the only major platform where text outperforms video by ~30%. This is because:
- X's culture is text-first (unlike TikTok, Instagram, YouTube)
- The algorithm can analyze text content more deeply for relevance matching
- Text posts have lower "skip" rates than video (less commitment to start reading)
- Dwell time on text is a stronger signal than passive video viewing
The optimal strategy: Text-first posts with supporting visual media. The text carries the algorithm score; the visual boosts dwell time.
Threads (multi-tweet sequences) get approximately 3× more engagement than equivalent single tweets because:
- More dwell time — multiple tweets to read through
- The algorithm treats threads as one unit — no diversity filter penalty
- Each tweet can surface independently — individual tweets from your thread can appear in other people's feeds
- Reply surface area — people can reply to any tweet in the thread, multiplying conversation opportunities
- Extended lifetime — active threads maintain visibility longer than single posts
The sweet spot is 5-7 tweets. Shorter than 5 doesn't justify the thread format. Longer than 10 risks drop-off.
Premium (verified) accounts receive a direct scoring boost:
| Tier | Boost |
|---|---|
| Free | 1× (baseline) |
| Premium | 2-4× distribution |
| Premium+ | Up to 10× in some cases |
This is applied at the scoring stage — it's a multiplier on your post's final score. For small accounts trying to grow, Premium is the single highest-ROI investment.
-
Before posting (10-15 min): Reply to 10-15 posts in your niche. This warms up the algorithm's model of your interests and makes you visible to potential audience.
-
Post (5 min): Publish your content at peak time for your audience.
-
After posting (15-30 min): Reply to every comment on your post. Substantive replies, not just "thanks!" This is the 75-150× multiplier in action.
-
Throughout the day (5 min scattered): Reply to 5-10 more posts in your niche. Build relationships. The algorithm notices your engagement patterns.