The Reply Hierarchy: Why 50 Replies Outperform 500 Likes on X

Most creators optimize for likes. Likes are the easiest engagement to generate: a quick double-tap, no thought required. But in X's engagement scoring hierarchy, likes sit at the bottom of Tier 1. Replies sit at the top.

The gap is not small. X's open-source algorithm reveals that the Heavy Ranker assigns disproportionately high weight to replies because they require the most cognitive effort from users. A reader who types a response has invested attention, formed a thought, and contributed to a conversation. The algorithm treats this effort as the strongest possible indicator that the content was worth distributing.

X's Phoenix transformer scores 19 engagement signals, but not equally. The hierarchy from the open-source code:

Tier 1 (article built around these):

1. Reply (highest weight): End with a statement people want to argue with, correct, or add to. Never neutral.

2. Retweet / Quote: One line must be quotable, worth screenshotting, makes the reader look smart for sharing.

3. Favorite: Hook + payoff must both deliver.

Tier 2 (structure drives these): Dwell time, profile click, share, click.

Tier 3: Follow, triggered by distinct POV and consistent voice.

Negative (distribution killers): "Not interested," Block, Mute, Report. These suppress future distribution, not just the current post.

Write Better Articles uses three ending strategies, each engineered for the reply signal:

1. The Arguable Statement. Take a position that reasonable, intelligent people could disagree with. "Most founders should fire their marketing team before hiring more salespeople." This forces readers who disagree to reply, and readers who agree to reply with their own evidence. Both outcomes generate reply signals.

2. The Complexity Admission. End by acknowledging what your article didn't cover or where your argument has edges. "This framework works for B2B. B2C is a different game, and I'm not sure these rules apply." Readers who see those edges reply with their perspective. Readers who disagree about the limitation reply with counterexamples.

3. The Moral Verdict. Close with a judgment the reader must take a side on. "So the question isn't whether AI replaces writers. It's whether you'll be the writer who uses it or the one it replaces." This forces an internal reaction that frequently becomes a reply.

What doesn't work: summaries ("In conclusion, we covered..."), explicit asks ("What do you think?"), and hedged endings that leave nothing to discuss.

A single reply is valuable. A reply chain is exponentially more valuable. When someone replies to your post and then a third person replies to that reply, the algorithm sees an active conversation. This cascading effect is why "reply to every comment in the first hour" is the highest-leverage post-publish action.

The Algorithm Brief's "First reply" field is specifically designed for this. It provides the ideal first reply to post immediately after publishing: the most provocative distillation of your central claim. This seeds the reply chain before organic replies arrive, signaling to early readers that discussion is welcome.

50 genuine replies, with 20 reply-to-reply chains, will outperform 500 likes and zero replies. The math is clear. The question is whether your article's ending is engineered for it.