How to get more engagement on X (Twitter)

In March 2023, X open-sourced its recommendation algorithm. Inside it is a model that scores every post by predicting how you will interact with it, then assigns each kind of interaction a weight and sums them into the number that decides what surfaces. The weights were in the repo. X has said the live values can be adjusted at any time, so treat the exact figures as the 2023 snapshot rather than today's production dial — but the relative priorities are the design intent, and they are blunt.

Read across the table. A like is the floor. A reply carries about 27 times the signal, and a reply the original poster answers back is the single highest positive weight in the whole model — 150 times a like. Two things you might assume matter, matter less than you think. A repost is only twice a like; the "a retweet counts as 20 likes" line you have read everywhere is simply wrong, because the repost weight is 1.0. And a passive like barely moves anything.

The negative side is the part almost nobody optimizes for. A mute, a block, or a "show less often" tap is weighted around −74, and a report is −369. One report cancels the signal from dozens of replies. Posting bait that annoys people does not just fail to help — it actively buries the post.

That reframes every tactic in the rest of this guide. Engagement on X is not a popularity score, it is a conversation score. Anything that starts a back-and-forth you take part in is worth orders of magnitude more than anything that earns a tap and a scroll.

The weight table points somewhere most tip lists never go. If a reply is worth 27 likes of signal and a reply the author engages with is worth 150, the fastest way to earn engagement is to spend it — by replying, early, with something worth answering, on posts your audience already reads.

The mechanics are specific. Reply in the first hour, before a post has hundreds of replies and yours is buried. Aim at accounts a notch or two bigger than yours in your niche, where the audience overlaps with the one you want. And add a real point — a counter-example, a number, a question the author would actually want to answer — not "great post." The target is that 150-weight outcome: a reply the original poster replies back to. One exchange like that does more for your reach than a day of posting into the void. It is a mechanism-backed tactic rather than a guaranteed-numbers one, so treat the returns as compounding, not instant.

The hard part is not writing the reply. It is finding the conversation while it is still early. Notifications only show you accounts you already follow, and scrolling the timeline hoping to catch the right post is not a system. The scalable version is monitoring: track the keywords, phrases, and handles your buyers actually use, and let the relevant posts come to you. That is the same habit behind catching brand mentions — the setup is in how to monitor brand mentions and the broader social media monitoring guide, and it is what we built ReplySocial to do. It is also how small accounts beat the follower math: their engagement rate ceiling is higher, so showing up in the right threads compounds faster.

The standard "10 tips to boost your engagement" lists are not wrong. They are just ranked by the wrong thing. Here is the honest version of the usual advice, sorted by whether it actually moves the number.

1. Write for replies, not likes. Asking a real question, taking a position someone can argue with, or leaving an obvious opening for a response beats a polished hot take that gets liked and scrolled past. Replies are the signal the algorithm pays for, and they are the part of X you can answer to turn a reader into a customer.

2. Post when your audience is online. Impressions are the ceiling on every engagement number — a great post seen by nobody engages nobody. Our best time to post tool reads the windows where a handle's audience is most active. On cadence, Sprout Social recommends roughly three to four posts a day, but the honest version is that posting more does not linearly buy more engagement. Consistency beats volume.

3. Use media, with realistic expectations. Tweets with images have historically out-performed text — Buffer's analysis found photos earned about 150% more retweets, and Twitter's own data once showed a 35% retweet lift — and video leads content-type engagement in RivalIQ's 2025 benchmarks. The "video gets 10x engagement" line you will see quoted has no traceable source; skip it. Media helps. It is not a cheat code.

4. Keep it tight. The highest-engagement length on X sits around 240-259 characters — a complete thought you can read in one glance. Padding to fill the limit lowers engagement. The mechanics of trimming without flattening the message are in writing tweets that fit.

5. Use one or two hashtags, at most. Past that the math is weak — extra hashtags read as spam, and the algorithm does not reward them the way it did in 2015. A single relevant hashtag to join a live conversation is fine; five to "maximize reach" is a tell.

Do all five. Just do not mistake them for the main move — none of them beats the reply-first system in the section above.

Two shortcuts get sold hard, and both make your numbers worse in the only way that counts.

Engagement pods and reply groups. The pitch is a group that likes and replies to each other's posts on cue to "trick the algorithm." X's Platform Manipulation and Spam policy bans exactly this — it prohibits "coordinating to exchange engagement in any X features, such as Likes, Polls, Replies, Reposts, Lists, Views, or Follows." Enforcement runs from anti-spam challenges to visibility limits to outright suspension. Set the rules aside and it still does not work: pod engagement comes from people outside your audience who will never become customers, and it trains the algorithm on a signal that does not match who actually wants your posts. The math is weak.

Bought engagement and bots. Fake likes and reply spam inflate the count on a post no real person cared about, and bot followers sit in your denominator dragging your real engagement rate down without ever engaging. We see this directly: when we score reply authors for bot likelihood with BotBlock, a meaningful share of the engagement on viral threads comes back flagged as automated. Before you celebrate a spike, it is worth knowing how much of it is real. Spot-check any handle with the free bot checker, and the 30 signals that flag a bot cover the manual version.

The thread connecting both: engagement you can buy or trade is engagement that does not compound. The kind that does is a reply from someone who actually wanted to talk to you.

"More engagement" is the right goal stated one level too shallow. A raw count climbs just because you posted more. What you actually want to know is whether the engagement is real and whether you can do anything with it. Three things to watch instead of the like tally:

1. Your follower-based rate, against your own tier. The rate tells you whether each post is landing, where the count only tells you that you showed up. What counts as good depends entirely on your follower size — the benchmarks by tier are in the engagement rate guide, and the engagement calculator tells you which percentile you hit.

2. Conversation rate — replies divided by followers. This is the metric that predicts the back-and-forth the algorithm rewards, and it is the slice of X that turns into pipeline. A post with 20 replies you can answer beats one with 200 likes you cannot do anything with.

3. Profile clicks and follows from a post. The algorithm pays 12 for a profile click that becomes a like or reply because it means you earned a new reader, not a passing tap. That is engagement that grows the audience instead of entertaining the one you already have. If followers are the actual goal, the longer game is in the X growth playbook.

The engagement worth chasing is the reply you can answer. Everything else is a number that feels good and does nothing.