X stopped counting replies toward the rewards programme, so a lot of small accounts stopped replying.

I think that’s backwards, and I spent a month testing it. 0 to 1,800 followers, and here’s the ratio that held every single day 👇 https://t.co/kYWznQpza2

Replies Don’t Pay. Reply Anyway.

X stopped counting replies toward the rewards programme. A lot of small accounts read that as a signal and stopped replying.

I think that’s exactly backwards, and I’ve spent the last month testing it.

Nobody knows who you are yet

When you have no followers, a post is a message with no distribution. It goes to a feed nobody is looking at. You can write the best thing you’ve ever written and it will do 40 views.

A reply is different. It borrows an audience that already exists. Under a post with real traction, your reply sits in front of people who have never heard of you and are already reading.

That’s the whole asymmetry, and it only exists while you’re small. It’s also why the advice inverts later.

Why people stopped, and why they’re wrong

The rewards programme counts verified Home timeline impressions. Replies are explicitly excluded. So if your goal is getting paid by X, replying is worthless.

But those are two different goals and people are conflating them.

Payouts need 500,000 verified impressions in 90 days. Almost nobody reading this is close. I’m not. Optimising for a threshold you won’t reach for months, at the cost of the thing that actually builds the audience, is a strange trade.

Build the audience first. Worry about what X pays for when the number is in reach.

My numbers

That chart is one month of a brand new account.

The shape is the point. The green bars are replies and they’re consistently many times taller than the blue ones. The account went from zero to roughly 1,800 followers in about a month on that ratio.

First full week: 451 replies, 60,300 organic impressions, 486 profile visits.

The number I actually care about is followers per reply, and it stayed remarkably flat. One day I wrote about 120 replies and gained around 185 followers. The next day, 55 replies and around 86 followers. Both land near 1.5 followers per reply.

Same ratio, different volumes, day after day.

What that ratio actually means

It’s good news and bad news in the same number.

Good: it’s predictable. It scales with hours, not luck. There’s no viral moment to wait for.

Bad: it doesn’t compound. The day I stop replying, growth stops that day. Not gradually. That day. It’s a direct exchange of hours for followers, and the machine stops the moment you stop feeding it.

Anyone selling you reply-based growth as a flywheel is selling you a treadmill.

The part that gets accounts killed

Volume is not the strategy. Volume is what triggers the filter.

X’s own search policy names repetitive content and bot-like behaviour as triggers for visibility filtering. Reported thresholds circulate — dozens of replies an hour, identical short responses pasted across threads — though those come from third-party testing rather than X.

The failure mode for a reply strategy is specific and it’s called a ghost ban: your replies still exist, but they get sorted to the bottom of the thread or tucked behind “Show more replies.” You keep working and your reach quietly flatlines. X’s own DSA transparency reporting puts the impression drop for filtered accounts at over 80%.

And this is where AI-generated replies bite. Not because they’re machine-written — because they’re interchangeable. Restate the post, add a generic observation, end with a question nobody asked. That reply could be pasted under a hundred other posts, which is precisely the pattern the filter is built to catch.

If your reply would fit under any post on the timeline, it isn’t a reply. It’s a signal.

The filtering is most of the work

I scan roughly 30 posts to find 4 worth replying to.

The other 26 get nothing, because the only thing I could add is agreement, and agreement is worth nothing to the author and nothing to anyone reading.

What makes one worth answering:

It’s minutes old, not hours. A reply under a 20-minute-old post gets read by the author and everyone arriving after. The same reply three hours later effectively doesn’t exist.

It has traction but few replies. High views with 40 comments means you’re comment 41 and invisible.

You have something specific to say. This is the real filter, and it’s the one that also keeps you out of the spam bucket.

When to flip the ratio

The asymmetry that makes replies work is that you have no distribution of your own. Once you do, it fades.

At some point your posts start reaching people without you going to find them, and every hour spent replying is an hour not spent on the thing that now compounds. That’s the moment to shift the percentages — not before, and not because a rewards programme told you to.

I’m not there yet. I’m still mostly replying.

What I’d tell you to do tomorrow

Pick one community where people actually build things. Be early — minutes, not hours. Skip the 26 out of 30 where you’d only be agreeing. Never send a reply you could have sent to someone else.

It’s unglamorous, it scales with hours rather than luck, and after a month it’s the only thing I can point to that demonstrably worked.

Finding those 4 posts out of 30 was the part that ate my time, so I built a Chrome extension that does the scanning. It scores posts in your timeline as you scroll and highlights the few worth answering — how fresh, how much reply room is left, whether it’s in your niche. Local, deterministic, no AI writing anything for you. winnowapp.app