The community playbook: Discord and Telegram

a community is a retention and feedback engine, rarely an acquisition channel. if your consumer AI app has active users who already want to talk to each other, a Discord or Telegram home compounds retention, word of mouth and product insight. if it does not, you are about to build a ghost town that signals a dead product. this page is the honest frame before you open one.

Retention and feedback, not acquisition

the job of a community is to keep the users you already have and to tell you why the rest left. it almost never brings new users in at meaningful volume, and starting one to “drive growth” is how most of them end up empty.

the people who run these full time agree. community managers rank retention as the top benefit of running one, per Staffordshire Business School’s 2025 research, ahead of anything acquisition-shaped. that matches what you see when you sit in a healthy server for a week: members answering each other’s questions, sharing what they made, telling you within minutes when something breaks. every one of those moments makes the product stickier and your roadmap sharper. none of them shows up as an install.

Discord is where AI power users already live

if your users are AI power users, they are probably on Discord already, and that is where the deepest version of this works. Midjourney famously ran its entire product through a Discord server: generation, feedback, culture, all of it in public channels. you do not need to go that far, but the lesson holds. Discord rewards products that give members something to do together, not just something to read.

Telegram is the other default home, and it plays a different game: broadcast reach with almost no friction, especially in markets where Telegram is the default messenger. the choice between them, and the third option of a subreddit you do not own, deserves its own decision, and borrowed rooms like the SAFE reddit playbook covers are often the smarter first move anyway.

The real cost: a product surface you run forever

a community is not a channel you test for a month. it is a product surface you now operate indefinitely. members expect answers, moderation, a reason to come back, and a host who shows up. an unanswered question sits there in public. a spam wave you slept through sits there in public. the silence after your last announcement three weeks ago sits there in public.

that cost is fine if you price it in. it is fatal if you thought you were launching a marketing channel you could pause. before you open a server, decide who spends time in it every day for the next six months. if the honest answer is nobody, do not open it.

The decision test before you start one

start a community only if your users already show demand for one. the test: can you point at real users talking to each other about your product somewhere you can see, replies under your posts, a thread you did not start, reviews that answer other reviews, people asking “is there a Discord”? that is pull. everything else is you pushing.

if the pull is there, build the home and the retention loop that feeds it, which sits closer to product-led growth than to any ad channel. if it is not there yet, a changelog, a newsletter or a broadcast channel does the update job without the empty-room risk, and you can graduate later.

if you want a straight answer on whether your app has earned a community yet and where it should live, the breakdown call is free. Let’s talk Growth.