AI Mastodon reply generator.

Mastodon isn't one network, it's thousands of servers that agree to talk to each other. A reply you write is delivered to the servers that know about the conversation, which means the audience for any given thread is genuinely partial — you may not see all the replies, and not everyone will see yours. Reply counts are unreliable for the same reason.

There is no engagement ranking anywhere in the stack. Timelines are chronological, and the only amplification mechanism is the boost. A reply doesn't get boosted often, so the point of replying is the conversation itself and the reputation you build on your server and among the people who follow you.

The default limit is 500 characters, though individual servers can and do raise it. More important than the limit are the community expectations: content warnings on anything sensitive, politics-heavy, or long; alt text on every image; no unsolicited pitching, ever. Servers have codes of conduct with actual moderators behind them, and a bad reply can get your whole instance defederated in extreme cases.

Mastodon's population skews technical, privacy-conscious, and allergic to marketing language. A reply that reads as though it came from a social media manager will be ignored at best. Sincerity and technical specificity are the currencies here.

Take the draft and remove anything that performs. No hooks, no rhetorical questions aimed at the audience rather than the person, no closing call to action. Then check whether the topic needs a content warning — if you'd hesitate to have it appear unexpectedly in someone's timeline, add one.

If you're replying across servers to someone you don't know, a short note of what you're responding to helps, since threads reassemble imperfectly across instances. Then stay for the reply. Conversations on Mastodon are slow, and the payoff is a small number of people who actually know who you are.