<b>Do engagement bots actually engage?</b>

<b>Do engagement bots actually engage?</b>

Leveling bots, XP, daily streaks — the standard "engagement layer." The evidence for them is thinner than the install counts suggest.

<b>What the data shows</b>
Comparisons of servers with and without leveling bots show a real lift in <i>message count</i> — often 20–40%. But when you decompose it, much of the lift is low-information messages: "gm", single emoji, grind-the-XP filler. On <i>retention</i>, the effect across studied cohorts was small and inconsistent; some servers saw the bot inflate vanity activity while real conversation flatlined.

<b>Why it happens</b>
XP rewards volume, and people optimize what's rewarded. You get more messages, not more <i>community</i>. The bot can also create a status hierarchy (level 50 vs newcomer) that subtly discourages the people you most need to retain.

<b>The caveat</b>
Bots vary enormously and most comparisons aren't randomized — servers that install engagement bots differ from those that don't. Telegram's bot ecosystem leans toward utility (anti-spam, captcha) over gamification, so a clean cross-platform comparison barely exists.

Open question: if you measured an engagement bot purely on <i>net new conversation threads started</i> rather than raw messages, how many would survive the cut?