People sleep. An account writing comments at four in the morning every day without a break differs from a person not in its wording but in its schedule — and that difference needs no content analysis to spot.
Why the schedule gives more away than the text
Text is easy to vary, especially with a language model. A schedule is harder to fake: either there is a rhythm or activity is flat across twenty-four hours, and the second does not occur among real users.
The same applies to even intervals. An action exactly every five minutes is not a person, however good each individual action is. Pauses have to differ in length.
What a believable schedule looks like
- A sleep window. Six to eight consecutive hours with nothing happening.
- Unevenness within the day. A morning and an evening peak, a dip in between.
- Different windows per account. Ten accounts falling asleep and waking together are linked by that alone.
- An occasional missed day. Perfect seven-day regularity is not a human trait either.
Time zones
The schedule should agree with the number's region and the proxy's. An account with a European number active on the other hemisphere's clock contradicts itself — see the account fingerprint.
What is lost and what is gained
It looks like the night pause throws away working hours. In practice it throws away very little: channels and chats are quiet at night, there is nothing to answer, and the first comment under a post published at 3 a.m. is read by nobody.
What is gained is account lifespan, which is the genuinely scarce resource in this work — see how many accounts you need and recovering a locked account.
How Neurogram does this
Everything above is manual work that runs into the number of hours in a day. Neurogram does it on your own accounts and without you: it collects channels and audiences, writes comments and replies with a language model, and keeps limits and quiet hours so the accounts survive. See pricing or look at your case.