Blog
Field notes on bots
Plain-language writing on automation patterns, scoring, and what the evidence actually shows.
The bot-and-grift glossary: every term in this series, in one place
Heard a term like "astroturfing" or "sockpuppet" somewhere and want the two-sentence version before committing to a full post? Here's every term from this series, defined plainly, with a link to the deep dive.
Read post →
Does the blue checkmark actually mean an account isn't a bot?
The checkmark used to mean something specific: a platform employee looked at proof of identity. Now it means eight dollars a month and a working phone number — and that's a very different bar to filter bots against.
Who actually pays for a bot farm, and why the math works out
A thousand fake likes costs less than a coffee. That price tag explains almost everything else in this series — it's not that bot operators don't care about getting caught, it's that getting caught costs them almost nothing.
AI-written replies aren't a novelty anymore. Here's what's actually changed.
The LLM-assisted account used to be the rare, cutting-edge bot pattern. It's not rare anymore, and the honest update is less about detection getting smarter and more about the sheer volume tipping over.
How much of X is actually bots? What the research actually says
Everyone's got a number — 5%, 15%, "half the internet." Here's where those figures actually come from, why they disagree so wildly, and what's genuinely known versus guessed.
Before you audit anyone else, check your own followers
It's easy to go looking for bots in someone else's audience. It's a lot more useful — and a lot more uncomfortable — to check your own first.
You've spotted a bot. Now what?
Mute, block, or report — they're not interchangeable, and picking the wrong one either wastes your time or lets something worse slide. Here's how to actually decide.
How to spot a bot yourself, before you ever need a tool for it
You don't need software to catch most bots — you need to know where to look. Here's the checklist we'd run through ourselves, and the honest point where doing it by eye starts to fall apart.
The account that says the same three things to everyone
A template reply spammer drops "Great point!" / "This 100%." / "Couldn't agree more 🔥" under completely unrelated posts — clearly not reading anything before it replies.
The account that never has an original thought (and why that's the point)
A retweet amplifier almost never says anything itself — it just pushes other people's posts, over and over, all day. Here's why that's one of the easiest bot patterns to spot once you know what you're looking at.
The account that talks to everyone and gets nothing back
A reply farm replies constantly — dozens, sometimes hundreds of times a day — but almost nobody ever replies back. That one-way silence is the tell.
The account wearing its third name this year
A sockpuppet wears its third name this year. Real people change a display name for fun; they almost never change the actual handle two or three times while posting at a pace no human keeps up.
The account whose replies are a little too smooth
An LLM-assisted account's replies have no typos, no slang, no stray thought — every one reads like it was written by someone who'd never typed something dumb on their phone at 11pm, because it probably wasn't a someone at all.
The account that's really just an ad wearing a person costume
A grift funnel is an ad wearing a person costume: blank or default avatar, a bio that's basically a pitch, and followers that look just as thin. Built to sell you something, not to talk to you.
The account following 40,000 people that 200 people follow back
A follow-for-follow growth account builds a following the cheap way — following tens of thousands and hoping a fraction follow back. Real audiences grow two-directionally; this math is almost always visibly lopsided.
The "American" account that keeps posting at 4am Eastern
A fake-location persona claims Texas or New York in the bio while posting at hours that fit somewhere else entirely. Bios lie for free; sleep schedules are hard to fake.
The account that keeps saying what another account just said
A coordinated campaign isn't one odd account — it's a dozen different handles posting near-identical text within minutes of each other. One of the clearest bot signatures there is, because it's genuinely hard to fake by accident.
The account built to publish, not to talk
A broadcast operator posts huge volume, but the second you look for an actual conversation, there isn't one. This type isn't hiding — it's just not built for two-way engagement.
The account that sat quiet for a year and then suddenly woke up
A bought or repurposed account wasn't born a bot — it was bought. A real, aged, previously-dormant profile gets picked up, handed to a new operator, and put back to work. Here's how to spot one that's changed hands.
How bot detection works (without the hype)
BotShark turns “this account feels off” into a score you can inspect — behaviour, evidence, limits.
The account that never seems to sleep
A 24/7 scheduler has no night. Real people go quiet for a few hours; these accounts stay flat around the clock, and that sleepless rhythm is oddly easy to spot once you plot it out.