Skip to content
AutomationAutopilot

Autopilot

Autopilot is what actually runs automation. Reply modes decide what is allowed; autopilot decides what is happening.

With autopilot off, nothing automatic happens: nothing is sent and no draft is written by itself. Generate always works, whatever autopilot is doing.

Three switches, all of which must be on before a reply is sent without a person


Autopilot writes to real people. First make sure that:

  1. the creator’s settings are tuned on real chats;
  2. your team has worked in Draft only long enough that drafts rarely need changes;
  3. you have decided which fans may be served, and set their reply mode.

See Reply Modes and Autonomy.


Autopilot is turned on in one place: the Autopilot tab of the side panel.

  1. Open the platform tab you want it to work in.
  2. Open the side panel → Autopilot.
  3. Under The autopilot works on, switch on the groups of fans it should serve. Each switch shows how many fans are in the group; one that says 0 fans, or is greyed out, has nothing to do.
  4. Under Where it works, choose Lock on one fan (the open fan’s name is shown in the option) or All fans in the list.
  5. For All fans in the list, choose Only answer messages from (see below).
  6. Press Start in the top row.

The Autopilot tab before anything is engaged: which classes it would serve, how many fans are in each, and what it would do

If no fan is set up yet, both switches show 0 fans. Set a fan’s reply mode on the Fan tab, or the creator’s default: click the creator’s name at the top of the panel to open their dashboard page.

It starts right away, and only from that moment. Nothing older is answered because of the start itself.

While it is on, the toolbar icon shows AP, so you can see it from anywhere in the browser.


Two big switches decide which fans this session serves. They do not change anyone’s reply mode; they only say which groups autopilot takes on right now.

SwitchWhat it does
Draft onlyWrites the reply and leaves it for you to send
Draft and sendWrites the reply and sends it without waiting for you

Each row starts with how many of this creator’s fans are in that group. That answers “who will this affect”, so it is the biggest thing in the row. Under the switches, one sentence says what the combination will do, and one more says how many fans are on Manual and will never be touched.

Turning on only the second switch makes sense: leave my hand-written fans alone, and handle the automatic ones.

A switch is greyed out, with one line of explanation, when your manager turned that group off for this creator or when no fan is in that group yet.


Open the fan, set the reply mode you want for him (the switch is inside this option) and press Start. Autopilot works with that fan and nobody else.

The lock is on the person, not on the window. You can read other chats in the meantime and nothing happens in them. When you come back, the panel is where you left it.

His chat must be open for a reply to go out, because a message can only be typed into an open chat. So if you go elsewhere, autopilot waits and says so; when his chat is on screen again, it continues. It does not switch off, and moving around costs nothing.

Every agency should start here. Your chatter sees every message go out, live, and can stop at once.

Good to know. When autopilot sends by itself, it presses the platform’s Send button from inside the browser, and a platform can technically tell that apart from a person’s click. When your chatter presses Send, it is an ordinary click.

Autopilot goes through the chat list and serves everyone who is waiting: opens a chat, reads it, replies, moves on. Each fan is handled by his own reply mode, and fans in a group you did not switch on are passed over.

This takes over the browser tab. It opens chats by clicking them, so your team cannot work in that tab while it runs. Use a separate window or a separate browser profile.

Before you start, this option shows how many chats are in reach, how many are unread (on OnlyFans and Fanvue; Fansly gives only the number of chats) and how long a full pass will take. So you can plan the session instead of wondering whether it is stuck. The chat list loads as you scroll, and autopilot only reaches the chats the platform has already shown, so on a very long list, scroll it down first.

Availability. Going through all chats works on OnlyFans, Fansly and Fanvue. On Fansly the panel cannot see how many chats are unread, so it shows “at least N” chats instead.


The most important setting for All fans in the list. It limits how old a message can be for autopilot to answer it:

WindowWhat it means
Last 6hOnly today’s active chats. The most careful choice
Last 24hThe default, and right for most agencies
Last 3 daysCatching up after a weekend
No limitEvery unanswered chat, however old

Choose carefully. On an account with history, No limit means every chat ever left unanswered gets a reply, possibly hundreds at once. The panel warns you before you choose it.

This window only applies to All fans in the list. A fan you locked on is always served: you picked him yourself, which says more than any timestamp.


Autopilot acts on exactly one condition:

The last message in the chat is from the fan.

That is all. It means the same as “a new message came” and “the creator has not replied yet”, and one rule instead of three keeps its behaviour predictable.


The Autopilot tab becomes a live screen. Its top turns green and stays green while it works, so “is it sending for me right now” is the first thing you see:

  • Who it is serving now, and what it is doing: opening the chat, reading, writing, waiting. When it is locked on a fan whose chat is not on screen, it says whose chat it is waiting for.
  • Progress through the current pass.
  • Counters for this session: sent, drafted, skipped, failed. They start from zero each time you press Start.
  • The activity log, which opens by itself when you start.
  • What it is running with, in a line under the top row.
  • Stop, in the top row, where Start was.

The Autopilot tab while it runs: who it is serving, the session counters, and what it has just sent

If you lock on a fan the session will never touch (because he is on Manual, or his group is switched off), the panel says so instead of looking busy.

While it goes through all chats, the Draft tab hides its controls and shows “The autopilot is driving this tab.” Chats open every few seconds, and buttons pressed by hand would act on whichever chat opens next.


  • The emergency stop stops all automatic sending for the creator, whatever else is on and whoever is working. It is in Creator Studio and on the panel’s Draft tab (Pause), on purpose not on this tab. A chatter who needs to stop their own browser presses Stop in the top row.
  • A reply is held back when… The four situations that hold a message for a person: the fan may be in crisis, the fan is complaining, a banned topic, and the middle of the creator’s night. Folded away with a count (“4 of 4 on”). Managers change them; everyone can see them, because they decide whether a reply goes out.
  • The pace: at most 20 messages an hour, fixed.

Every message is saved: who it went to, what was sent, when, and the reason for every chat that was opened and then passed over.

Fans skipped without being opened at all (because their reply mode is not in a group you switched on) are counted, not listed. On a team where most fans are manual, a line for each would fill the log and push the real sends out. Their numbers are in the counters and on the two group switches.

The log is folded with its count while autopilot is off and opens by itself when you press Start.

Read it every day for the first two weeks. Nobody watches a full pass live. The log is how you review it, and how you learn which chats should not have been served.


Autopilot works at a human pace, on purpose:

The countdown before an automatic send, with the button that stops it

  • a short random pause before each message, shown as a countdown with a Cancel button;
  • a minimum gap between messages (8 seconds);
  • a strict limit: 20 messages per hour.

This is not a technical limit, it is the point. An account that answers forty chats in ninety seconds does not look like a person.


Autopilot stops cleanly and says why:

What happenedWhat the panel says
You press StopYou switched it off.
The tab it worked in is closedThe tab was closed.
The page reloadsThe page reloaded.
You change the creator you write asYou switched creator.
The browser restartsThe browser restarted.
Five attempts fail in a rowFive attempts failed in a row.
Your monthly drafts are nearly used upIt stops; the panel shows the code quota-floor

Going to another chat does not stop it. A session is locked to a person, so where you browse does not matter.

Autopilot never survives a restart, a reload or a closed tab. This is on purpose: automation that looks like it is running while nobody runs it is the most dangerous state there could be. You always turn it on again yourself.

Your switches, choice of where it works and the time window are remembered, so turning it on again is one click. The lock on a fan is not: it belongs to one session.


On top of the checks in Reply Modes and Autonomy, autopilot has its own:

  • Both group switches start off, and a manager must allow automatic sending before it can be switched on at all. A fresh install does nothing.
  • Chats where the creator never replied are skipped. Either it is a brand-new fan who deserves a person, or something is not read correctly. Skipping costs one manual reply; not skipping could cost an incident.
  • The same messages are never answered twice. Starting autopilot over a chat that already has a draft does not make a second reply.
  • Every message checks the fan again right before sending. If the chat on screen changed, the send stops.
  • Repeated failures turn it off, instead of carrying on at volume.
  • Your drafts are protected. Autopilot stops before the monthly drafts run out, leaving some for the drafts your team asks for by hand. Generate is never blocked.

Phase 1 — watched, drafts only, one fan. Lock on one fan with only Draft only switched on. Nothing is sent; your chatter reads what the AI writes.

Phase 2 — watched sending, one fan. The same lock, with Draft and send on. Your chatter watches every message go out. Do this for several shifts.

Phase 3 — review. Read the log from start to end. Every skip reason should make sense to you. Every sent message should be one you would have approved.

Phase 4 — all fans, in a separate window. Start All fans in the list with a 6- or 24-hour window, in a separate window so your team keeps working. Check the log after each pass.

At first, do not run phases 2 and 4 together. If something goes wrong, you want to know whether it is autopilot or the creator’s settings, and running both at once hides that.


  • Autopilot is turned on per chatter, per tab. It is not an account setting someone can leave on by accident.
  • What a chatter may turn on is a manager’s setting, per creator, in Creator Studio.
  • Every automatic send is saved with the fan, text, time and result.
  • Chats where automation stopped and a person is needed appear in Needs attention on the dashboard.
  • The creator’s emergency stop beats everything, instantly.