Skip to content
KeepAnswers

Guide

Auto-approving Facebook group members without wrecking the group

There are two kinds of auto-approval, and one of them quietly destroys the membership answers you turned questions on to collect. Here is the difference, and a pacing policy that behaves like an admin rather than a script.

Updated 7 minute read

The short version

  • Facebook's Admin Assist can approve requests on criteria, at no cost.
  • It cannot read the content of a written answer, and it saves nothing.
  • Approval is what deletes the request card, so auto-approval deletes the answers first.
  • Third-party risk comes from pace: bursts, no quiet hours, several groups at once.
  • Record the answers, then approve. Never the other way round.

Facebook’s own auto-approval

Admin Assist, inside the group’s admin tools, will approve join requests that meet criteria you choose. The criteria available have moved around over the years, and have included things like how long the account has existed, whether there are connections in common, and whether all membership questions were answered. Open Admin Assist in your own group and read the current list rather than trusting any article’s version of it, including this one.

It is free, it is first-party, and for a group that does not use membership questions it is often all you need. Two things it does not do are worth understanding before you switch it on.

It cannot read an answer

Admin Assist can check that all three questions were answered. It cannot check what the answers say. “I run a competing group and I’m here to recruit” satisfies the criterion perfectly.

It does not save anything

This is the part that catches admins out, and it is worth being precise about. The membership answers exist on the pending request card and nowhere else. Approving the request removes the card. Admin Assist approves requests. So the sequence, in a group with questions enabled and auto-approval on, is:

  1. A person types their email address into your question.
  2. Admin Assist checks the criteria and approves them.
  3. The card leaves the queue.
  4. The address is gone, and you never saw it.

The combination to avoid

Membership questions plus Facebook’s built-in auto-approval, with nothing recording the answers, is the worst of both. You are asking people for information and then destroying it automatically, at speed.

Either record the answers first, or turn one of the two features off.

Automating approvals without losing the answers

The order is the whole thing. Read the card, write the row, send the promised email, then approve. Everything in this category of tool is a version of that sequence, and the guide on saving answers covers the recording half in detail.

  1. 1

    Capture

    The three answers, the name, the profile URL and a timestamp, written to your sheet while the card is still on screen.

  2. 2

    Validate

    Check the email is shaped like an email. Flag blanks and keyword hits for a human rather than approving them blind.

  3. 3

    Deliver

    Send what the question promised. Doing this before approval means the member arrives having already got something.

  4. 4

    Approve, at a human pace

    Which is where most of the risk in automated approval actually lives.

What “safe” means, concretely

Facebook publishes no approval rate limit, and anybody quoting a specific number is guessing. What is observable is the shape of behaviour that draws attention: a burst of identical actions at identical intervals, at 3am, across several groups at once, from an account that was idle for a week.

So the policy is not a magic number, it is a set of habits that make the automated queue indistinguishable from an admin working through it on a Tuesday morning.

A pacing policy for automated approvals
SettingA reasonable defaultWhy
Interval between actions40 to 90 seconds, randomisedA fixed interval is the most obvious signature there is. The variance matters more than the average.
Daily capA number a person could plausibly have clickedThe cap protects you on the day a post goes wide and 400 requests land at once.
Quiet hoursOvernight in your own time zoneAn admin in London does not clear a queue at 4am. Neither should something acting as them.
Session lengthShort bursts, not a continuous dripPeople work in sittings. Twenty minutes twice a day reads better than one action every four minutes for 24 hours.
Gap between groupsMinutes, not secondsApproving in three groups in the same minute is a pattern no person produces.
Backlog behaviourSpread it over daysClearing 600 held requests in one evening is the single riskiest thing you can do.

Scroll for the rest

Multi-group spacing is worth calling out, because it is the setting most often left to the admin. Group Leads’ own guidance asks customers to space their multi-group intervals by hand to avoid Facebook activity flags, which works right up until the evening you forget. KeepAnswers works out the spacing across every group on the licence itself, which is one fewer thing to remember and one fewer thing to get wrong.

A decision rule worth writing down

Full automation only makes sense if you have decided in advance what should not be automatic. Three buckets:

  • Approve: all three questions answered, email shaped like an email, rules checkbox ticked, nothing in the hold list.
  • Hold for a human: a blank answer, an account created in the last couple of weeks, or a keyword you care about. Two of these a day is a minute of work.
  • Decline: whatever your group’s line is, stated in the rules so the decline is defensible if anyone asks.

Write the hold list before you turn anything on. Written afterwards, it is always a reaction to the one person who got through.

The first 24 hours after approval

A new member is more likely to engage on day one than any day after it, and an automated pipeline is well placed to use that. The lead magnet has already landed. What helps next is small and specific: a welcome post that names people rather than a generic one, or a single question in the feed that the new member’s own answer suggests they can answer.

That is the argument for capturing the answers even in a group where you would have approved everybody anyway. The approval is the boring half. The answers are the half that tells you what to post.

Where to go from here

If you have questions switched on, sort out the recording before you touch any auto-approval, first-party or otherwise. The Google Sheets guide covers where the rows should land, and the tool comparison covers which tools do the approving as well as the capturing, and which of them need your browser open to do it.

Questions people ask

Does Facebook have a built-in auto-approve for group members?
Yes. Admin Assist can approve join requests automatically when a request meets criteria you set, along the lines of account age, mutual connections and whether all membership questions were answered. Open Admin Assist in your own group to see the current list, because the criteria have changed over time.
Why does auto-approval lose my membership question answers?
Because approval is what removes the request card, and the card is the only place the answers exist. Facebook's own auto-approval does not copy the answers anywhere first, so switching it on with questions enabled means the answers are gone before you ever read them.
Is using a third-party tool to approve members risky?
The risk is pace, not automation. Clicking approve 200 times in four minutes looks nothing like a person working through a queue. A tool that spaces actions, caps the daily total and stops overnight behaves like an admin who is paying attention. One that races through the queue does not.
Should I approve everyone who answers the questions?
For most public-interest groups, yes, and the questions do their filtering by existing. For a group where fit matters, keep a hold rule: anything that trips a keyword, a blank answer or a profile created last week waits for a human. A hold queue of two a day is manageable. Reading every request is not.
How many members can I approve in a day?
Facebook publishes no number, and anyone quoting one is guessing. The useful frame is behaviour rather than a limit: keep the total in a range a person could plausibly have clicked through, spread it across the day, and stop overnight.

Keep reading