AutoModerator setup guide for new mods (with an example config)
AutoModerator decides whether your subreddit feels like a community or a spam bucket. The day-one config we ship with branded subs, with working YAML.
Originally published April 13, 2026
AutoModerator is the unglamorous piece of infrastructure that decides whether a new subreddit feels like a community or a spam bucket. Soar is a community marketing agency that has run 4,200+ community campaigns across 280+ brands since 2017, and the pattern across our branded-subreddit launches is consistent: subreddits that ship a tuned AutoMod config in week one stay clean, and subreddits that do not drown in their own mod queue by month three. This post is the exact day-one setup we hand new brand mods, with working YAML you can paste.
What AutoModerator actually is
AutoModerator is Reddit's built-in rule engine. It reads every post and comment submitted to your subreddit, compares them against rules you define in YAML, and takes automatic actions: remove, filter (send to the modqueue for human review), approve, report, or auto-comment. It runs for free and is the single most important tool a new mod team has. Reddit's Mod Help center hosts the canonical docs, and the full reference wiki covers every supported field.
Without a tuned config, a human handles every piece of noise manually. With one, 80 to 95 percent of the noise never reaches a human at all. For a branded subreddit averaging 200 daily submissions, that is the difference between a community manager spending two hours a day in the queue and spending fifteen minutes.
Where AutoMod lives and YAML basics
AutoMod rules live in a wiki page at /r/yoursub/wiki/config/automoderator. You edit the page, save it, and rules are live within seconds. No deploy step. The page is YAML, each rule separated by three dashes (---), four-space indentation, no tabs. Reddit refuses to save malformed YAML, which is the first sanity check you will learn to trust.
Every rule is a set of conditions and an action. The main fields are type (posts, comments, or both), trigger fields like author, title, body, and domain, and action fields action, action_reason, and comment. The action field accepts remove, filter, approve, or report. The author field takes subfields like comment_karma, post_karma, and account_age, with comparison operators passed as strings ("< 10", "> 100", "30 days"). The tilde (~) prefix means "not" and is the single most useful piece of syntax in AutoMod.
Rule 1: Remove posts from low-karma accounts
The single highest-value rule you can ship on day one. It removes submissions from accounts with very low comment karma, which is where the overwhelming majority of spam comes from, and auto-replies with a friendly explanation to prevent the modmail cycle that follows silent removals.
type: submission
author:
comment_karma: "< 10"
action: remove
action_reason: "Low karma account"
comment: |
Hi! Your post was removed automatically because your account has very little karma. Please contribute to other communities first.
Start with a threshold of 10. Loosen if you see false positives, tighten if spam slips through. Our default for new brand subs is 5 to 50 depending on category — gaming and crypto subs benefit from a higher floor (50+), B2B SaaS subs run cleanly at 5 to 10.
Rule 2: Auto-approve trusted users
Once a subreddit has a regular contributor base, you want their posts to go live immediately. The approve action marks a post as human-approved, which overrides filters further down the chain. Approve rules belong at the top of the file so they run first.
type: any
author: [username1, username2, username3]
action: approve
The type: any line applies the rule to both posts and comments. Maintain the list by hand: when a user has demonstrated consistent, good-faith contribution, add them. For a brand sub, seed this list with the brand's own employees and a handful of known community members on day one.
Rule 3: Filter banned keywords in titles
Category-specific spam follows predictable patterns. SaaS subs see "growth hacking" and "guaranteed traffic" pitches. DTC brand subs see affiliate spam. List the terms and let AutoMod filter anything that contains them.
type: submission
title (regex, includes-word): ["spam-keyword-1", "spam-keyword-2"]
action: filter
action_reason: "Banned keyword in title"
Use filter rather than remove for keyword rules because keywords produce false positives. Filtered posts land in the modqueue for human review. The includes-word modifier makes the match word-based, avoiding false hits inside unrelated compound words.
Rule 4: Ban link domains
Link shorteners that hide the final destination, known affiliate networks, and competitor domains all belong on a domain block list.
type: submission
domain: ["bit.ly", "tinyurl.com", "spammy-site.example"]
action: remove
action_reason: "Disallowed domain"
Removing legitimate shorteners can frustrate casual users, so most brand subs ship a short list and add domains as they appear in the queue. The domain match is against the final URL's hostname, not the link text.
Rule 5: Require post flair
If your subreddit uses post flair to organize content by category, you can require every post to have one. AutoMod removes the post and auto-comments an explanation if the submitter skipped the step.
type: submission
~flair_text (includes): ["Discussion", "Question", "News", "Bug"]
action: remove
action_reason: "Missing post flair"
comment: |
Your post was removed because it does not have a required flair. Please resubmit and select a flair from the list: Discussion, Question, News, or Bug.
The tilde is load-bearing. It reads as "when the post flair does not include any of these values, take the action." Without it, the rule would remove every flaired post, which is the kind of error that gets mods panicking at 2am.
Rule 6: Auto-comment a removal reason
Any removal rule can include a comment field that auto-replies with a friendly explanation. This cuts modmail volume dramatically because most removed users just wanted to know what went wrong. Include a comment on every remove action by default.
type: submission
body (regex): "(?i)affiliate|referral code|click here"
action: remove
action_reason: "Promotional content"
comment: |
Your post was removed because it appears to contain promotional or affiliate content. This community does not allow referral codes, affiliate links, or "click here" pitches. If you think this was a mistake, please message the mods.
The (?i) prefix makes the regex case-insensitive; | separates alternatives. Test regex patterns in an external tester before pasting them in.
Rule 7: Flag possible reposts
Reposting is hard to catch with a single rule, but you can flag obvious cases. This rule reports (rather than removes) submissions matching common low-effort patterns, leaving the decision to a human mod.
type: submission
title+body: "['giveaway', 'contest', 'free stuff']"
action: report
action_reason: "Possible repost or low-effort promo"
For real repost detection, combine AutoMod with a third-party bot like RepostSleuthBot. AutoMod on its own is not a duplicate-detection engine.
How to test rules safely
Use action: filter instead of action: remove for any new rule in its first week. Filtering sends posts to the modqueue without removing them, so false positives are visible and recoverable. Once the false-positive rate is acceptable, promote the rule to remove. Write rules incrementally: add one, save, watch it run for a day, then add the next. Shipping ten rules at once is how new mods accidentally remove every post for six hours on a Saturday. A solid reference config to read is NoahLE's public config on GitHub.
Common mistakes we see
Vague keyword rules that catch real posts. A rule that filters the word "free" will remove half of a community around free-tier SaaS. Pair broad keywords with narrower context using
title+bodyorincludes-word.Regex errors. A broken regex silently matches nothing or matches everything. Test in an external tester before shipping.
Conflicting rules. Order matters. Approve rules should live at the top of the file so they run first.
No comment on removal. Silent removals produce angry modmail. A friendly auto-comment produces almost none.
No filter week. Shipping new rules straight to
removeis how brand subs accidentally remove their own employees' posts. Filter first.
The brand-sub angle
AutoMod for a branded subreddit is tuned differently than for a general community. A brand sub needs to allow mentions of the brand and product (which would otherwise look like self-promotion) while filtering competitor spam and affiliate schemes. That usually means an approve rule scoped to brand-related terms, combined with aggressive keyword filters for known competitor pitches and affiliate domains. For the wider context, see our 2026 guide to running a branded subreddit and the semi-official subreddit model.
Frequently asked questions
What karma threshold should a new brand subreddit start with?
Most brand subs run cleanly at a 5–10 comment-karma floor. Higher-spam categories (crypto, gaming, fitness) benefit from a 50+ floor. Start at 10, watch the modqueue for a week, and adjust based on the false-positive vs spam ratio you actually observe rather than the one you guessed.
Should I use `remove` or `filter` for new rules?
Always filter for the first week of any new rule. Filtered posts go to the modqueue without disappearing from the subreddit, so false positives are visible and recoverable. Promote to remove only after seven to fourteen days of low false-positive rate. This is the single change that prevents the "I accidentally removed every post on a Saturday" failure mode.
Can AutoMod replace a human moderator entirely?
No. A tuned config handles 80–95 percent of daily noise, but AutoMod cannot judge nuance — sarcasm, brigading sentiment, partial brand mentions, or borderline self-promotion all need a human. The right framing is that AutoMod handles the predictable patterns so humans can spend their time on the calls that actually require judgment.
How do I prevent AutoMod from filtering my own brand mentions?
Add an approve rule scoped to your brand and product terms at the top of the config, before the broader self-promotion filters. Approve rules run first and override later filters. The pattern is type: any, then body+title+url (includes-word): ["yourbrand", "yourproduct"], then action: approve with a tight scope so it does not over-approve unrelated posts.
What breaks AutoMod silently?
Malformed YAML (Reddit refuses to save), but also: regex patterns with unescaped special characters, incorrect operator syntax for karma thresholds (use strings like "< 10", not bare numbers), and conflicting approve/remove rules where the wrong one fires first. Diff the wiki edit history when AutoMod stops behaving as expected.
Are there third-party bots that complement AutoMod?
Yes. RepostSleuthBot for duplicate detection, ContextModBot for cross-subreddit ban context, and Toolbox (browser extension, not a bot) for queue management. None replace AutoMod; they layer on top of it. Brand subs typically run AutoMod plus RepostSleuthBot from launch and add others as the community matures.
Conclusion
AutoModerator is not glamorous, but it is the difference between a subreddit that runs itself and one that drowns its mods. Ship the low-karma filter, the trusted-user approve list, and a few category-specific keyword rules on day one. Promote from filter to remove once you have watched them for a week. Auto-comment every removal. Done right, AutoMod handles 80 to 95 percent of the daily noise, which is the difference between a community manager doing their job in 10 hours a week and drowning in 30.
