How to deal with trolls and toxic comments

To deal with trolls, act early and don't argue. CommentCamp, $0 for site owners, builds in a moderation queue and spam filtering. In a 2017 study, seeing troll posts raised trolling odds by 68%. Publish clear rules, hold first-time comments, delete rule-breaking posts fast and ban repeat offenders.

US adults harassed
41% (Pew)
Severe harassment
25%, up from 15% in 2014
Left a community
34% of US adults
Bad mood effect
+89% odds of trolling

Why people troll, and why it spreads

Some trolls really are in it for the cruelty. In a 2014 study by Buckels, Trapnell and Paulhus, 5.6% of 418 US adults said trolling other users was what they most enjoyed doing in comment sections, and sadism was the strongest personality predictor of trolling.

Most toxic comments come from ordinary people, though. In a 2017 experiment by Cheng and colleagues with 667 participants, 35% of posts were trolling when people were in a good mood and read a civil thread. That rose to 68% when they were in a bad mood and read earlier troll posts. A bad mood raised the odds of trolling by 89%, and seeing troll posts raised them by 68%.

The same paper analyzed more than 16 million CNN.com posts. Users who had a post flagged in one discussion were twice as likely to troll in their next discussion (4.6% vs 2.1%).

The scale is real. In Pew Research Center's latest survey on online harassment, published in January 2021, 41% of US adults said they had been harassed online. 25% had faced severe harassment such as stalking or physical threats, up from 15% in 2014, and 64% of adults under 30 had been harassed.

Start with clear, public rules

Hostile threads cost you readers who never complain. In an October 2026 Pew Research Center report on online communities, 34% of US adults said they had left an online community. Of those, 57% cited too much arguing and 43% said people were being harassed.

Write a short rules page and link it right next to your comment box. Our community guidelines template gives you a starting point. Only list rules you will actually enforce: no personal attacks, no slurs, no threats, no posting private information, stay on topic.

Then make the report button easy to find. In a 2017 Coral Project survey of more than 12,000 commenters at 20 news sites, only 62% knew their site had a way to report offensive comments, even though every site had one.

Reward the commenters you want

Penalties alone make a comment section feel like a courtroom. Answer good questions, thank people who add sources and quote the best comments in follow-up posts.

CommentCamp, the free comment system we're building for independent publishers, bloggers and WordPress sites, rewards regulars with optional member levels, quests and helpfulness badges, with levels local to each site. A moderation queue and automated spam filtering are built in. It costs site owners $0 and is in private early access: join the waitlist for an email invite as rollout expands.

On Discourse, members gain more privileges as they climb its 5 trust levels, which gives regulars a reason to stay civil.

Don't feed trolls: what to do instead

No controlled study has tested "just don't reply" on its own. The case for it is indirect but strong: in the Cheng experiment, seeing troll posts made more trolling likely. Every quote-reply or public argument keeps the bait near the top of the thread, where the next reader sees it.

So don't debate bad-faith posters. Remove the post, and spend your replies on people who asked something real. In the same Coral Project survey, more than 70% of commenters said they wanted journalists to answer factual questions in the comments.

Tip: Watch the first replies on a new post closely. In the CNN.com data, after a flagged first post, later posts by other users were flagged 3.1% of the time, compared with 1.7% when the first post was clean.

The hide, delete, ban ladder

Escalate in steps and treat the same behavior the same way every time: delete the first offense, delete and warn on the second, then ban. Deleting beats hiding in small threads. A 2023 study of 412 million Facebook comments by Ribeiro, Cheng and West found that deleting rule-breaking comments reduced later rule-breaking in threads with 20 or fewer comments, including by other people. Hiding them had small, statistically insignificant effects.

Behavior Response Tool
Drive-by insult Delete it before it sets the tone Manual delete from the moderation queue
Slurs and repeated swearing Hold or delete automatically Keyword lists or a profanity filter
Off-topic bait Delete early, with no public reply Moderation queue
Flooding a thread Make people wait between posts Discourse slow mode
Pile-on from outsiders Hold replies from new accounts and non-members Reddit Crowd Control or first-comment approval
Threats or doxxing Save evidence, delete, ban WordPress Disallowed Comment Keys with their email and IP
Heated but sincere disagreement Leave it up and answer factual questions Your published rules

Add friction for new accounts

You can't read every comment at 2 a.m., so let your settings slow down strangers while regulars post freely:

  • WordPress: "Comment author must have a previously approved comment" is on by default, so first-time commenters wait for approval, and comments with 2 or more links are held. Comment Moderation keys hold matching comments, while Disallowed Comment Keys send them straight to the Trash. Both match inside words.
  • Discourse: 5 trust levels. New (TL0) users can't flag posts, post more than 2 links in a post or post more than 10 replies. Topic slow mode, added in version 2.6.0.beta5 in October 2020, makes people wait a staff-set time between posts.
  • YouTube Studio: Basic or Strict holding for potentially inappropriate comments, blocked words and hidden users. Held comments stay in review for up to 60 days.
  • Reddit Crowd Control: filters or collapses comments from accounts with negative community karma, new accounts and non-members.

These limits also blunt ban evasion, because a fresh account starts at the bottom again. See our comment moderation guide for queue and keyword setups.

Protect your moderators

Reading abuse for hours wears people down. In a 2019 experiment by Riedl, Masullo and Whipple, 747 participants each moderated 78 comments, and those who saw only uncivil comments reported more emotional exhaustion and lower task satisfaction.

Abuse isn't spread evenly either. When The Guardian analyzed the 70 million comments left on its site since 2006, 8 of its 10 most abused writers were women.

Note: Rotate moderation duty, mix queue work with other tasks and let filters and first-comment holds take the first pass, so nobody reads only the worst comments all day.

Frequently asked questions

Does ignoring trolls actually make them stop?

Not reliably. No controlled study has tested ignoring trolls on its own, but a 2017 study by Cheng and colleagues found that seeing earlier troll posts raised the odds of more trolling by 68%. Deleting rule-breaking posts quickly works better than leaving them up or arguing with them.

Is it better to hide or delete toxic comments?

Deleting works better in small threads. A 2023 study of 412 million Facebook comments by Ribeiro, Cheng and West found that deleting rule-breaking comments reduced later rule-breaking in threads with 20 or fewer comments, while hiding them had small, statistically insignificant effects.

Why do normal people act like trolls online?

Mood and the tone of the thread matter. In a 2017 experiment with 667 participants, 35% of posts were trolling when people were in a good mood and read a civil thread, rising to 68% when they were in a bad mood and saw earlier troll posts.

How do I stop banned trolls from coming back with new accounts?

Make new accounts earn trust. Keep WordPress's default setting that requires a previously approved comment, add the banned person's email and IP to Disallowed Comment Keys, and use tools like Reddit Crowd Control or Discourse trust levels, which limit what brand-new accounts can do.

Will strict comment moderation drive away my readers?

Unchecked fighting drives away more. In an October 2026 Pew Research Center report, 34% of US adults had left an online community, and 57% of them cited too much arguing. A 2016 Center for Media Engagement survey found argumentative comments kept 40.9% of comment readers away.

Sources

  1. pewresearch.org/internet/2021/01/13/the-state-of-online-haras...
  2. pewresearch.org/wp-content/uploads/sites/20/2026/10/PI_2026.1...
  3. arxiv.org/abs/1702.01119
  4. arxiv.org/abs/2210.10454
  5. www2.psych.ubc.ca/~dpaulhus/research/DARK_TRAITS/ARTICLES/PAI...
  6. guides.coralproject.net/12000-people-have-something-to-say
  7. mediaengagement.org/research/survey-of-commenters-and-comment...
  8. mediaengagement.org/research/moderating-uncivil-comments
  9. theguardian.com/technology/2016/apr/12/the-dark-side-of-guard...
  10. blog.discourse.org/2018/06/understanding-discourse-trust-levels
  11. meta.discourse.org/t/discourse-2-6-0-beta5-release-notes/168756
  12. wordpress.org/documentation/article/settings-discussion-screen
  13. support.google.com/youtube/answer/9483359
  14. support.reddithelp.com/hc/en-us/articles/15484545006996-Crowd...

Prices and plan limits checked on October 30, 2026.