The short answer: Publish a comment policy, sort stories into risk tiers before they go out, limit comments on the handful that always go wrong, let rules-based automation hide abuse and scam links overnight, and have one editor review what was hidden each morning. Nobody has to sit up until 4am watching a crime story.
Why do news posts attract the worst comments?
Three things happen at once under a news post that do not happen under a brand post.
First, the subject is a real conflict. A court case, a protest, a local planning fight. People arrive with a position already formed and an opponent already in mind. Second, reach is wide and mostly strangers. A bakery's followers know the bakery; a regional paper's followers include anyone the algorithm handed the story to. Third, news posts run hot at the worst hours. A story filed at 22:00 gets its first thousand comments while the newsroom is empty.
Add to that the specific abuse journalists get: named reporters attacked personally, pile-ons after a byline is shared in a hostile group, and threats that sit under the post all night. And on busy pages, scammers follow the traffic. Fake giveaway links and cloned accounts show up under the most-shared story of the week because that is where the audience is.
Are you responsible for what readers write under your story?
Often, yes, and news publishers carry more of that risk than most page owners. Australian and European courts have treated page owners, including news outlets, as responsible for unlawful comments left under their posts once they could have acted. The practical rule is the same everywhere: act promptly once you know, and be able to show what you did. For the cases behind that rule and what they mean in day-to-day terms, read are you legally responsible for comments on your Facebook page.
What should a news comment policy say?
Write it once, publish it where readers can find it, and link it in the page's About section. It does two jobs: it tells readers where the line is, and it gives your moderators something to point at instead of improvising.
Keep it short and specific:
- What gets hidden: threats, incitement, slurs and dehumanizing language about a group, personal attacks on reporters or named private people, doxxing (addresses, phone numbers, workplaces), scam and spam links, impersonation of the publication or its staff.
- What stays: criticism of the article, of the reporter's work, of the publication's choices, strong political disagreement, and accusations of bias. Hiding criticism of your own journalism is the fastest way to lose the argument about moderation.
- What happens on repeat: a reader whose comments are hidden repeatedly may be banned from the page.
- How to appeal: an email address, and a promise to answer.
One line is worth adding explicitly: we do not hide comments for disagreeing with us. You will need to be able to prove it, which is why what you hide and why should be recorded somewhere you can retrieve. For the harder balance between a cleaner thread and a real debate, see how public-facing pages reduce toxic comments without silencing real debate.
How do you triage stories by risk before you post?
Most newsroom comment trouble comes from a small, predictable set of stories. Deciding the comment setting at the moment you schedule the post costs about ten seconds and saves the overnight mess.
| Story type | Typical comment risk | Before posting | Overnight setting |
|---|---|---|---|
| Sport, weather, culture, local events | Low. Banter, some spam. | Normal post. | Automated spam and abuse filtering only. |
| Politics, elections, council decisions | Medium to high. Pile-ons, party brigading, insults between readers. | Post normally, flag to the morning editor. | Automation on, hidden items reviewed first thing. |
| Crime, courts, named suspects | High. Presumption of guilt, naming of victims, vigilante talk. | Strip names from the caption where possible. | Automation on, strict. Consider limiting who can comment. |
| Migration, religion, minority communities | High. Slurs and dehumanizing language, coordinated accounts. | Decide in advance whether comments are open at all. | Limit comments or close the thread overnight. |
| Deaths, accidents, suicides, grieving families | High harm, low volume. | Consider posting with comments limited. | Limited or closed. Check the family's wishes. |
| Ongoing investigations and legal proceedings | Legal risk. Defamatory claims about people not yet convicted. | Legal sign-off on the caption. | Limited, with a named person on call. |
Tiering also tells your automation what to do. A rule that reads "on crime and court stories, hide any comment that accuses a named person of a crime they have not been convicted of" is a sentence any editor can write and any editor can later audit.
Can you close or limit comments on a sensitive story?
Partly. Facebook does not give Page posts a blanket off switch the way Groups get one. What it does give you is a per-post control over who can comment, from everyone who can see the post down to only the people and Pages you tag. Narrow it enough and the thread is effectively closed to the public while the post stays up. The menu wording has changed more than once, so check it on your own Page rather than trusting a screenshot from a blog.
Use it sparingly. Closing comments on a story about a protest reads as hiding from the audience, and readers say so. Closing them on a story about a child's death reads as decency. The difference is whether you are protecting people in the story or protecting yourself from criticism.
How do you cover the night without a night shift?
The honest options are three: pay someone to be awake, close comments between roughly 22:00 and 07:00, or automate the obvious cases and review them in the morning. Most regional newsrooms cannot afford the first and lose engagement with the second.
Automation is worth it for the parts of the job that need no judgment. Threats, slurs, doxxing, scam links and copy-pasted spam are the same at 3am as at 3pm, and they are exactly where a delay does the most damage. What automation must not do is make a silent call on a borderline comment. A comment that might be an unfair accusation and might be a legitimate tip should be flagged, not hidden, and a person should see it with their coffee.
Two properties make overnight automation defensible rather than risky. Hiding instead of deleting, because a hidden Facebook comment is reversible and still there in the morning for an editor to put back. And a written reason attached to every action, so the morning review takes minutes and an editor can answer a complaint with a record instead of a guess.
What does the morning review look like?
Give it to one person, on a fixed slot, as part of the first editorial hour. Fifteen minutes is usually enough.
- Flagged comments first. The ones automation would not decide. Hide, keep, or escalate to the editor on the story.
- Spot-check the hidden pile. Not all of it. Pull the ones hidden under your highest-traffic story and check nothing legitimate went with the spam. Unhide anything that was just rude about the paper.
- Escalate real threats. Anything aimed at a named reporter goes to the editor and, where appropriate, to the police. Screenshot before you hide, because evidence matters more than tidiness.
- Read the top thread for tips and corrections. See below.
- Note anything the policy did not cover and adjust the rules that evening, not in the middle of a story.
If you publish in more than one language, the rules have to work in each of them, which keyword lists famously do not. That problem has its own guide: how to moderate comments in multiple languages.
What do you get out of the comments?
Two things worth the moderation cost. Corrections: a name spelled wrong, a street in the wrong district, a date off by a day. Readers catch these faster than anyone, and a correction made within the hour is worth more than an argument about whether comments are worth having. Build a path from the moderation queue to the newsdesk so a credible correction reaches a journalist, not just a community manager.
And tips. People who were at the scene, who work in the hospital, who sat in the public gallery. Some of it is noise. Occasionally it is the follow-up story. A moderator whose only instruction is "hide the bad stuff" will scroll past both.
Where a tool fits
You can run this workflow with Facebook's own controls: a keyword list, the profanity filter, hiding by hand, banning repeat offenders. For a page posting a few times a day, that is enough. It stops being enough when one story brings thousands of comments overnight in more than one language, or when you need to explain a year later why a specific comment was hidden.
That is the gap Social Moderator AI was built for: you write the newsroom's rules in plain English, the AI reads every new comment against them and either hides it, flags it for a person, or drafts a reply in your voice, with a written reason for every action and an audit log behind it. Facebook and Instagram comments arrive in real time, YouTube is checked about every ten minutes, and TikTok is polled. There is more on the newsroom setup on the news media page.
Frequently asked questions
Should we just turn comments off on Facebook news posts?
On a handful of stories, yes. As a standing policy, it costs you reach, reader relationships, corrections and tips, and it does not remove the discussion. It moves it to groups and shared posts you cannot see or moderate at all.
Who in the newsroom should own comment moderation?
One named person with a daily slot, plus a clear escalation path to the editor for legal and safety issues. Shared ownership means nobody reviews the overnight queue.
Is hiding a comment censorship?
Hiding removes a comment from public view on your post. It does not stop anyone publishing the same thing on their own profile. The defensible position is a published policy, consistent application, and a record of what you hid and why, including when the comment attacked your own journalism.
What do we do about comments that defame someone in a story?
Act quickly once you are aware of them and keep the record, and take legal advice for your own jurisdiction.
Do we still need a person if moderation is automated?
Yes. Automation handles the repetitive, unambiguous cases at hours when nobody is awake. Judgment calls, escalations, corrections and tips need an editor, which is why flagged comments should wait for one rather than be decided silently.