Protect your community without switching off the comments
Turning comments off is not protection, it is surrender, and it costs you the audience the channel was built for. Sence reads every comment and direct message on arrival, hides anything that breaches your rules before your audience sees it, and leaves the rest of the conversation running. Your team keeps the final say and a record of every decision.
Moderation queue@grimjab_779:47 pmHidden
NegativePersonal HateViolent Language
@nineteenrows_back9:52 pm
Refs got that one right for once 😅
@dealhound_dana10:04 pmNot visible
MAKE $500/DAY FROM HOME — link in bio
NeutralSpamWebsite Link
Trusted by teams using Sence
Most of the harm arrives after everyone has gone home
Sence has analysed millions of social media comments in the last six months. In that time 21% of the comments left on brand communities were flagged as harmful, and the mix is not the one most teams brace for. Spam and scams make up 37% of it, harassment 27% and hate comments 26%, with violence, sexual content and everything else making up the rest.
The timing is the harder half. 72% of comments are posted outside regular working hours, peaking between 6pm and 10pm, which is exactly when the people who own the channel are not looking at it. The volume is going up as well: since Meta reduced its platform moderation in early 2025, Sence has seen a 30% increase in harmful comments on Facebook.
21% of comments are harmful
Of the comments left on brand communities in the last six months, 21% were flagged as harmful by Sence.
72% arrive out of hours
72% of comments are posted outside regular working hours, peaking between 6pm and 10pm, when most teams cannot respond.
30% more harm on Facebook
Since Meta reduced its platform moderation in early 2025, Sence has seen a 30% increase in harmful comments on Facebook.
Every comment and message read on arrival
Thanks for having me London!
Sence connects to your channels and reads what lands, comment by comment, across Facebook, Instagram, YouTube and TikTok, plus direct messages on Facebook, Instagram and TikTok. Nothing waits in a queue for someone to open a tab in the morning. At peak it runs at up to 10,000 comments a minute, which is what a final, a product launch or a bad news cycle actually does to a comment section.
- Comments on four networks, direct messages on three
- Read on arrival, 24/7, never batched overnight
- Up to 10,000 comments a minute at peak
Judged against your rules, not a generic score
HideANYPersonal HateORHarm WishANDRaceORReligionANDNegativecomments
Applies to: 2 channels · News section
@terracecal9:12 pm
We were robbed, that ref needs glasses 🤓
NegativeOpponent Hostility
@grimjab_779:14 pmHidden
NegativePersonal HateRace
Matched rule: Identity Based Abuse
Sence does not ship a default classifier. Every classifier is built from your own moderation decisions and your own policy taxonomy, so the line between a fan roasting the referee and a comment aimed at a person sits where your team drew it. Anything that breaches your rules is hidden before your audience sees it. Anything that does not is left alone.
- Moderation across eight categories of harmful comment
- Banter stays, targeted abuse goes
- Automatic in real time, or flagged for your team to decide
Nobody carries it alone, and nothing goes unrecorded
One policy
Redmount RoversClub channel
@teaomarama.hcky_13Personal channel
Audit log
HiddenPersonal HateNegative10:41 pm
matched Harm + Target rule
ApprovedCoachingNegative10:48 pm
banter about the game, left visible
Flagged for reviewPersonal HateNegative11:02 pm
aimed at Other Commenter, held for a person to decide
The abuse that does the most damage is aimed at a person, not a brand. Sence runs across club, league and athlete channels alike, so the people who cop the worst of it get the same cover as your official pages, without adding headcount to your social team. Every hidden comment stays on record with its classification and the reasoning attached, which is what turns a bad week into something your team can report on rather than relive.
- The same protection on personal channels as on your main pages
- A full record of what was hidden and why
- Between 83 and 410 hours a month saved on manual moderation
Thanks for having me London!
Three jobs, one classification pass
Everything runs off the same pass over the same comment. Comment & DM moderation software handles what arrives, custom content classification decides what counts as harmful in your community, and talent & team abuse protection covers the people it is aimed at.
01 A comment arrives
02 Classified
- Category
- Violent Threat
Aimed at the post author and describing harm, not reacting to the game.
03 Sentiment and theme
- Sentiment
- Negative
- Theme
- Controversial selections of players
04 Action taken
Hidden
Not visible to your audience; the classification and the reasoning are logged.
05 In the dashboard
2,442
comments on this post, every one read on arrival
Controversial selections of players
12% positive, 15% neutral, 70% negative, 3% mixed01 A comment arrives
killed it out there tonight lads 💀 gutted the season’s over
02 Classified
- Category
- Suspect Word
Violent vocabulary reading as celebration, so it is held for a person to call.
03 Sentiment and theme
- Sentiment
- Mixed
- Theme
- Match-day atmosphere
04 Action taken
Flagged for review
Held in the review queue. Your team decides whether it is hidden, deleted or approved.
05 In the dashboard
132
comments flagged for review on this post
Match-day atmosphere
64% positive, 15% neutral, 18% negative, 3% mixed01 A comment arrives
absolute robbery, that ref should never work again 😤
02 Classified
- Category
- Opponent Hostility
Hostility at the match, not at a person — loud is not the same as harmful.
03 Sentiment and theme
- Sentiment
- Negative
- Theme
- Refereeing frustration
04 Action taken
Approved
Left visible, exactly where it was. Banter stays.
05 In the dashboard
288
comments tagged Opponent Hostility on this post
Refereeing frustration
8% positive, 16% neutral, 72% negative, 4% mixedComment & DM Moderation
Every comment and message on Facebook, Instagram, YouTube and TikTok read as it lands, with anything that breaches your rules hidden before your audience sees it.
Learn moreCustom Classification
No default toxicity model. Your classifier is built from your own moderation decisions, so it draws the line exactly where your team draws it.
Learn moreTalent & Team Protection
Abuse aimed at players, staff and officials hidden in seconds, on personal channels as well as your official pages.
Learn moreA toxicity score cannot tell banter from abuse
A generic model gives every comment a number and leaves you to pick a threshold. Set it high and you lose the fans shouting at the referee, which is most of what makes a comment section worth having. Set it low and the comment that names the street someone lives on goes straight through, because there is not a swear word in it.
Sence does not ship a default classifier. Every classifier is built from your own moderation decisions and your own policy taxonomy, so the line sits where your team drew it, on your topics, in the language your community actually uses. That is what produces fewer false positives and fewer missed violations than a generic approach.
- No default toxicity model, and no threshold set by someone else
- Trained on your own moderation decisions, not on the open internet
- Different rules for different topics, where your policy needs them
- Every correction your team makes feeds back into your classifier
Protection is the first job, not the only one
Engage
The harmful comments are gone. The good ones still need answering, in your brand voice, across every channel you moderate.
ExploreUnderstand
The pass that classifies a comment also reads it. Turn what your community says into sentiment, themes and trends you can act on.
ExplorePlatform
How protection, engagement and insight run as one workflow instead of three tools your team switches between.
ExploreFrequently asked questions
No, and switching comments off would. Most people who see your post read the comments under it, and many read them specifically to learn what others say about a brand, so an empty comment section costs you the audience it was meant to protect. Sence removes the harmful comments and leaves the rest of the conversation running, which keeps your posts open and algorithm-friendly.
You do. Sence moderates across eight categories of harmful comment: hate, harassment, spam and scams, violence, extremist content, illicit activity, self-harm and sexual content. Where the line sits inside each category is set by your own policy, in workshops with your team, and it can differ by topic.
Yes. Sence can moderate automatically in real time, or your team can review flagged comments and decide whether to hide them. Every correction your team makes feeds back into your classifier.
It is flagged rather than hidden, and it waits for a person. Human review is available at any classification point, so the borderline calls sit in a queue for your team while the obvious ones keep running automatically. Your team decides whether it is hidden, deleted or left alone.
Platform filters work one network at a time and mostly on keywords, and a keyword list is always a step behind a community writing around it. Sence applies one policy across Facebook, Instagram, YouTube and TikTok, reads the comment in context instead of matching strings, and keeps every decision in one record. Platform tools also cannot tell you afterwards what was hidden or why.
Sence clients are saving between 83 and 410 hours each month that would otherwise be spent manually moderating comment sections. The rest of the return sits on the other side of the ledger: clean comment sections keep reach and engagement up, and paid ads can perform better when harmful comments are cleared in real time.