In a study of content creators, the large majority could recall a specific hate or harassment incident.
CHI 2022 studyFor creators who read their own comments. In a study of content creators, most could recall a hate or harassment incident, and for about one in three it is a regular occurrence. Here is the reorder that helps.
For creators who read their own comments
Stop dreading your own comment section.
Troll Guard hides the targeted, cruel comments on your screen until you choose to look. Your channel stays public, nothing gets deleted, and nobody else reads them first.
Runs on-device first. Uncertain comments may go to the scoring service. Nothing is removed from your channel. Refund if it doesn't shield your comments in your first session.
Live shield demoyoutube.com/watch · comments
This actually helped me ship my first app. Thank you.
Troll Guard shielded this before you read it.
86% risk · appearance / voice attack
I disagree with your point at 6:20, but the framework is solid.
Troll Guard shielded this before you read it.
91% risk · creator-targeted attack
underrated channel, the pacing on this one is perfect
Troll Guard shielded this before you read it.
95% risk · self-harm baiting
3 of 3 attacks held back this session. Your real fans stayed visible.
The 90-second walkthrough
Skip the lecture. See the reorder in six steps.
Interactive walkthrough1 / 6
The 2 a.m. scroll
You publish. Then you brace.
Most creators do not open their comments with excitement. They open them with a held breath, thumb hovering, ready to find the one line that ruins the night.
The honest version
This isn't about being thin-skinned.
You read your comments because your real fans are in there, your pinned questions are in there, and your next idea is in there. So you wade through the rest to find them. The fix isn't to toughen up or to swear off the comments you actually need. The fix is to not read the cruel one first, by default, on someone else's schedule.
What the research shows
You're not imagining the pattern. It's documented.
Roughly a third say targeted harassment is a routine part of creating, and some describe self-censoring or leaving.
CHI 2022 studyHave faced online harassment, and the more severe forms have been rising over time.
Pew ResearchCitations are linked in the report below. Troll Guard reports these to explain the problem, it does not claim to treat any health outcome.
You posted something good. Now there is a number next to the comment tab and you do not want to open it. You make coffee first. You tell yourself you will check later, after you have eaten, when you feel more like a person.
If you make videos in gaming, streaming, news, politics, fitness, beauty, or you show your face online as a woman, you know this stall. It is not a character flaw. It is the cost of searching for useful feedback inside a feed that also contains targeted cruelty.
The platforms moderate after you have already read the comment. By then the line has landed. That sequence is the problem Troll Guard changes.
The platforms moderate after you have already read the comment. By then the line has landed. That sequence is the problem.
Troll Guard changes the order. It folds toxic-scored comments before you read them, keeps real criticism visible, and leaves every public channel action in your hands.
Sources: CHI 2022 study of content creators, Pew Research on online harassment.
What it looks like
Three nights a lot of creators recognize.
Representative composites drawn from creator-harassment research, not specific individuals.
The 1 a.m. pile-on
You post late. Within the hour the thread fills with remarks about your voice and whether you should quit, and you read every one.
The cruelest lines fold behind a card before you see them. The fan asking a real question is still right there.
The coordinated raid
A smaller channel gets brigaded. Dozens of near-identical insults land faster than you can scroll.
Repeated and obfuscated attacks get scored, folded, and held in the Vault instead of in your face. You look when you want to.
The slow pullback
After months of appearance attacks, you quietly stop showing your face on camera, and your reach drifts down with it.
With the cruel comments off your screen by default, the thing that was quietly editing your choices loses its grip. You make the video you wanted to make.
How it works
Scan. Score. Fold. Review.
Scan the comment surface
A light content script reads the comments on a YouTube watch page using a self-healing selector config, so a YouTube layout change does not break the shield.
Score in fast batches
New comments are scored through the scoring service in batches. If scoring is unavailable, local phrase rules and safety guards still protect obvious cases while the remainder may pass through until scoring returns.
Fold it before you read
A comment scored toxic folds behind a small card with its risk score. The line never lands on you by default. Your real fans stay exactly where they are.
Review on your terms
Action history sits in a local vault. Reveal, Approve, Keep hidden, or Block similar. Every choice tunes the shield to the triggers you flag.
sample-pack comments handled correctly, checked before every release
of your Vault actions are logged locally so the shield keeps learning
the scoring key lives only on the backend, never in the extension
Vault actions and preferences stay on your device by default; scoring does not create a full comment archive.
Why it is different
Other tools clean up after the comment lands. This one holds it before you read it.
The old way
- You read the comment, then react.
- The line lands on you first.
- You clear a review queue in the same feed that just stung you.
- Tools optimize for the audience's view, not your reading.
With Troll Guard
- Toxic-scored comments fold before you read them.
- Real fans and fair criticism stay visible.
- You review the rest later, on your terms.
- Local-first, on your device, you in control.
The offer
Start free for 14 days. Then $19/mo.
Your first 14 days are free, card required, cancel anytime in one click, and we remind you before it renews. After that it's $19/mo, or $190/yr to save two months. You get the full-volume shield, the risk-score cards, the four actions, and the local learning that tunes to the triggers you flag. Up to about 150,000 scored comments a month, which is more than most channels in the 10k to 500k range will touch. Past the cap, new comments show unscored for the rest of the cycle and we email you to bump your tier. No silent failure.
Refund if it doesn't shield your comments in your first session. Cancel anytime, no retention maze. Nothing is removed from your channel, so leaving undoes nothing.
Prefer to pay once? A Founding Creator Pass exists, capped at 100 and enforced in code. It is a one-time acquisition option for early entrants.View founder availability.
Start my 14-day free trialOr pay yearly, $190 (save two months)Troll Guard is a comment-filtering tool, not a mental-health service. The classifier is automated and imperfect: it misses some comments and flags some it should not. Most content is processed with on-device rules first, with uncertain comments sent for scoring. Nothing is removed from your YouTube channel.
What you get
Folds toxic-scored comments before you read them
Risk-score card with four actions
Local learning, tuned to the triggers you flag
Local action history, on-device controls
Monthly$19
Plans
Start free. Move up only when the comment tab does.
14-day free trial
then $19/mo
The full shield for 14 days. Card required, cancel anytime in one click, and we remind you before it renews. Up to about 150,000 scored comments a month.
Start free trialAnnual
$190/yr
Same full shield, billed once a year. Saves you two months versus paying monthly.
Pay yearlyRunning a team or an agency roster? Email us and we'll set you up. Troll Guard is a comment-filtering tool, not a mental-health service. The classifier is automated and imperfect: it misses some comments and flags some it should not. Most content is processed with on-device rules first, with uncertain comments sent for scoring. Nothing is removed from your YouTube channel.
Straight answers
The questions every careful creator asks.
Does this delete comments or touch my channel?
No. Troll Guard runs in your browser and changes what shows up on your screen. It does not delete anything, does not hide anything from your audience, and does not moderate your public comment section. Uninstall it and your channel is exactly as it was.
Isn't YouTube's free hold-for-review enough?
For protecting your public section it is genuinely useful, and you should use it. Here is where it stops short. The held comments land in a review queue, and to clear that queue, you read them, one after another, with no buffer. YouTube's tool protects your audience from seeing the comment. It does not protect you from reading it. Troll Guard works on the creator side, folding what the classifier scores as toxic behind a card so you decide whether to open it.
Is this human-reviewed moderation?
No, and we won't pretend it is. It is a classifier, software scoring each comment, running locally. That makes it fast and private, and it also makes it imperfect. It will sometimes hide a fine comment and sometimes miss a bad one. That is why every card gives you Reveal and Approve, so you stay the final judge.
Where does my comment data go?
Most comments are filtered by local rules first. When scoring is needed, the comment text is sent for classification and not saved as a permanent transcript. Your choices, score cache, and preferences are logged locally so your shield can learn your thresholds.
Is the 100-founder cap real?
Yes. The cap is enforced in code, not marketing copy. When the hundredth founder pass is claimed, founder checkout closes and the price moves to a recurring plan.
What if it doesn't work for me?
Try it on a real upload. If it doesn't shield your comments in your first session, email us and we refund you. Cancel anytime. And since nothing was ever removed from your channel, walking away leaves it untouched.
Not ready to buy?
Get the free desktop shield and the launch notes.
Creators get priority when they can test on an active YouTube channel. No card required for the beta.
Ordinary disagreement stays visible. Toxic-scored comments fold behind a card.
Local-first by default. The scoring key never ships in the extension.