Free tool · no sign-up
Is this comment toxic? Check it in two seconds.
Paste a reader comment. OpenModeration, the moderation engine behind Logora newsrooms, returns a verdict (publish, review or reject), a category and the reason, written in the language of the comment.
0 / 1200 · Try a sober argument, then a hostile one. The verdict updates live.
—
This is the Logora moderation engine running today on production newsrooms. It runs on OpenModeration: on client deployments, our AI handles the clear cases with each publisher’s rules, and doubtful cases, reports and contested decisions are handled by our team, included in the offer.
Reading the result
Four signals, and why they can disagree.
-
01
The verdict
Publish, review or reject. It is an editorial decision that weighs what the comment says and who it targets, not just the words it contains.
Logora : in production, doubtful cases are handled by our team, and each publisher sets its own rules and sees every decision in real time.
-
02
The category
Toxicity, spam, hate speech and other policy categories, so you know what kind of problem was found.
Logora : categories feed the DSA statement of reasons and the annual transparency report.
-
03
The reason
One or two sentences explaining the decision, written in the language of the comment.
Logora : this is the basis of the statement of reasons the reader receives when a contribution is removed.
-
04
The toxicity score
A probability between 0 and 1 from a classic classifier. It reacts to words, not to context.
Logora : a high score alone never rejects a comment. Harsh criticism of a public figure can score high and still be legitimate debate.
What to test
Try the cases that break generic filters.
Generic toxicity models were trained on social media and misread strong opinion as abuse. Test the hard cases your moderators see every day.
- Insults and personal attacks on other readers
- Hate speech targeting a group
- Spam, scams and promotional links
- Harsh but legitimate criticism of politicians
- Comments in the language of your readers, each analysed in its own language
- Heated debate that stays on the argument
The engine behind the tool
The same model newsrooms run in production.
-
7 years
of training on millions of messages; each comment analysed in its own language
OpenModeration
-
AI + team
clear cases by our AI, doubtful cases, reports and contested decisions by our team, included
OpenModeration
-
Real time
every decision and its written reason visible to the publisher, who can override it
Moderation admin
-
99.96%
API uptime in Q1 2026, monitored publicly on status.logora.fr
status.logora.fr
Questions
About the toxic comment checker.
-
Is the toxic comment checker free?
Yes. It is free, with no account and no email required. You test one comment at a time, as many times as you like.
-
What does the toxicity score mean?
It is the probability, between 0 and 1, that a classic classifier would reject the text. It is a useful signal but it only reads words. Learn more in our entry on toxicity detection.
-
Why was a comment with a high toxicity score approved?
Because the verdict looks at context. Calling a minister incompetent in blunt terms is part of public debate; attacking another reader is not. Generic models often confuse the two, which we explain in why generic toxicity models fail on opinion pieces.
-
Which languages does it support?
The model is multilingual: each comment is analysed in its own language. For the best result, use the version of this page in the language of the comment (language menu at the top). See multilingual comment moderation.
-
Can I moderate my whole comment section this way?
Yes, with Logora. Logora's moderation runs on OpenModeration: our own AI, cross-checked by other AI models and adapted to your rules, handles the clear cases; doubtful cases, reports and contested decisions are handled by our team, included in the offer. Every decision comes with a written reason, visible in real time, and you can override it. See AI moderation.
-
What happens to the text I paste?
It is sent to the Logora moderation API to compute the result. Do not paste personal data or private messages.
Moderating thousands of comments a day?
A 60-min call with Pierre or Henry. We look at your volume, your languages and your false positives, and come back with a scoped pilot.
Wie viel staatliche Förderung verträgt die deutsche Autoindustrie?
Wirtschaft · 3 Std · 847 Reaktionen · Moderiert von der Redaktion
Christine G. · vor 2 Std · ★ Top contributor
Subventionen ohne Bedingungen zu Standortzusagen sind politisch indefensibel. Der Steuerzahler trägt das Risiko, die Konzerne nehmen die Gewinne mit.
▲ 284 · ↩ 18 Antworten · Empfehlung der Redaktion
Jörg S. · vor 1 Std
Standortzusagen wurden in den 2010ern getestet, am Ende nicht eingehalten. Bedingungen müssen vertraglich an Beschäftigungszahlen geknüpft sein, nicht an Absichtserklärungen.
▲ 137 · ↩ 4 Antworten
Maria K. · vor 32 Min · Verifizierter Abonnent
Die wirklich offene Frage ist die der Transformation. Subventionen für die Elektrifizierung der Lieferketten, ja. Subventionen für den Status quo, nein.
▲ 92 · ↩ Antworten