About · how it works
About TrueMantis
AI-powered thread analysis for smarter online discourse
Score the thread with evidence.
AI analysis of online discussions. We don’t pick a side.
Every comment in a discussion gets classified sentence by sentence: evidence, refutations, genuine questions, name-calling, deflection. The thread earns a discourse score you can actually inspect.
No hot takes. No vibes. Just the arguments, graded.
TrueMantis is an independent service, not affiliated with, endorsed by, or sponsored by any social media platform. We analyze publicly available discussions, and every analysis links back to the original thread.
What the analysis measures
Six lenses on one discussion. Every score is earned inside a specific public conversation, and links back to the line it came from.
Sentence-Level Classification
Every sentence is classified against the 64-element Periodic Legend of Discourse and highlighted in place, so you can see exactly which line was evidence and which was a dodge.
Discourse Score (−100 to +100)
The whole thread earns one score from -100 to +100. Pervasively constructive discussions climb toward +100; bad-faith pile-ons sink below zero. No participation trophies.
Discourse Health Axes
Three 0-100 bars decompose the score: Justification (evidence & reasoning), Civility (interpersonal conduct), and Constructiveness (actually engaging the point).
Rhetoric Profile
Classical appeals, measured per comment: Ethos (credibility), Pathos (emotion), and Logos (logic), so you can see how an argument persuades, not just whether it does.
AI Fact Check
The opening post gets an automated accuracy read with a confidence rating and sourced claims, and Spot Check extends the same check to any reply in threads you capture. Always labeled as AI analysis, not an editorial opinion.
Context Clues
When a post quotes or clips someone, TrueMantis traces the line back to its original source and flags whether the context is intact, missing, or changed, so a cropped screenshot can't hide what was really said.
Leading Participant
Each analysis shows the participant whose arguments contributed most to that one discussion, derived from that single thread only, never a cross-thread profile.
Who uses TrueMantis
Built for everyone who reads the internet and wishes the arguments came with receipts.
Curious Readers
Skip the noise. Discover discussions worth your time, see which arguments actually hold up, and read the internet by quality instead of volume.
Journalists & Researchers
A searchable vault of AI fact-checks with sources, claim-level verdicts, and discourse data across every public discussion we have analyzed. Verify before you cite.
Educators & Critical Thinkers
Real arguments, graded in the wild: sentence-level classifications, rhetoric profiles, and quality axes make every analyzed thread a teaching artifact.
Community Moderators
See your community's discourse health at a glance: citation density, civility trends, and the threads driving change. Community-level insight, never user surveillance.
How it works
From mention to full analysis in three steps.
Tag @TrueMantisBot
Mention @TrueMantisBot on X or @truemantisbot.bsky.social on Bluesky, in any thread. Our bot always identifies itself as automated.
It reads the conversation
The conversation up to where you tagged us is captured and every sentence is classified, scored, and checked. No cherry-picking, no skimming.
Get the analysis
A shareable archive page with the discourse score, quality axes, highlighted sentences, fact check, and the leading participant of that thread.
Works on X and Bluesky today. More platforms are on the roadmap.
The Periodic Legend of Discourse
Our unified taxonomy: 64 discourse elements in eight families, arranged from constructive at the top to bad-faith at the bottom. The argumentative family runs the full ladder from refuting the central point down to name-calling, and the other seven families capture everything a single ladder alone can't.
Argumentative
The argumentative ladder, from refuting the central point down to name-calling.
Refutation · Counterargument · Ad Hominem
Information
Evidence quality: substantiated claims, cited sources, statistics, and misinformation.
Source Cited · Statistics · Misinformation
Speech Acts
What a sentence is doing: genuine questions, requests, commitments, apologies.
Genuine Question · Request · Apology
Social Bonding
How participants relate: agreement, respectful disagreement, inclusive or divisive humor.
Agreement · Respectful Disagreement · Divisive Humor
Emotional
Genuine expression and empathy versus manipulation and mockery.
Empathy · Legitimate Appeal · Mockery
Identity
Credentials, experience, and how in-groups and out-groups get treated.
Expertise · Disclaimer · Dehumanizing
Meta-Discourse
Talk about the talk: clarifications, definitions, boundary-setting, or derailing.
Clarification · Definition · Derailing
Bad Faith
The patterns that poison discussions: strawmanning, sealioning, gaslighting, trolling.
Strawman · Whataboutism · Trolling
Every archived analysis shows the full interactive Legend beside the thread.
Behind TrueMantis
No team, no investors, no ad network deciding what you see. Just an independent project trying to make online arguments easier to read honestly.
That independence is the point. TrueMantis can label discourse the way it actually reads, constructive or bad-faith, evidence or dodge, because it isn't built to flatter anyone or chase engagement.
Every analysis you capture is yours to export and keep. Paid plans and pay-as-you-go captures are what fund the work, and every dollar goes straight back into keeping TrueMantis independent and improving.