RedFlag vs LiarLiar.ai vs Liars.ai: what each actually does

Last Updated: September 22, 2026

If you've searched for a "lie detector" tool, you've probably landed on one of these three names. They do different things, run in different places, and — this is the part worth reading before you pick one — they make very different claims about what they can tell you. Here is what each one actually is, as plainly as we can put it.

The core difference, up front

LiarLiar.ai and Liars.ai both lead with "lie detection" as the pitch. LiarLiar markets itself around "polygraph technology" and is pivoting toward a product called PolygrAI. Liars.ai calls itself an "AI lie detector fun tool." Both are, in effect, selling certainty about something science doesn't currently make certain.

RedFlag doesn't do that. It measures how a speaker's pitch variation, blink rate, and speech continuity deviate from normal ranges — including their own personal baseline — over a video, and shows you where the deviation is largest. That's it. We call it baseline-relative behavioral analytics, not lie detection, because that's what it is. The full scoring method — every signal, threshold, and gate — is published at /methodology/ so anyone can check it, and it's tested against hand-labeled reference videos rather than just asserted. If you want the deeper case for why no video tool can detect lies at all, see can AI detect lies from video? and the 54% test.

How each one runs

RedFlag runs entirely in your browser: paste a public YouTube URL, and the face and voice analysis happens client-side, on that one video, after the fact. Nothing to install.

LiarLiar.ai is a desktop app for Windows and Mac that overlays live on video calls — Zoom, Meet, and similar — analyzing the other person's face and voice in real time while you talk to them. That's a genuinely different use case from analyzing a public recording: it's built for live conversations, not YouTube clips, and it requires installing software that can see your call.

Liars.ai is a web tool, positioned as entertainment ("fun tool") rather than a serious analysis product. It appears largely inactive and low-traffic at the time of writing.

Free tier and pricing

RedFlag: 1 video/day (up to 10 minutes) with no account at all; 2/day (up to 30 minutes) with a free account; unlimited analysis for $9.99/month, $21 per 3 months, $36 per 6 months, or $60 per 12 months. Code REDFLAGFRIEND takes 90% off during the current beta.

LiarLiar.ai: roughly $5.99/month, or a lifetime-access beta deal around $29.99, per its public pricing.

Liars.ai: a small free trial, unlocked by inviting friends rather than a flat daily allowance.

Side by side

Where LiarLiar's feature set is genuinely different

To be fair to them: LiarLiar's live video-call overlay with rPPG-based heart-rate estimation is a real, distinct feature set, not a lesser version of what RedFlag does. Watching a live call in real time and estimating heart rate from facial video are capabilities RedFlag doesn't have and isn't built for — RedFlag analyzes a public, already-recorded YouTube video, once, after the fact. If what you actually want is a live overlay during a call, that's a different job than RedFlag is built to do, and LiarLiar is the one built for it. Where we differ is the claim layered on top of that capability: "polygraph technology" implies a validated lie-detection instrument, and no webcam-based system — RedFlag's, LiarLiar's, or anyone else's — has that validation.

Where Liars.ai fits

Liars.ai reads as a novelty product, and it says so itself ("fun tool"). There's nothing wrong with that framing as long as it's clear going in — the risk is only when a tool built for entertainment gets treated as if its output means something in a real decision. Its invite-based free trial and low activity suggest it's not aiming to be a serious analysis product, which is a reasonable, honestly-framed niche to occupy.

Why RedFlag positions this way

Two reasons, and we'd rather state both than let you assume just one. First, it's accurate: the science of detecting deception from video or voice doesn't support a "detector" claim, and saying otherwise would be a fact we invented to sound more impressive. Second, it's practical — payment processors are wary of tools marketed as lie detectors, for the same underlying reason. But the order matters: we didn't land on "not a lie detector" because Polar prefers it. We'd say it even if that weren't true, because it's the honest description of what pitch monotony, blink elevation, and voiced-continuity deviation actually measure. A red flag means a speaker's measured behavior moved outside their normal range at that moment. It doesn't mean they lied. You interpret it — RedFlag just shows you where to look, transparently enough that you can check its work.