Learn: the research behind the signals

By the RedFlag team

RedFlag measures body-language and voice signals in video. These explainers cover what the peer-reviewed research actually says about each signal — including the parts that cut against the hype. No technology can detect lies, and we say so on every page; what can be measured is where a speaker's signals diverge from their own baseline.

Best body language & voice analysis tools (2026)

An honest roundup of the tools people reach for to check a video — and which question each one actually answers.

Deepfake detection vs behavioral analysis

“Is the media fake?” and “how does this real person come across?” are two different questions. Here’s which tool answers which.

How to tell if a video is real

A practical checklist: provenance, deepfake signs, selective editing, and what a person’s own behavior can and can’t tell you.

How to analyze an interview

Establish a baseline, watch for changes, and read the signals honestly — with a worked example.

How to analyze a political speech

Prepared remarks are rehearsed by design — so the measurable part starts at the Q&A. Plus what a wide podium shot can and can’t give you.

How to analyze a podcast guest

Long-form podcasts are great behavioral footage — if you handle the two-shot, the shared audio track and the three-hour drift.

Body language in video calls

What webcams, gallery view, noise suppression and lag do to the signals — how to review your own calls, and where analysis shouldn’t be used.

RedFlag vs LiarLiar.ai vs Liars.ai

What each tool actually does, how it runs, what it costs, and why RedFlag is deliberately not a lie detector.

Case study: Bernie Madoff in 2007 — zero flags

A year before his Ponzi scheme collapsed, a calm, fluent delivery. The counter-example that shows why signals are not verdicts.

Case study: Sam Bankman-Fried at DealBook, 2022

15 flagged moments three weeks after FTX collapsed — flat stretches and agitated bursts, both off his normal band.

Case study: Jodi Arias’ 2008 jail interview

36 flagged moments while she denied involvement — and why an expressive speaker gets judged against her own baseline.

Case study: Chris Watts’ 2018 porch interview

Recorded while his family was reported missing. 11 flagged moments, driven by an unusually flat voice.

Case study: Scott Peterson’s 2003 TV interview

Close to half the interview flagged, almost entirely vocal — and what archival audio does to the measurement.

Case study: Bill Clinton’s 1998 Lewinsky denial

The analyzer flagged the famous line itself. Measured signals, not a verdict — the record is settled.

Case study: Anthony Weiner’s 2011 photo denial

12 flagged moments in the CBS interview six days before his admission. Pitch flattened well below his own baseline.

Case study: the 1999 Lance Armstrong doping denial

We ran a famous denial through the analyzer. 6 flagged moments, driven mostly by pitch flattening — measured signals, not a verdict.

Can AI detect lies from video? What the evidence says

Short answer: no — and any tool that claims otherwise is overselling. What the research supports instead, and what is honestly measurable in a video.

Blink rate and deception: what the research shows

Do people blink more under pressure? The evidence on blink suppression, the post-load rebound, and why only person-relative baselines make sense.

Does your voice change under stress? Pitch, monotony and the evidence

What decades of paralinguistics research says about pitch under cognitive strain — and why commercial voice-stress analyzers scored at chance.

Cognitive load in speech: pauses, fillers and speaking rate

Thinking hard leaks into speech. The research on pauses, speech rate and fillers, and what a careful listener (or analyzer) can actually pick up.

The 54% test: how good is human judgment?

Eight famous statements, each later confirmed or contradicted by the documented record. Judge them yourself against the 54% research average.

Want the exact thresholds and test procedure behind the verdict? Read How RedFlag scores a video.