Screen what you watch.

What is Kinexis.AI

Kinexis helps anyone ask: “Is this video real, and how is it trying to persuade me?” We combine AI-powered authenticity screening with structured behavioral analysis to help people evaluate what they see and hear online.

All analysis is performed by AI models, not human analysts. Our AI screens video for signs of synthetic manipulation, then processes it through multiple analytic lenses (expression, voice, language, body, context) and synthesizes the results into a structured report for further review.

Authenticity Screening

We screen each video for indicators of AI generation or manipulation: synthetic faces, altered timelines, audio-visual mismatches, behavioral anomalies, and other artifacts that are difficult to spot with the naked eye. This screening examines both technical signals (pixel-level artifacts, audio inconsistencies) and contextual signals (behavioral plausibility, provenance, content consistency).

You get a summary of authenticity signals with confidence levels, not a simple “yes/no” verdict. Screening produces a probability estimate, not a guarantee. A clean result lowers suspicion but does not prove a video is authentic. Modern AI-generated media can be technically flawless, so contextual analysis is often more revealing than pixel-level inspection alone.

Behavioral Analysis

Once authenticity is assessed, we analyze how the video communicates across six modalities:

• Facial expressions. Expression patterns, intensity grading, and congruence with speech content
• Vocal patterns. Pitch, pace, pause patterns, and disfluencies
• Body language. Gestures, posture shifts, and self-touch behaviors
• Gaze behavior. Direction, fixation patterns, and visual attention
• Linguistic patterns. Pronoun use, hedging, detail distribution, and tense shifts
• Interaction dynamics. Dominance, deference, turn-taking, and rapport indicators

We surface patterns in delivery and communication technique. We do not claim to know what someone “really feels” or whether they are “lying.” Behavioral patterns have many possible explanations, and our reports present alternatives alongside every finding.

Narrative & Information Operations

Beyond individual behavior, we assess the information environment. Video content exists within a communicative context. It may inform, persuade, mislead, or some combination of all three.

We surface framing choices, bias indicators, omissions, narrative structure, and audience targeting. Content can be strategic communication whether or not it's labeled as such. The patterns are presented for analysts to draw their own conclusions.

What You Get

You get a structured report with:

• Authenticity screening with manipulation signals and confidence levels
• Headline and narrative summary
• Timestamped behavioral events with confidence scores
• Per-person body language analysis with gesture patterns and posture shifts
• Per-person emotional trajectory across the video
• Credibility-relevant indicators, including supporting signals, concerns, and alternative explanations
• Visual and production analysis covering scene composition, camera work, and on-screen elements
• Context enrichment covering topic identification, event and issue context
• Information operations assessment covering narrative structure, bias, and persuasion tactics
• Full transcript with speaker attribution

Not all sections appear in every report. Videos without visible people will not include face-specific behavioral analysis. The AI adapts its output to what is observable.

Our Principles

Probabilistic, not definitive

We surface indicators and hypotheses, never verdicts. Findings carry confidence scores, whether they relate to authenticity signals or behavioral patterns. Our assessments are one input among many.

Multi-modal, not single-channel

We cross-reference face, voice, body, language, and context to reduce reliance on any single signal.

Transparent uncertainty

When we're less certain, we say so. Alternative explanations are part of every assessment.

Analysis, not judgment

We detect and describe patterns in the content, not the character of the person in the video. Kinexis is not a lie detector. We don't determine guilt or intent.

Screening as media literacy

Disinformation thrives when people lack the tools to question what they see. Every Kinexis report is designed to make the techniques visible — deepfake artifacts, emotional manipulation, narrative framing — so you can evaluate content on its merits, not its production value. Build a habit of checking before you share.

Limitations

• All analysis is generated by AI models, which can hallucinate, misidentify expressions, misinterpret context, or produce inconsistent results across runs. AI output should always be reviewed by a qualified human before acting on it.
• No screening system catches all deepfakes. Synthetic media techniques evolve rapidly, and our models may miss novel approaches. A clean result lowers suspicion but does not guarantee authenticity.
• Video-only inference is inherently limited. There is no physiological data, no interviewer control, and camera angle, lighting, and compression artifacts affect what can be observed.
• Cultural norms, fatigue, medical conditions, camera anxiety, and media format expectations can produce patterns that mimic signals of interest. Behavioral indicators are not proof of deception, intent, or any particular mental state.
• This system cannot distinguish between neurodivergent traits, personality pathology, or any other cause of atypical presentation. Its output should never be used to infer diagnoses, personality disorders, or to dismiss or excuse flagged patterns.
• Our analysis is one input among many and should complement, not replace, traditional investigative and analytic methods.
• Confidence scores reflect the AI model's self-assessment, not an objectively calibrated probability. They indicate relative certainty within a report, not an absolute measure of accuracy.

Disclaimer

Kinexis.AI reports are for informational and educational purposes only. They do not constitute legal, medical, psychological, forensic, or professional advice of any kind. Do not use Kinexis.AI output as the sole basis for legal, employment, medical, financial, or safety-critical decisions.

Kinexis.AI does not guarantee the accuracy, completeness, or reliability of any analysis. All output is generated by AI models and reflects probabilistic inference, not verified fact. Users are responsible for independently verifying any findings before relying on them.

Kinexis.AI is not a lie detector, a forensic tool, or a substitute for professional investigation. The presence or absence of any signal in a report does not prove or disprove deception, authenticity, or intent.