Get the newsletter
Meta · AI / ML · Filed Dec 16, 2025 · USPTO published Jul 2, 2026 · Jul 2 release

Meta's New Patent Wants to Track Your Mood by Listening to Your Voice All Day

Meta has filed a patent for a system that continuously listens to your voice, detects how you're feeling, and builds a timeline of your emotional trends. The AI ties your mood to where you are, what you're doing, and what time it is.

Meta Patent: AI That Tracks Your Emotions Over Time, figure from US 2026/0182881 A1
Figure from the official USPTO publication.
Publication number
US 2026/0182881 A1(US20260182881A1)
Applicant
Meta Platforms, Inc.
Filed · Published
Dec 16, 2025 · Jul 2, 2026
Inventors
Lachlan Dunn
Status when we published
Waiting for an examiner
Document
20 claims
US classification
434/PCA.026
Parent application
Claims priority from a provisional application 63738984 (filed Dec 26, 2024)

What Meta's voice-based mood tracking actually does

Imagine if your phone kept a journal of how you sounded every time you spoke, and then a month later handed you a report: "You tend to sound stressed on Monday mornings and noticeably calmer after exercise." That's roughly what Meta's patent describes.

The system records your spoken words across many situations, transcribes them, and runs the audio through an AI that picks up both what you say and how you say it. It looks at tone, pace, and other nonverbal cues. It then notes the context: the time of day, your location, what you were doing, or what app you were using.

Over time, it pieces those signals together into a summary of your emotional trends, complete with specific quotes from your own conversations as evidence. Meta's patent frames this as a fitness coaching tool, but the underlying technology is a persistent emotional monitoring system tied to voice data collected across your daily life.

How the AI links voice cues to context and builds a mood log

The patent describes a multi-step pipeline that turns ongoing voice recordings into an emotional profile.

  • Audio capture and transcription: The system receives spoken audio continuously across different contexts and converts it to text.
  • Emotional-state ML model: A machine learning model (an AI trained to recognize mood signals) analyzes both the words spoken and nonverbal cues like tone, pitch, and pacing to assign emotional indicators to each segment of audio.
  • Contextual correlation: Those emotional indicators are linked to contextual factors: the time of day, the user's physical location, their current activity, or what digital service they were interacting with.
  • Citation generation: The system identifies specific passages in the transcripts that best support each emotional finding. Think of these as footnotes the AI uses to justify its conclusions about your mood.
  • Trend summary: All of this feeds into a periodic report summarizing emotional patterns over time, with those transcript citations included as supporting evidence.

The patent mentions fitness coaching as a framing, but the claims are written broadly. Nothing in the core method limits it to workout contexts. The system is designed to work across contexts, which is the key phrase in the filing.

Why always-on emotional listening raises real questions

The obvious tension here is that this system requires persistent, ambient voice recording to work. It is not triggered by a wake word or a single session. It builds a picture of your emotional life from accumulated audio captured across your day, tied to where you were and what you were doing. For users, that means the privacy stakes are not just about what you say but about how you sound over time.

Meta operates Ray-Ban smart glasses and is developing further wearable devices, which are natural form factors for always-on audio. A system that can tell whether the person wearing your glasses sounds happier at home than at work, or calmer after a walk than a meeting, would be genuinely useful as a wellness tool. It would also represent one of the most detailed behavioral datasets any company has ever collected.

Source: US 2026/0182881 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Meta patent news breaks.

There are more where this came from

We read the new patents from Big Tech, carmakers and more and send you the ones worth knowing. Plain English, free, every week.

Watchlists this filing belongs to

All watchlists →