Every question.
No sales call required.

Emobot is a passive, continuous, multimodal AI platform that monitors a patient's emotional state between clinical visits. The patient installs the EmoDTx app on their iPhone in 3 minutes; after that, monitoring is fully passive — no daily check-ins, no surveys. The AI analyzes facial micro-expressions, voice prosody, and digital behavior on-device, producing a daily mood score and trend. When the score dips, the patient is nudged to rebook with their clinician; if the patient opts in, the clinician sees an anonymized trend line. Think of it as continuous Holter monitoring, for mood.

Yes — these are our two primary clinical use cases. For TMS: detect responders early (week 1–2 of treatment), confirm sustained response between maintenance visits, and catch silent relapse before the next PHQ-9. For Esketamine / Spravato: confirm the 24-hour antidepressant effect, track consolidation through induction, and detect when maintenance dosing needs adjustment. The mood-score curves on /try-it are grounded in published response-kinetics for both treatments (TMS τ ≈ 8 days, esketamine fast τ ≈ 2 days).

Emobot was clinically validated for major depressive disorder, treatment-resistant depression, and bipolar disorders. It also produces useful signal for anxiety disorders, PTSD, and recurrent depressive disorder. The platform is most differentiated in interventional psychiatry contexts (TMS, esketamine, transcranial direct current stimulation) where between-visit signal is hardest to obtain by survey.

No. Emobot is a patient-facing wellness app. The daily score, trend, and nudges live inside the patient's own app. When signals deteriorate, the patient gets an in-app invite to rebook with you. If the patient chooses to opt in, they can share a clean trend line with their clinician, but you never passively surveil anyone. This distinction is why US clinicians can deploy Emobot today, before the formal regulatory pathway concludes.

Yes. Emobot is designed for minimum data footprint. Facial analysis runs entirely on-device; raw video is never transmitted. Voice audio is sent to our servers in a 30-minute ephemeral window for voice-match verification and emotion analysis, then the recording is discarded and only the numerical emotion score is retained. Activity data (step count, distance, walking speed, flights climbed) and location metrics (trip count, hours away from home) are processed server-side and stored as numbers, not raw biometrics. All transmission uses TLS 1.3; storage uses AES-256 at rest. Patient records are kept separate from EHR systems. A BAA is available for US clinics.

Video is never recorded or transmitted. Facial analysis runs on-device and the raw buffer is discarded immediately; only a numerical score leaves the phone. Voice audio is captured during natural phone use and transmitted to our servers, where it lives for no more than 30 minutes while we run voice-match verification and emotion analysis. After that, the recording is discarded and only the numerical emotion score is retained.

Only if the patient opts in to share. In that case, you see a single anonymized daily score (0–100) plus signal-level sub-scores (facial, vocal, actigraphy, digital behavior), think of it as receiving a lab result, not the blood sample. No audio, no video, no raw biometrics. Opt-in is for a fixed period and cannot be revoked mid-period; at the end of the period, the patient can extend, narrow, or end the share. If the patient doesn't opt in to share, you simply see them rebook in your calendar.

Yes, and they're the primary audience. The EmoDTx app is a wellness companion: patients see their daily score, mood trend, and an activity summary. When the score dips, the app gently invites them to rebook a visit. The design is transparent and consented end-to-end, which is also what drives the 80% activation rate when a clinician proposes it and installs it with the patient during the visit.

Patients can stop monitoring at any time by uninstalling the app or revoking permissions in phone settings. They can also revoke the clinician opt-in share in one tap. All their data is deleted from Emobot servers within 30 days of a deletion request.

By default, nobody but the patient. Your clinician sees nothing unless you explicitly opt in to share — at which point they see only an anonymized daily score plus signal-level sub-scores, never raw audio or video. Emobot staff never see individual patient data; only de-identified aggregate analytics for clinical-validation research, and only after IRB approval. We never sell data, never use it for advertising, and never share it with third parties.

3 minutes. Patients download EmoDTx, complete a brief onboarding form, and grant permissions (camera, microphone, motion, Apple Health). After that, monitoring is completely passive. No daily check-ins, no surveys, no reminders.

Any iPhone running iOS 16 or later. No wearables, no external sensors, no additional hardware required.

Battery impact on iPhone is approximately 7–10%, comparable to or less than most navigation apps. Battery efficiency is a core engineering KPI. We reduced battery consumption by 30% in the past year and continue to optimize.

Emobot is designed to capture signal during natural phone usage; the monitoring adapts to how the patient already uses their phone. For patients with very low phone usage, some signal streams may have lower confidence, which is reflected in the dashboard uncertainty indicators.

80% of patients activate when a clinician proposes the app and installs it with them during the visit. This compares to 30–40% completion rates for PHQ-9 sent via email. Proposing the app and installing it with the patient during the visit is the single biggest driver of activation.

Because you've been told to track your mood — and you've never actually done it. Mood journals work for a week, then you stop. Emobot does the tracking for you, passively, on your phone. You see your trend the way you see your sleep score. And when something shifts that you didn't notice, the app gently invites you to rebook with your clinician — before things get worse.

A mood journal asks you to log how you feel; PHQ-9 asks you nine questions every few weeks. Both require effort, and both get less honest over time. Emobot tracks the involuntary signals — facial micro-expressions, voice prosody, daily activity rhythm — so the data is honest by default. You don't have to remember anything, fill anything in, or perform for the survey.

When your daily score dips below your personal baseline for several days, the app sends a gentle nudge to rebook with your clinician. Not an alarm, not a notification asking you to journal — just a single tap to schedule. The goal is to put the appointment in your calendar before the symptoms get bad enough that you'd reach out yourself.

You don't have to. Emobot's primary purpose is to give you visibility into your own mental health. If you choose to share with your clinician (one tap, anytime, revocable), they see a clean anonymized trend line — same as showing them an Apple Health export of your sleep. Sharing makes follow-up appointments dramatically more productive because your clinician walks in already knowing where to focus.

No. The AI analysis runs on-device, so very little data is transmitted. Daily upload is roughly 1–3 MB — about the size of a single photo on iMessage. Voice clips are processed in a 30-minute ephemeral window then deleted; only the numerical emotion scores leave your phone. Safe on cellular, won't burn through your data plan.

Ask your clinic if they're enrolled. If yes, they'll send you a one-tap enrollment link. If not, you can share /for-patients with your clinician — most interventional-psychiatry practices are eligible to deploy Emobot in under a day, and the enrollment is free for the patient.

In the pooled analysis of 3 prospective studies presented as a late-breaking poster at ASCP 2026 (Emobot × NRx Pharmaceuticals × CentraleSupélec × Johns Hopkins, n = 45), the EMOCARE multimodal score correlates with clinician-administered MADRS ratings at repeated-measures r ≈ 0.895 (p = .016). A correlation of 1.0 would be perfect agreement; 0.895 is near the ceiling of what's achievable with any passive monitoring technology. Most digital health biomarkers land at 0.5–0.7. The multimodal fusion (facial + vocal + actigraphy + digital behavior) is what lifts the number. The full poster is on /articles.

Sensitivity to change is arguably the more important clinical property. PHQ-9 Δ Spearman ≈ 0.834 (p < .001) and HAM-D17 Spearman ≈ 0.833 in the same pooled ASCP 2026 analysis, meaning the EMOCARE score moves when clinical severity moves. Preserved across three independent cohorts.

Our research stack is on the record. Clinical validation: late-breaking poster at ASCP 2026 (Perzo/Sapko/Petelot/Séguier/Javitt — Emobot × NRx Pharmaceuticals × CentraleSupélec × Johns Hopkins), plus the medRxiv companion preprint (2026). The speech side is published at IEEE ICASSP 2024 (synthetic data augmentation for speech emotion recognition, Ibrahim/Perzo/Leglaive) and ASSTA SST 2024 (symptom-level prediction from speech). The multimodal work is in Springer LNCS 15614 (data-driven validation of Russell's circumplex model of affect, 2025). The LLM adaptation for speech emotion is on arXiv (EmoSLLM, 2508.14130). Representation learning is on OpenReview (T-JEPA). Seven papers total, all linked on /articles with DOIs, authors, and open PDFs where available.

10+ clinical studies across France, the United States, Canada, and Germany, plus the EMOKET × COGPSY integration at GHU Paris. Flagships: EMC1 (Pr Schwan, CHU Nantes/Rennes, NCT06601140), MOODELING (Pr Vinckier, GHU Paris, 600 patients), Must-D (Dr Suzi Hong, UCSD/FHCSD, ~800 patients), DeeP-DD (Dr Benrimoh, McGill), and the US rTMS & Esketamine pragmatic pilots. Full list with NCT numbers on /clinical-evidence; published outputs on /articles.

The multimodal AI detects subtle changes in facial microexpressions, vocal prosody and rhythm, and physical activity (step count, walking speed, flights climbed) before clinical symptoms surface. When the score dips, the patient's app gently invites them to rebook with you, up to 48 hours before clinical signs would otherwise appear.

Emobot today is a patient-facing wellness app that a patient can use on their own smartphone; no device-class gate applies. In parallel, we're running the regulatory pathway on both sides of the Atlantic: the CE Mark pathway (EMC1 is 80% complete under Pr Schwan at CHU Nantes/Rennes; EMC2FR and EMC2-BD are 50% complete) and a dedicated US FDA clearance study (EMC1-MDD-US, N=92, Q2 2026 → Q1 2027). A French reimbursement RCT (REMOOD, N=504) is also in flight.

No EHR integration is required. Clinicians access a standalone web dashboard at portal.emobothealth.com from any browser; no installation on your side. Patient data is kept separate from your EHR by design, which simplifies compliance significantly.

Most clinics are fully operational within one day. You create an account at portal.emobothealth.com, add your clinical team, and start enrolling patients. We provide onboarding support and a short staff training video.

You send a patient enrollment link (via text or email) directly from the dashboard. The patient receives a link to download EmoDTx and complete their 3-minute setup. You can also scan a QR code together during the appointment.

PHQ-9 by email achieves 30–40% completion and captures one point in time every 6–8 weeks. Emobot is continuous, passive, and objective, and it sits with the patient, not on your dashboard. It gives the patient a daily signal and nudges them to rebook with you when it dips. That's the 99% of the treatment course you're currently blind to, turned into actual appointments.

Yes. De-identified aggregate data exports are available for research purposes. Contact us at contact@emobothealth.com to discuss research collaboration; we actively support academic partnerships.

80% activation when a clinician proposes it and installs it with the patient in the visit — independent of demographic. The app is 3 minutes to install and then asks nothing of the patient. There are no daily reminders, no surveys, no streaks to maintain. Patients in their 60s and 70s install it the same way they install any other clinic-recommended app. The trick is proposing it and installing it with the patient in the visit, not the user interface.

Approximately 2–5 minutes per patient at enrollment (sending the link or scanning the QR code during the appointment), and zero ongoing overhead unless a patient's score deteriorates — at which point the patient sees a nudge and reaches out to YOU to rebook. No staff member has to monitor dashboards. The signal travels to the patient first; the clinic gets a booked appointment, not an alert.

Yes — and the ROI compounds. A multi-site network runs Emobot at a single per-patient-month tier across the entire network, but captures value through two distinct P&L pathways: LTV extension on self-pay clinics, reimbursed maintenance throughput on commercial clinics. One SaaS line item; revenue lift in two places. The /roi-calculator page has a Multi-site Network toggle that runs the math for both pathways simultaneously.

Usage-based SaaS billed on Active Patient-Months across 3 tiers. You only pay for patients actively being monitored. Pricing is disclosed on a sales call and inside the Stripe checkout when you start your subscription; book a demo and we'll walk you through the tier that fits your clinic size. Multi-clinic groups receive a dedicated Customer Success Manager to maximise activation and outcomes.

A typical IP clinic sees a 4× net ROI, net of Emobot's subscription cost, not a gross revenue multiple. The primary driver: recapturing patients from silent drop-offs. A single recaptured patient can represent $8k–$19k in annual retained revenue, so the subscription typically pays for itself on the first rebooked visit. The bigger return is clinical: re-treating a relapsed patient early lifts sustained remission from ~70% to ~85%.

Activate your clinic account directly at portal.emobothealth.com/signup: self-serve in a few minutes, no procurement cycle. Once activated, you can invite your first patients the same day. Pilot arrangements (e.g. a scoped evaluation window, custom onboarding for multi-clinic groups) are negotiated on the sales call; book a demo if that's the path you'd prefer.

No long-term contract required. Monthly subscriptions can be cancelled at any time. Annual plans are available at a discounted rate for clinics ready to commit.

Today, Emobot is a clinician-pays SaaS subscription — the clinic absorbs the per-patient-month cost, and the ROI comes from the revenue side (recaptured patients, sustained TMS / Spravato courses). We have an active CMS reimbursement pathway in progress for the US (RPM 99453/99454/99457 codes once FDA clearance lands) and a French ROOD reimbursement RCT (REMOOD, N=504) for the European pathway. Until those land, the math we publish on /roi-calculator assumes no insurance reimbursement; any future reimbursement is upside on top.

Processed data for US clinics is stored in AWS US-East infrastructure under an AWS HDS (Health Data Hosting) configuration, segregated from EHR systems. Raw video is never stored. Voice recordings are kept in an ephemeral 30-minute window for voice-match verification and emotion analysis, then discarded; only the resulting numerical score is retained. Activity metrics from Apple Health and location metrics are stored server-side as numbers. All transmission uses TLS 1.3; at-rest storage uses AES-256. UK and EU patient data is stored in AWS eu-west (Ireland/Frankfurt) under UK GDPR / EU GDPR terms.

Yes. US clinics sign a Business Associate Agreement with Emobot before any PHI is exchanged. Our architecture is designed to minimise BAA scope: raw video never leaves the patient's device, voice audio is only ever held in a 30-minute ephemeral window before being discarded, and only processed numerical scores and metrics reach long-term storage. The agreement covers that narrow slice of de-identified data plus clinician metadata. Contact compliance@emobothealth.com. Terms are summarised on /terms (US) and /terms/uk (UK).

Within 30 days of the effective termination date, at your written option Emobot will either (i) return Customer Data via secure export in a readily usable format, or (ii) delete the data and certify deletion. Production-system deletion completes within 30 days; backup purge completes within 90–120 days. PHI return or destruction is governed by the BAA.

For US clinics under HIPAA: aggregated clinical records are retained for 10 years from the last interaction, in line with medical record retention standards. For UK/EU patients under GDPR: 5 years after the last account activity, after which data is anonymised or deleted. Patients can request deletion at any time via the EmoDTx app; all data is removed from production within 30 days. Full detail on /privacy (US) and /privacy/uk (UK).

Patients consent inside the EmoDTx app before any monitoring starts. They're told exactly what is collected (facial expression, voice samples, activity metrics, digital behavior signals), how it's processed (facial analysis on-device; voice audio held in a 30-minute ephemeral window on our servers then discarded; activity stored as numerical metrics), what is ultimately retained (numerical scores, never raw video or images), and for how long. They can opt in or out of sharing with their clinician, revoke consent in the app, and request deletion at any time. The consent language is the basis of the /privacy page.

You subscribe on a per-Active-Patient-Month basis, pay monthly in arrears by card or ACH, and get a clinician dashboard plus the EmoDTx patient app. Initial Term is 12 months from the Service Start Date, then auto-renews in 12-month cycles unless either party gives 60 days' notice. You may terminate for convenience after the first 6 paid months with 30 days' notice. Either party may terminate for material breach not cured within 30 days (10 days for undisputed payment breaches). Full US terms on /terms; UK equivalent on /terms/uk.

Liability is capped at the total fees paid in the six (6) months prior to the claim (updated from an earlier 12-month cap). Excludes indirect, consequential, and punitive damages. Does not cap your undisputed payment obligations, either party's wilful misconduct or fraud, our IP infringement indemnity, or a breach of confidentiality / BAA data-security obligations.

Yes. For customers operating two or more clinical sites under common ownership, Emobot dedicates a named Customer Success Manager whose mandate is to maximise activation, monitor outcomes, and coordinate Quarterly Business Reviews across all sites. Included at no additional fee. Written into Section 4 of the Terms.

Email support Monday–Friday, 7am–4pm EST (US federal holidays excluded). Target initial response for technical product issues: within 48 hours during support hours. Multi-clinic groups additionally get a dedicated CSM and Quarterly Business Reviews. P1 incidents (platform down) are responded to within 4 business hours.

Terms of Service: /terms (United States) and /terms/uk (United Kingdom). Privacy notice: /privacy (US, HIPAA-aligned, includes BAA summary) and /privacy/uk (UK, GDPR / DPA 2018 aligned, ICO complaint route). Governing law for US agreements is New York (federal or state courts, New York); UK agreements are governed by English law. Sub-processor list and our Data Processing Addendum are available on request from compliance@emobothealth.com.