An AI medical chatbot may help someone organize symptoms, support a clinician's research, or handle a hospital appointment request. Those are different products with different responsibilities. Choosing by a single “medical AI” label can put the wrong tool in the wrong part of care.
For personal symptom information, start by evaluating Ada. For an organization building triage and care-navigation workflows, consider Infermedica. Glass Health belongs in a clinician's decision-support evaluation; Hyro focuses on patient-access work such as scheduling; Wysa supports mental-wellbeing self-help. These five options are compared by intended use, not by an invented diagnostic-accuracy score.
No chatbot in this guide should be used to rule out an emergency or replace appropriate professional care. If symptoms are severe or rapidly worsening, seek urgent medical help instead of continuing a chatbot conversation.
Medical chatbots at a glance (October 2026)
| Product | Who it's for | Type | Pricing (public) |
|---|---|---|---|
| Ada | Patients organizing symptoms | Consumer symptom assessment app | Free app |
| Infermedica | Health systems, insurers | Triage engine / API | Contact sales |
| Glass Health | Clinicians | Decision support + ambient documentation | Lite $0; Starter $20/mo; Pro $90/mo; Max $200/mo |
| Hyro | Hospitals, clinics | Patient-access & scheduling assistant | Contact sales |
| Wysa | Individuals, employers | Mental-wellbeing self-help | App with free tier; enterprise contracts |
| Amazon Health AI | US consumers | Consumer health assistant linked to care | See Amazon Health |
| Clearstep | Health systems | Care navigation | Contact sales |
| Microsoft Azure Health Bot | Organizations building their own bot | Developer platform | Azure usage pricing |
Eight tools, different users and clinical boundaries
| Product | Intended evaluation context | Boundary that matters |
|---|---|---|
| Ada | An individual organizing symptoms and possible next steps | Its assessment is not a medical diagnosis |
| Infermedica | A healthcare organization integrating symptom assessment and triage | Deployment requires a defined clinical workflow and escalation route |
| Glass Health | A clinician evaluating decision-support and documentation assistance | The clinician must review the evidence and retain decision authority |
| Hyro | A health system automating patient-access interactions | Booking an appointment is different from making a clinical decision |
| Wysa | A person or care program using mental-wellbeing self-help | It is not an emergency or crisis service |
| Amazon Health AI | Consumers seeking personalized health information and access to One Medical care | Availability, Prime offers and clinical treatment eligibility vary |
| Clearstep | Health systems deploying symptom triage, navigation and scheduling | Validate clinical governance, routing and integration for the local system |
| Microsoft Azure Health Bot | Healthcare organizations building patient-facing conversational experiences | It is a platform to configure and deploy; validate content, compliance and support locally |
The interface alone tells you little about those boundaries. A friendly conversation can sit on top of a structured symptom assessment, a scheduling system or a generative model. Ask what information the product uses, what output it produces, and who acts when the conversation cannot safely continue.
For people seeking information about symptoms
Ada: structure the information you bring to a healthcare professional
Ada offers a free symptom-checker experience that asks follow-up questions about the information you provide. Its assessment instructions describe a report of possible causes based on symptoms, demographic information and risk factors. The practical value is a structured record of what is happening and what questions to raise next.
For example, someone preparing for a non-urgent appointment may find it useful to organize when a symptom began, how it changed and what other symptoms occurred. Keep the report as supporting information, while telling the clinician anything the questionnaire did not capture. An answer is only as relevant as the information entered.
Ada explicitly states that its solutions are not intended to provide a medical diagnosis or replace medical advice. Its intended-use explanation is more important to the buying decision than a marketing claim about accuracy. A list of possible causes does not confirm one cause or safely exclude another.
The consumer app is described as free. Organizational deployments are a separate proposition; do not infer that a free consumer experience includes integration, governance or support for a clinical service. Also check availability and supported use for the person who will actually use it.
For organizations implementing assessment and navigation
Infermedica: evaluate the whole triage pathway
Infermedica provides clinical AI for symptom assessment, triage and care navigation. It is relevant to an insurer, telehealth service or healthcare provider integrating those functions into a broader patient journey. This is not simply a consumer chat subscription with a medical theme. Infermedica's platform overview describes the organizational offering.
Its triage API documentation explains that the triage endpoint categorizes cases using reported observations and the severity of likely conditions. It also describes combinations of symptoms or risk factors that can trigger a higher triage level. Those are implementation details with direct consequences for how the result should be displayed and handled.
During evaluation, follow one case through the complete service: intake, questions, recommendation, handoff and record retention. Check what happens if a person gives contradictory answers, cannot complete the assessment, or reports something outside the supported scope. A technically valid API response is not the same as a usable care pathway.
Pricing and deployment need to be scoped with the vendor. The developer documentation notes that trial access limits triage calls; it does not establish an unlimited production entitlement. Ask for the contracted usage unit, integration work, clinical configuration, support and change-management responsibilities. A low apparent API cost can be a small part of the overall deployment effort.
For clinicians evaluating assistance inside their work
Glass Health: assess the evidence behind a clinical suggestion
Glass Health positions its product around clinical intelligence and decision support, with documentation-related workflows also described in its product material. It is a candidate for clinicians evaluating assistance with a case, rather than for patients looking for an automated substitute for consultation. Start with the official product site and the current clinician plans.
The useful evaluation is specific: can a clinician inspect the basis of a suggestion, identify missing context and revise the output before it enters the record? A polished differential or plan is not enough. Check the relevance of cited evidence to the patient population, the date of the guidance and the assumptions made from incomplete information.
Keep research assistance separate from an approved clinical action. The responsible clinician needs to judge contraindications, competing explanations and local practice requirements. An assistant that drafts a note must also preserve the distinction between what the patient reported, what was observed and what remains uncertain.
Glass offers different access and workflow packages. The current pricing page was not consistently readable during this review, so we are not repeating an old flat subscription price. Confirm the specific plan's usage limits, any EHR connectivity and organizational terms directly before procurement. A personal account should not be assumed to include an enterprise deployment or every integration shown in marketing material.
For patient access and administrative conversations
Hyro: connect the conversation to an actual appointment workflow
Hyro's healthcare offering covers patient-access interactions through voice and chat. Its scheduling page describes appointment booking, rescheduling, verification and cancellation, including integration with Epic scheduling rules. This makes it relevant when the operational problem is getting patients to the right service and managing appointments.
The distinction matters: answering “When is my appointment?” and deciding “What treatment do I need?” involve very different responsibilities. A system can be useful at the first without being authorized for the second. Define which questions are administrative and how clinical questions reach a qualified person.
A meaningful demonstration should use your actual booking rules. Include a provider with restricted appointment types, a cancellation, a reschedule and a patient who needs staff assistance. Confirm that the scheduling system is updated correctly and that the user receives a clear confirmation. The chat transcript alone does not prove the appointment was booked.
Hyro is an organizational purchase, with scope and integration details discussed through its sales process rather than a simple consumer price. Request a quote that identifies channels, volumes, supported workflows, implementation and ongoing support. Evaluate successful completion and appropriate handoff, not just how many conversations avoid a human agent.
For mental-wellbeing self-help
Wysa: distinguish structured self-help from clinical treatment
Wysa describes its app as a space for self-help and self-monitoring of mental wellbeing. Its conversation and exercise library are intended to help users reflect on thoughts and feelings and practise skills. The official FAQ explains both the intended use and the limits of the service.
That makes it a different choice from a general symptom checker. Someone may use it to structure a reflection or work through a wellbeing exercise, while a care organization may evaluate a particular Wysa program alongside other services. Neither context justifies treating every Wysa conversation as therapy or a medical assessment.
The FAQ states that versions offered through institutional partners can differ: some are primarily rule-based, while others combine rules and language models. Ask which version and services your account includes. Do not assume an employer benefit, an insurer program and a directly downloaded app provide the same features or access to people.
Wysa is not recommended for crisis situations. Human support, where offered, is a separate service with its own availability and terms. Check the current app or program pricing and what is included rather than assuming that free conversational access includes a therapist. For immediate danger or a mental-health emergency, contact local emergency or crisis support directly.
Amazon Health AI: connect a consumer conversation to provider care
Amazon describes Health AI as a personal health assistant that can review health records, answer questions and connect users with One Medical services. That makes it different from a standalone symptom checker: the workflow may continue into clinician messaging or appointments. Eligibility, geography and promotional treatment access matter, so verify current terms before comparing costs. A bot response should not be treated as a diagnosis; check how escalation to a licensed provider works for the user’s specific concern.
Clearstep: assess navigation across the care journey
Clearstep offers healthcare organizations conversational symptom assessment, triage and care navigation. Evaluate the entire path from the patient’s initial question through routing, scheduling and handoff, rather than judging only the conversation. Ask how the deployment is governed, how local clinical protocols are maintained, and what happens when a patient’s symptoms fall outside supported pathways.
Microsoft Azure Health Bot: build an organization-specific assistant
Microsoft’s Azure Health Bot is a service for healthcare organizations to build conversational experiences, including symptom checking and patient navigation. It is a configurable platform rather than a ready-made consumer chatbot. Buyers need to assess the clinical content, integrations, privacy and compliance controls, and ongoing maintenance required for their own implementation.
Questions that change the purchase
For personal use, confirm the product's intended purpose, cost, availability and data controls. Be especially cautious about interpreting a reassuring response as a reason to delay care. Keep important health information and a route to professional help outside the app.
For an organization, use a written deployment scope. It should identify the users and supported populations, the decisions the system may assist, the data it receives, and the person or service responsible for escalation. Review privacy terms and any required agreements for your jurisdiction and use case; a general security badge does not answer those questions.
Ask for evidence relevant to the proposed task. A study about a different population, product version or setting may not support your deployment. Separate diagnostic performance, administrative task completion, user satisfaction and response speed. They measure different things and should not be combined into a single “best chatbot” score.
Run a bounded evaluation before expanding
Start with a narrow workflow and an agreed review process. For an appointment assistant, verify completed bookings and staff handoffs. For clinician support, have qualified reviewers examine evidence, omissions and unsupported statements. For symptom navigation, evaluate the complete escalation pathway with clinical governance in place.
Record failures as carefully as successes. An uncertain answer that reaches the right person can be preferable to a confident answer outside the product's scope. The appropriate tool is the one whose intended use, evidence and deployment conditions fit the task you actually need to perform.
How to choose a medical chatbot
- You are a patient: use a consumer symptom tool like Ada to organize symptoms before a visit. For emergencies, call 911 or go to the ER instead.
- You run a clinic or health system: separate administrative automation (Hyro: scheduling, FAQs) from clinical triage (Infermedica, Clearstep). They need different governance.
- You are a clinician: evaluate Glass Health or similar decision-support tools against the evidence they cite, and keep final judgment.
- Mental health: Wysa offers structured self-help, not treatment or crisis care.
- Always check: HIPAA eligibility and a BAA, escalation to a human, regulatory status, and published validation.
Can general chatbots like ChatGPT Health be trusted for triage?
Use caution. A 2026 Nature Medicine stress test of ChatGPT Health (launched January 2026) found it under-triaged 52% of gold-standard emergencies in clinician-written scenarios, while handling classic emergencies like stroke correctly. A 2026 npj Digital Medicine meta-analysis found LLM triage accuracy similar to clinicians' on average, with large variation and methodological weaknesses. Treat general chatbots as information tools, not triage.
Frequently asked questions
What is the best AI medical chatbot for patients? Ada is a widely used free symptom assessment app. Its output is information to bring to a professional, not a diagnosis.
Are medical chatbots HIPAA compliant? Some enterprise platforms (for example Microsoft Azure Health Bot) are offered as HIPAA-eligible. Consumer apps generally are not covered by HIPAA. Organizations need a signed BAA.
Can an AI chatbot diagnose me? No product here is a substitute for a clinician. Use them to prepare questions and decide when to seek care.
Is there a free medical AI for doctors? Glass Health has a free Lite plan for students; paid clinician plans start at $20 a month.
Related guides
Sources (accessed October 15, 2026)
- Glass Health pricing: https://glass.health/pricing/clinicians
- Ada: https://ada.com/app/
- Infermedica: https://infermedica.com/
- Hyro: https://www.hyro.ai/healthcare/scheduling-management/
- Wysa FAQ: https://www.wysa.com/faq
- Clearstep: https://www.clearstep.health/
- Amazon Health AI: https://health.amazon.com/health-ai/learn-more
- Microsoft Health Bot: https://www.microsoft.com/en-us/research/project/health-bot/
- Nature Medicine, ChatGPT Health triage test: https://www.nature.com/articles/s41591-026-04297-7
- npj Digital Medicine meta-analysis: https://www.nature.com/articles/s41746-026-02409-8
Updated October 15, 2026. This comparison uses topic-specific research and the linked official product documentation. It is a product-selection guide, not medical advice, a clinical validation study or a claim that we tested these systems on patients.
