A symptom checker is software that asks a user about their symptoms and then suggests possible conditions and how urgently they should seek care. Modern ones, like Ada, are built on probabilistic reasoning over a knowledge base of conditions and findings; the user answers a branching series of questions, and the app outputs a ranked list of candidate conditions plus urgency advice (for example, self-care, see a doctor, or go to the emergency room). They are marketed as a front door to care or a way to decide whether a symptom is worth a visit - explicitly as triage aids, not as a substitute for diagnosis.
The category has been studied carefully. A 2020 BMJ Open vignette study by Gilbert and colleagues compared eight symptom-checker apps to general practitioners on 200 clinical scenarios and measured two separate things: did the app suggest the right condition, and was its urgency advice safe. No app matched doctors on diagnosis - the best, Ada, suggested the correct condition among its top picks 70.5 percent of the time versus 82.1 percent for GPs - but a few apps gave urgency advice as safe as the doctors’. The study illustrates the central design tension: a symptom checker can be tuned to be cautious and safe at the cost of being less precise, often by over-referring.
Symptom checkers also carry business and regulatory weight. Babylon Health built a high-profile business partly on its checker and chatbot before collapsing, and the gap between marketing claims and validated accuracy has drawn scrutiny throughout the field.
Why business readers should care: the symptom checker is the clearest example of why health AI is sold as triage rather than diagnosis - safety and accuracy are different goals, regulators watch the claims, and over-caution has real downstream costs.