The number of individuals who attend medical consultations with a specific medical label in mind is increasing, though many of those conclusions are inaccurate. In a survey of 2,345 adults in the United States published by JAMA Dermatology on April 15, 2026, the use of an artificial intelligence tool for skin conditions increased the proportion of participants who correctly identified a condition from 7.86% to 22.79%. A separate group that received information from medical specialists achieved a rate of 36.20%. The proportion of participants willing to suggest any diagnosis rose from 41.21% to 62.26%. To a medical office, the statistics indicate that more individuals present a specific medical term, approximately 75% of those terms remain incorrect and the individuals possess higher levels of certainty than in previous years. It is an architectural challenge for the intake process before it becomes a clinical difficulty.
The measurements within the dermatology research
The researchers provided participants with anonymous medical cases and requested that they identify the condition. The control group achieved a correct identification rate of 7.86%. The group using the artificial intelligence tool reached 22.79%. The group provided with information from specialists reached 36.20%.
The increase compared to the control group receives the most attention – but the 13-point difference between the artificial intelligence group and the specialist supported group is the more informative data point. Even the participants with the highest level of information were incorrect in nearly 66% of cases. The group using artificial intelligence was incorrect in more than 75% of cases. Google Research, which produces the technology for those tools, states that the systems exist to help individuals perceive and explain observations rather than to provide a formal diagnosis.
The authors of the study emphasize a second point regarding operations. In their findings, the increase in patient certainty happened more quickly than the increase in functional medical knowledge, especially concerning the necessity of treatment, the level of urgency and the timing of follow up care. Individuals became more willing to name a condition faster than they became accurate in identifying it.
The effect of rising certainty on intake documentation
Intake documents rest on a premise that is no longer valid for many individuals – this premise assumes that the individual completing the form does not know their condition and will describe what they see or feel.
A patient who believes they have identified their diagnosis provides a description of that diagnosis instead of their symptoms. “Eczema on my forearm, three weeks” replaces “red scaly patch, itchy at night, started after I changed detergents”. The second phrase is helpful for clinical work. The first phrase is a final conclusion that lacks evidence, which requires the clinician to work in reverse to find the information that the patient omitted.
It is not necessary to view those individuals as untrustworthy – but it is necessary to ask different types of questions.
Intake structures that fail when the patient provides a diagnosis
Three standard methods of data collection fail in specific ways:
| Intake method | Effect on a patient with a self diagnosis | Functional solution |
|---|---|---|
| Open ended complaint field | The individual enters the name of a condition instead of observations – specific details about symptoms disappear | Divide the field: “What do you think this is?” and “What are you noticing?” as two different questions |
| Branching logic for triage | The individual provides answers to support the label they already chose – the logic follows their mistake | Base the initial questions on visible facts like duration, location and fever before using any medical names |
| Image uploads with brief history | The quality of the image is high but the written history becomes a single word | Require specific fields for history alongside the image rather than making them optional |
The trend in the three methods is that the document encourages the individual to shorten their entire experience into a single label.
Information to collect instead
Four questions gather the details that a label removes without suggesting that the individual is incorrect:
- What is your opinion on what this might be? Collect this information in a specific place. If there is no specific field, the individual places it in the primary complaint field and influences all subsequent data.
- What information led you to that conclusion? This could be a search engine, an artificial intelligence assistant, a relative with a similar condition or a previous occurrence. Each source provides a different level of proof.
- What actions have you already taken because of this belief? Treating oneself based on an incorrect label is a significant clinical fact and is a detail that digital intake forms often miss.
- What are you specifically observing? Mention the start time, the duration, changes over time and factors that improve or worsen the state. On the form keep this field separate from the diagnosis field so the two types of information do not merge.
Structured fields provide a capability that a free text box does not. The system is able to measure how frequently patients arrive with a self diagnosis and how frequently those patients are correct. There is a measurable data set regarding your patient population after multiple hundred visits rather than an individual opinion.
The impact on appointment duration
The act of correcting a patient answer requires more time than obtaining an original answer. The clinician is required to identify what the patient believes, determine the reasons for that belief and establish a new perspective before the medical history begins – that labor occurs most often during the shortest appointment categories, where the schedule lacks flexibility to accommodate the extra time.
The pressure is not equal across a medical office – telehealth represents 7.2% of primary care interactions as of June 2026, according to Epic Research, but the distribution concentrates in specific areas. The data from FAIR Health shows that mental health conditions represent 52.1% of telehealth insurance claims nationally in the first quarter of 2026 – this is the primary diagnostic category for every age group. Medical offices with many virtual appointments for short durations experience this effect first.
Circumstances where this change is beneficial
An advantage exists that you can plan for – asynchronous care and store-and-forward models receive improved data from a patient who has researched their condition, provided you ask specific questions at the start.
A patient who researches their medical condition often provides a clearer photograph, a more exact timeline and a more detailed description of previous treatments when the form requests the items. In an asynchronous process where the clinician is not able to ask a follow up question immediately, a detailed structured intake makes the model functional. The same patient behavior that increases time in a seven minute live visit can eliminate an entire message cycle in an asynchronous one.
Necessary adjustments and implementation difficulty
The task involves system configuration rather than software development. Adding two data fields, changing the order of triage logic so facts appear before diagnostic labels and requiring a history for photo uploads are changes a medical office can perform.
But the feasibility of this depends on the software – this is one reason that medical offices evaluating white label telehealth platforms prioritize the ability to change intake forms as much as video quality or legal compliance. Custom forms and branded onboarding are standard features in modern software. Healee’s technology serves over one million patients and five million appointments across more than 200 clinics. On this platform the workflows function at a high volume. If you want to observe how the intake builder manages those patterns, you can request a demo.
The general conclusion remains the same for any software – consumer AI tools change the information patients provide at the start of a visit. The intake form determines if that change assists you or creates a cost and the form is the least expensive part of the system to modify.
Frequently asked questions
What is the accuracy of consumer AI tools for skin conditions?
In the JAMA Dermatology study, participants using an AI tool for dermatology correctly identified the condition 22.79% of the time – this compares with 7.86% without a tool and 36.20% for those with human expert support. The tools increase understanding but most users receive an incorrect answer.
Should I prevent patients from entering a suspected diagnosis?
No. To hide the field does not remove the patient’s belief – it only hides the belief from the clinician. It is better to collect the diagnosis in one field and collect observations in another so both are available.
Which appointment types experience the most impact?
The impact is greatest on short live appointments where correcting a patient assumption uses a large portion of the time. Virtual urgent care and short follow up visits experience this before long consultations.
Is the effect different for asynchronous care?
Yes. Asynchronous workflows benefit from patients who conduct research if the intake process requests a structured history alongside images or messages.
How much time is required to change intake forms?
On a system with adjustable settings, adding fields and changing logic takes only hours. On systems with fixed forms, the change requires a request to the vendor and a software update cycle.
How would I identify if this is occurring in my medical office?
You add a structured “what do you think this might be” field and compare the data against the final diagnosis after a few hundred clinical visits – this specific assessment indicates the exact frequency at which patients present with a prior self identification of their condition. It also shows how frequently those patients are accurate in their initial self assessments.
Sources:
- Consumer Understanding of Skin Concerns With an AI-Powered Informational Tool – JAMA Dermatology, 2026
- Research into how AI can help users understand skin conditions – Google Research, 2026
- Mental Health Conditions the Top-Ranking Telehealth Diagnostic Category in Every Age Group in First Quarter 2026 – FAIR Health via PR Newswire, 2026
- Telehealth Utilization Data Tracker – Epic Research (Cosmos), 2026