Dermatology doesn't need another AI chatbot. It needs a reference.

It's 5:30 p.m.

You've seen your last patient, but your workday isn't over. There are charts to finish, patient messages to answer, pathology results to follow up on, and tomorrow's schedule to prepare.

You became a dermatology professional to care for patients. Yet more and more of your day goes to the information surrounding that care.

Artificial intelligence promises to help. You've seen the demos: models that describe a rash, summarize a paper, draft a note or answer a medical question in fluent paragraphs.

But here's a question worth asking:

If AI is becoming so capable, why does using it still feel like more work, not less?

Often, it's because you have to check everything it says. An answer you can't verify isn't help. It's another task.

That's the reason we're building DermLiving.

The problem isn't that AI isn't smart enough

Imagine hiring an exceptionally well-read medical assistant.

This assistant has read thousands of textbooks, understands the terminology, and can explain nearly any dermatologic condition.

But there's a catch.

They can't tell you where any of it came from. They don't know which guideline they're paraphrasing, or whether its recommendations have changed since they read it. They can't tell the difference between what they know and what they're guessing. And they'll answer questions they should leave to you.

So despite being remarkably intelligent, they create work: everything they say has to be checked.

That's the challenge facing much of medical AI today.

General-purpose AI models are becoming extraordinarily capable. But capability alone doesn't make a system clinically useful. To be useful in dermatology, AI needs reliable specialty knowledge, sources you can check, and clear limits on what it will answer.

Dermatology is more complicated than it looks from the outside

Consider a patient with a persistent inflammatory skin condition.

For an experienced dermatologist or PA, evaluating that patient involves much more than identifying a rash. You're weighing morphology, distribution, duration, symptoms, prior treatments, medication history, associated conditions and, often, several competing diagnoses.

The knowledge behind those judgments is scattered across clinical guidelines, published research, drug labeling and reference texts. Each source holds a different piece of the picture, and each one changes on its own schedule.

Even the names don't line up. A patient says "eczema." The chart says atopic dermatitis. The bill says ICD-10-CM L20.9, and ICD-11 says EA80.Z. Software has to know these mean the same condition, and to notice when they don't: "eczema" can mean more than one thing.

The opportunity for AI isn't simply to generate an answer. It's to bring the right information, with its source, at the right moment.

That's a fundamentally different challenge.

What a better clinical tool could look like

Here is what that difference could feel like.

During a visit, you meet a less familiar presentation and ask the tool your practice uses about it. Instead of a confident paragraph, you get a short answer where every sentence carries a footnote: the source, the section and the version it came from. Open a footnote and you can read the passage yourself.

If the reference doesn't cover your question, the tool says so instead of guessing. Ask it what to prescribe, and it declines plainly. That decision is yours.

Now imagine the same discipline in the other software around your practice. A patient-education draft in which every statement points to a source you can review. A coding helper that knows atopic dermatitis is L20.9 and EA80.Z, and asks before it assumes what "eczema" means.

These aren't promises that every workflow can or should be automated. They're examples of what becomes possible when AI tools are grounded in reliable knowledge and built to know their limits.

They also illustrate an important distinction:

The most valuable AI in dermatology may not be the AI that makes the diagnosis. It may be the AI that shows its work, and knows when to stop.

Why hasn't this already been built?

Because most AI applications start from general-purpose technology.

A software company can connect a powerful language model and ask it questions about dermatology. That produces impressive demonstrations.

A dependable clinical application needs much more. Its developers have to organize dermatology-specific knowledge, connect terminology across code systems, retrieve supporting evidence, keep track of updates, handle uncertainty and keep patient information out of places it doesn't belong. They have to decide what the AI may answer, what it must refuse, and how its behavior will be tested.

Every company building dermatology AI faces some version of these problems. Many solve them separately, and much of their engineering goes into rebuilding the same foundation instead of improving what clinicians actually see.

We believe there's a better approach.

Introducing DermLiving

DermLiving is a dermatology knowledge API: a reference that software builds on, so each company doesn't have to build its own.

Think about how online payments work. When a store takes a card, it doesn't build a payment network. It connects to one built for that job.

We're applying the same idea to dermatology knowledge. When an app asks DermLiving a question, here is what comes back:

Underneath, every condition is a structured record with its ICD-10-CM and ICD-11 codes. Every answer is split into claims, and each claim names its source, the source's version and the section it came from. Every response says which version of the knowledge it read. When no passage supports an answer, DermLiving says so.

For dermatologists and PAs, the technical details aren't the important part. What matters is what this foundation makes possible: AI tools that show their evidence, decline what isn't theirs to decide, and say plainly when they don't know.

DermLiving is early, and we want to be precise about where it stands. Today it runs as a free sandbox for software developers, on a small synthetic dataset that shows the format and the limits. It holds no licensed clinical content yet, and it is not for patient care. Licensed sources, each with its usage rights recorded, and review by clinicians come before any real content is served. The clinical experiences described in this post are our vision, not capabilities that have been deployed or clinically validated.

Your expertise becomes more valuable, not less

There's an understandable concern whenever AI enters a medical specialty.

Will it replace clinicians? Will it encourage patients to bypass professional care? Will organizations put automation ahead of clinical judgment?

Those are legitimate questions.

We believe the most promising role for AI in dermatology is to support clinical expertise, not imitate it.

Dermatology requires judgment that goes far beyond recalling information. It means recognizing uncertainty, interpreting findings within a patient's broader history, understanding what the patient wants, and making decisions under real-world constraints.

Those responsibilities stay with qualified clinicians. That's why DermLiving's limits are part of its design, not an afterthought: it does not diagnose, score malignancy, choose or dose treatment, or triage. It returns reference knowledge, and it says where that knowledge came from.

AI can help organize information and retrieve evidence. It must also be evaluated for accuracy, bias, privacy and safety.

Consider two futures.

In one, clinicians are asked to trust increasingly powerful black-box systems.

In the other, clinicians use tools that make their evidence transparent, take friction out of finding information, and leave the judgment to them.

We're building toward the second.

Why we're starting with dermatology

Dermatology is a specialty where visual examination, precise terminology, observation over time, medical evidence and patient communication all matter.

Consider how much of a dermatology professional's day moves between different kinds of information: clinical photographs, treatment histories, medication lists, pathology reports, published research, patient instructions and follow-up notes. Each often lives in its own system.

The challenge isn't only that there's too much information. It's that the information isn't connected in ways that support the work.

We're starting with the knowledge layer because dependable clinical AI has to be grounded in reliable information before anything built on it can be trusted.

Built with clinicians, not just for them

The people building healthcare technology don't always understand what happens during a busy day in a dermatology practice.

A feature that looks impressive in a demonstration may be impractical in a real clinic. An AI assistant that saves thirty seconds on one task but adds three review steps isn't an improvement. And a tool that doesn't fit the way you already work may never be used, however sophisticated it is.

That's why clinical input is essential.

We want to understand where dermatologists and PAs actually lose time to information. Not where technology companies assume the problems are. Where they are.

Which questions interrupt patient care most often? What information is unnecessarily hard to find? Where would faster access to reliable evidence make a real difference?

Clinicians will also shape the reference itself. We're forming a network of dermatologists, PAs, NPs, residents and students to review reference passages before software can cite them, to mentor students, and to bring care to communities where dermatology is hardest to reach.

A better future for dermatology

We believe the next major advance in medical AI won't come from making chatbots sound more like physicians.

It will come from connecting capable AI systems to reliable clinical knowledge, with clear limits, and to the professionals responsible for patient care.

That's the future we're working toward with DermLiving.

Not artificial intelligence replacing dermatologists. Not another disconnected application demanding your attention. A foundation for technology that understands the specialty, respects clinical judgment and shows its sources.

Because the measure of healthcare technology isn't how impressive it looks in a demonstration.

It's whether it makes life better for the people delivering care, and for the patients who depend on them.

Help us shape DermLiving

We're listening to dermatologists and dermatology PAs to understand the problems they face in everyday practice. One question matters more than the rest:

If you could hand one part of your clinical day to a tool you could trust, which part would it be?

The practitioners page explains how clinicians will take part. Sign-ups aren't open yet; they open there first.