For clinicians who use AI
Capture how you practice once. Use it in Claude, ChatGPT, Gemini, OpenEvidence - anywhere you work with AI. No more re-explaining yourself in every session.
The problem
You open Claude, ChatGPT, or Gemini. The AI doesn't know you're a board-certified physician. So you get hedged answers. Disclaimers. Mechanism explanations you already know. Citations from secondary aggregators instead of primary literature.
You can fix it by typing custom instructions into each tool, every time. Most clinicians don't, because it's a tax on every single session.
Idiolect makes the fix permanent - and portable across every tool you use.
How it works
37 questions covering your specialty, communication style, evidence preferences, mechanism teaching, and the things you don't want AI to do. One screen per question.
Idiolect generates a ready-to-paste prompt for each major AI tool - plus an editable source-file portfolio underneath so nothing's locked in.
Paste into Claude Projects, ChatGPT Custom GPT, OpenEvidence Dotflow, or Gemini Gem. Step-by-step guides for each. From then on, every AI knows who you are.
What you get
One prompt per platform, tuned for that tool's quirks and character limits. Copy it once, paste where Idiolect tells you, and every clinical question you ask from then on gets the right context.
For Claude and ChatGPT, you also get a set of Knowledge files to attach for richer context. For OpenEvidence Dotflows and Gemini Gems, the prompt itself is the whole package.
Under the hood
The prompts are assembled from ten plain-text markdown files — your specialty, your tone, your evidence preferences, your hard "don'ts." Edit any one. Share them. Version them. Re-export to any AI tool that comes along.
identity.mdSpecialty, training stage, practice settingcommunication-style.mdTone, terminology, structure, lengthclinical-identity.mdSubspecialty focus, priority domainsclinical-reasoning.mdMechanism preferences + proactive surfacingevidence-preferences.mdCitation depth, format, guideline freshnesspreferences-and-constraints.mdWhat you don't want AI to dotool-preferences.mdPer-tool configurations, ready to pastecurrent-projects.mdActive projects AI should know aboutteam-and-relationships.mdPractice partners + consult patternsgoals-and-priorities.mdLearning, career, panel/community goalsWorks with
Project instructions + Knowledge files. Strongest compliance with proactive-surfacing directives.
GPT instructions + Knowledge attachments. Compressed to fit Custom Instructions limits.
Pre-built Dotflow text to paste into OpenEvidence settings. (Dotflows launched April 2026.)
Compressed Gem instructions. Tone and structure land well; surfacing is variable.
Why Idiolect
Pathophysiology recall calibration. Proactive surfacing for "things you might have missed." Behavioral prohibitions like "don't lecture me on basics." The difference between an AI that feels like a colleague and one that feels like a chatbot.
OpenEvidence has Dotflows. ChatGPT has Custom GPTs. Each works only inside its own product. Your Idiolect portfolio works in all of them - and any new AI tool that comes along - because markdown is the universal interchange format.
Pseudonymous by default. Your portfolio lives in your browser. Create an account only if you want to come back and edit. We store nothing about you without explicit opt-in.
Idiolect captures professional identity. It doesn't diagnose, treat, recommend, or store patient information. Use it to configure the AI tools you already trust - without becoming dependent on yet another platform.
Privacy
FAQ
Dotflows are great - but they only work inside OpenEvidence. Your Idiolect portfolio works in OpenEvidence and in Claude, ChatGPT, Gemini, and any new tool you adopt. Idiolect actually generates a Dotflow for you to paste into OpenEvidence as part of the output.
Idiolect adopts the same file convention as NLW's framework, but with clinical-specific categories - mechanism preferences, evidence specification, proactive surfacing for things you might miss, behavioral prohibitions. Your Idiolect files work in NLW's MCP server too if you want.
Not by default. Everything lives in your browser. You can create an account (optional) if you want to sync across devices or come back to edit later - but you're not required to.
The schema is built to handle any specialty. We have tested it against Family Medicine, Cardiology, OB/GYN, and Emergency Medicine. Some chip lists are FM-centric in v1; the free-text fallbacks absorb specialty-specific content well.
Compliance varies by tool. Claude follows them most consistently. ChatGPT is inconsistent (especially on the proactive-surfacing directive). Gemini is variable. We're honest about this on each integration guide.
Free for individual clinicians at launch. Future tiers may exist for health systems and pro-tier feature expansions, but the core individual experience will remain free.
Fifteen minutes once. Used in every AI session forever.
Build your portfolio →No signup. No patient data. Your files, your control.