Protect Patient Data
Learn what can and cannot go into an LLM. Use the PHI safety checker to build good habits from day one.
A free clinical knowledge resource that helps speech-language pathologists use large language models for documentation, goal writing, and clinical workflows.
I/O = input/output. In computing, it's how a system receives and sends information. In SLP, it's what we study every day: how people take in language and produce it. This guide sits at that intersection.
This isn't about learning to use LLMs. It's about articulating what you already know (your clinical expertise) clearly enough that any tool can work with it.
Learn what can and cannot go into an LLM. Use the PHI safety checker to build good habits from day one.
Six patterns that operationalize your clinical knowledge so it translates into better LLM output.
Put it into practice. The specificity of the output depends entirely on the clinical knowledge you bring to it.
Watch how a well-crafted prompt produces clinically useful output. These are pre-generated examples: no API calls, no data sent anywhere.
Enter from any angle. The guide is organized around four questions: What do I need to know? What am I trying to do? Where do I work? Did I do it right?
Core principles, PHI safety, how LLMs work, prompt patterns, evidence base, ethical decision trees, and model comparison.
8 guidesEvery term defined from both sides, clinical and technical. SLPs and developers finally speaking the same language.
25 termsStructured summaries of peer-reviewed studies on LLMs in healthcare and clinical documentation, focusing on what the evidence actually says.
6 studiesEach prompt is an inventory of what a competent clinician knows (assessment frameworks, legal requirements, clinical reasoning) made actionable.
12 promptsMulti-step guides with decision points where clinical judgment overrides the tool. Not just prompts, but sequences.
6 workflowsGoal writing, documentation, communication, compliance, clinical reasoning, and material creation.
6 tasksSchools, medical/acute, SNF/rehab, early intervention, private practice, and university supervision.
6 settingsArticulation, language, fluency, voice, pragmatics, cognitive-linguistic, dysphagia, AAC, literacy, and hearing.
10 domainsMap your clinical philosophy to LLM interaction strategies. Naturalistic, discrete trial, family-centered; each approach is different.
6 patternsBefore-and-after examples that reveal clinical gaps, not LLM mistakes. If you can't spot what's wrong, that's the lesson.
8 antipatternsDe-identified clinical scenarios showing real SLP workflows with LLMs: what worked, what failed, and lessons learned.
5 casesAdaptable LLM governance templates for schools, hospitals, private practices, and clinical education programs.
4 templatesDigital media literacy, full references, changelog, and supporting materials.
3 resourcesSLPs spend more time on paperwork than almost any other clinical task. The people doing that work should have a say in how it evolves, and the tools to make it better should be free.
Documentation eats clinical hours. New tools can help, but only if they're built on real clinical knowledge, not generic LLM hype. Clinicians needed a resource made for them, by one of them.
This resource is built by a school-based speech-language pathologist in New York who also writes software. That dual lens — clinical practice and software development — drives everything here.
Paperwork is where clinical knowledge becomes a legal document. If new tools are going to touch that process, clinicians (not vendors) need to define the standards.
SLP/IO isn't pro-LLM or anti-LLM. It's pro-clinician. This resource is free because the knowledge it organizes already belongs to the profession. We're just making it easier to use.
Every AI query costs electricity and water. A shared, maintained reference is a better use of resources than thousands of clinicians asking chatbots the same questions independently. When you do use an LLM, make it count.
The position: Being thoughtful about AI isn't just about clinical ethics. It's about resource ethics. Reformatting a three-line note isn't worth the compute. Surfacing patterns across a semester of data, or finding language that honors a student's dignity while meeting legal documentation requirements — that's worth it.
SLP/IO is a Harmonic Systems project, built by a speech-language pathologist and software developer who has spent years writing the same IEP goals, progress notes, and evaluation reports that you write.
This resource is free because the knowledge inside it isn't new. It's what competent clinicians already know: assessment frameworks, legal standards, clinical reasoning, ethical boundaries. We just organized it in a way that makes it usable with modern tools. Clinicians deserve to participate in how their paperwork evolves, not have it decided for them.
The question was never "should SLPs use LLMs?" It's "who defines what good looks like?" The answer should be clinicians. This is our attempt to make that happen.
This resource was developed with LLM writing tools and reviewed through clinical judgment.
This guide gets better when clinicians challenge it. If you've found a use case we missed, a principle that needs nuance, or something that's flat-out wrong, that's exactly the kind of feedback that matters.
Send feedback