Honest answer: for a one-off question, go ahead. For a health system that has to read research, grade it, and run for years, a chat window is the wrong shape of tool.
ChatGPT, Claude, and Perplexity are excellent at the question you just asked. They are not the right tool for the thing health actually requires: a system that runs for years, sitting on a structured library of every paper that matters.
ProtocolEngine is that system. The same six-factor grading runs across 35,000+ papers. Every day a new pass picks up what changed. Your stack, goals, and constraints carry forward across sessions. Here is what that means in concrete terms.
ChatGPT answers questions. ProtocolEngine maintains a protocol.
| Capability | ChatGPT / Claude | ProtocolEngine |
|---|---|---|
| Reads, grades, and reconciles the full research base | Training cutoff plus web search. No structured grading of what it finds. | Daily scan of PubMed, bioRxiv, and medRxiv. 35,000+ papers indexed, claims extracted, deduplicated, scored on the same six factors. |
| Same grade for the same question, every time | Answers vary by session, model, and prompt. | The same six-factor algorithm runs on every paper. The grade does not change because you opened a new chat. |
| Every grade traces back to specific papers | Copy-paste, no guarantee citations exist. | Click any study to see how it scored. DOI and PMID on every paper. Exportable summary for your doctor. |
| Remembers your stack, goals, and constraints | Per conversation only. Starts from zero each session. | Persistent, encrypted, under your account. GDPR Article 9. |
| Schedules doses and habits to your calendar | Cannot write to your calendar. | Google Calendar OAuth and .ics export, with reminders. |
| Tracks what you took and how you felt | No logging layer. | Daily journal with energy, sleep, mood, focus, and notes. |
| Tells you when new research changes your protocol | You have to think to ask. | Automatic evidence alerts when grades on your stack change. |
| Flags interactions between your stack and meds | Will mention if you ask. No warning when you add something that conflicts. | Drug-supplement interaction database. Surfaces conflicts at the moment you add to your stack. |
| Uses your genetic data to personalize | Cannot read your files by default. | Upcoming: 23andMe and AncestryDNA raw file upload, with key SNPs mapped to recommendations. |
| Operates under EU health data rules | US consumer product. Your prompts may train future models. | EU hosting. GDPR Article 9 handling. Right to erasure implemented. |
A language model is a brilliant advisor for a one-off question. ProtocolEngine is the operating system for what happens after the question: the structured grading, the schedule, the reminders, the journal, the alerts, the re-grading when new research lands.
The two fit together well. One thinks. The other runs the operation, month after month, without you having to re-explain yourself.