Start with PDFs, not screenshots
ChatGPT reads uploaded PDFs directly. Export your reports as proper PDF files, not screenshots, so the model gets full text, tables, and variant notations.
If you have many reports, upload them in batches grouped by topic (for example, all pharmacogenomics reports together) rather than dumping hundreds of files at once.
Set the context in your first message
Tell ChatGPT what you want and what it is looking at. Example: "These are clinical genomic reports from Dante Labs. Summarize pathogenic and likely pathogenic variants across all uploaded files and list associated conditions."
Ask the model to cite which report each finding came from. This makes it easier to verify claims against the original PDFs.
Useful analysis tasks
Summarize key findings in plain language for a non-specialist audience.
Build a table of all flagged variants with gene, classification, zygosity, and associated condition.
Compare pharmacogenomics results and flag drug-gene interactions worth discussing with a prescriber.
Draft questions to bring to a genetic counselor based on your specific results.
What AI cannot replace
ChatGPT is not a licensed clinician. It can misread tables, hallucinate variant details, or miss nuance in ACMG classifications. Always treat AI output as a draft for your own review, not a medical diagnosis.
For clinical decisions (starting a medication, reproductive planning, or confirming a pathogenic finding), consult a genetic counselor or physician who can interpret results in the context of your full medical history.
Privacy considerations
Uploading health data to any cloud AI service involves privacy tradeoffs. Review OpenAI's data usage policies for your account type. Avoid including your full name, date of birth, or address in prompts if you want to minimize identifiable information.
For maximum privacy, use local AI tools that run on your machine, or redact identifying headers from PDFs before upload.
Need help?
Questions about backing up your reports? Join our Discord.