Here is a thing nobody warns you about modern healthcare: your lab results show up in the patient portal the instant the machine spits them out, days before any human being is scheduled to explain them to you. So at 11pm on a Tuesday I got a push notification, opened the app in bed, and was greeted by fourteen rows of numbers, half of them wearing little orange 'flag' icons like tiny sirens. My ferritin had a flag. My TSH had a flag. Something called 'AST' had a flag and I had no idea I even owned an AST.
I did what any reasonable adult does. I read each one out loud to my partner in an increasingly panicked voice, then I googled 'ferritin 11 low,' which is a direct portal to the worst forty minutes of your week. By midnight I had diagnosed myself with four conditions, none of which I had, and had read at least one forum post by a person who was very confident I had a fifth. I slept badly. I woke up and immediately reopened the app to check that the numbers hadn't gotten worse overnight, as though my blood was being updated live.
Why a table beat my 11pm spiral
The next morning, with slightly more blood reaching my brain, I pasted the whole panel into the Decode My Lab Results Into Plain English Before the Follow-Up Call prompt and ran it on Gemini 3 Pro Preview. I chose Gemini on purpose: a fourteen-row panel is a long, structured input, and Gemini is genuinely good at holding a big table in its head and turning it into another clean table without dropping rows or hallucinating a fifteenth test. That mattered. I did not want a chatty paragraph that buried the signal. I wanted a grid I could scan in one pass.
What came back was exactly a grid. One row per test, my value next to the reference range, a plain-English 'what it measures,' and — this was the part that actually unclenched my shoulders — a 'common things that move it' column. For my low ferritin it noted that recent blood donation, diet, and menstrual timing can pull it down. I had donated blood nine days earlier. I had completely forgotten. The orange siren was, very likely, a Red Cross appointment and a cookie.
The three sections that did the real work
Below the table the prompt gave me a 'Worth asking about first' section that ranked which two values were actually worth raising on the call and which were noise. It put my borderline TSH near the top with a one-line question to ask, and gently parked the AST flag with a note that mildly elevated values are common and have many ordinary causes. Crucially, it never told me anything was fine and never told me anything was dangerous. It described, then handed me back to my doctor — which is the only honest thing an AI should do with a lab result, and the thing I'd actively distrust it for if it did otherwise.
It also opened with the readability-aid disclaimer, in plain text, every single time. That sounds like boilerplate, but it did real work on my anxiety. It kept reminding me that I was reading an explanation of words my doctor had already used, not receiving a verdict. The difference between 'understanding my results' and 'diagnosing myself' is the entire ballgame, and the prompt stays on the right side of that line so consistently that I stopped trying to drag it across.
- It opened with the readability-aid disclaimer, so I stayed calibrated about what I was and wasn't learning
- The 'likely innocent explanations' section only used facts I'd given it — it caught my blood donation but didn't invent anything
- It turned fourteen scary rows into three things worth asking and eleven things to stop refreshing
- The 'bring to the appointment' list reminded me to dig out last year's panel for comparison
- It never once told me a number was 'bad' — it described and redirected, which is the only behavior I trust
The boring, eleven-minute appointment I actually wanted
On the call, I was the most boring patient my doctor had all week, in the best way. I led with the two ranked questions. She confirmed the ferritin was almost certainly the donation, put the TSH in context of nothing-yet-worth-treating, and we were done in eleven minutes. No spiral, no four imaginary diagnoses, no 1am WebMD rabbit hole with a stranger telling me to get my affairs in order. The whole reason an eleven-minute appointment can be useful instead of frustrating is that you arrive already understanding the words, so the doctor's scarce minutes go to interpretation and next steps instead of vocabulary.
If your portal also likes to ambush you at bedtime, grab the Decode My Lab Results Into Plain English Before the Follow-Up Call prompt on Prompt Dock and run it on your own panel before you let Google anywhere near it. Worst case you understand your bloodwork. Best case you remember you gave blood last week and you get to go back to sleep. When I pair this with my Prep Questions Before a Doctor Appointment prompt, I show up with the panel decoded AND an agenda, and the appointment runs like I'm a tiny project manager of my own bloodstream.