Mismatched medication records are a well-known, preventable source of medication errors - a dose changed by one prescriber that never made it into another’s chart, a pharmacy’s records still showing a medication that was stopped months ago, a supplement or over-the-counter product nobody thought to mention because it did not feel like “real” medication. Every one of these gaps is a text-comparison problem before it is anything else, and text comparison is something AI can genuinely help with, carefully.
This is a method for reconciling a medication list from multiple sources into one document - built entirely from what your labels, prescriptions, and records actually say, with every mismatch flagged rather than resolved. The reconciliation itself, deciding which record is correct and what to do about a conflict, stays with a pharmacist or your clinician.
Never ask AI to recommend starting, stopping, combining, substituting, or changing the dose of any medication - and never let a mismatch between records tempt you into deciding on your own which one to follow. If you find a discrepancy, treat it as a question for your pharmacist or clinician, not as something to resolve yourself before that conversation.
Step 1: Gather every source before you start comparing
Collect every record that lists your medications: pharmacy printouts, prescription labels themselves (not just the bottle - the label, which often has more detail), your patient portal’s medication list, and any paper list from a recent hospital stay or specialist visit. Do not rely on memory for any of this - the entire value of reconciliation comes from comparing what is actually printed on each source, not what you believe you take.
Step 2: Transcribe exactly, then let AI format
Type each source’s medication information exactly as printed - resist the urge to “clean it up” as you type, since AI is much better at standardizing formatting than at guessing what an ambiguous abbreviation on a label meant.
Here is a medication list exactly as printed on [source name]:
[paste, exactly as written, including any abbreviations or unclear
formatting]. Format this into a table with columns: medication name,
dose as printed, schedule as printed, and prescriber if listed. Do
not interpret, standardize, or correct anything - if something is
ambiguous or hard to read, mark it as "unclear" rather than guessing.
Repeat this for each source separately before combining anything - comparing already-merged lists hides exactly the discrepancies this method exists to surface.
Step 3: Compare sources side by side and flag differences
Once each source is in its own clean table, ask AI to compare them and flag every difference - without commenting on which one is likely correct or what should be done.
Here are medication lists from [N] different sources: [paste each
table, labeled by source]. Compare them and produce a combined table
with one row per distinct medication. For each one, show what each
source says (name, dose, schedule) side by side, and flag clearly
whether the sources agree or disagree. Do not suggest which source is
more likely correct, and do not suggest any change to any medication -
only identify and flag matches and mismatches.
A mismatch is not always an error - a dose may have changed recently and one record simply has not caught up yet. That judgment, too, belongs to whoever reconciles the list professionally, not to the comparison tool.
Step 4: Add the fields a pharmacist actually needs
A useful reconciliation table needs more than name and dose. For each medication, record when each source was last confirmed accurate, and who prescribed it - both make it much faster for a pharmacist or clinician to resolve a flagged discrepancy.
Add two columns to this reconciliation table: "last confirmed" (the
date I last verified this entry against its source) and "discrepancy
flag" (yes/no, plus a one-line note on what disagrees). Do not fill
in dates or flags I did not provide - leave them blank if I have not
given you that information.
Step 5: Do not forget what does not feel like “medication”
Reconciliation commonly misses supplements, over-the-counter products taken regularly, and anything prescribed by a dentist, physiotherapist, or specialist seen only once. Add a dedicated section for these rather than folding them quietly into the main list, since they are exactly what gets left out of a verbal medication history and exactly what can interact with prescribed medications in ways only a pharmacist can properly assess.
Step 6: Bring flagged discrepancies to a pharmacist or clinician - do not resolve them yourself
Once the table is complete, every flagged row is a specific question, not a decision for you to make. “This pharmacy’s record shows 10mg, this prescription says 20mg - which is current?” is exactly the kind of question a pharmacist can resolve quickly with access to the full prescribing history, and exactly the kind of question that should never be settled by picking whichever number appears more often, or newer, or on a source you trust more.
The full medication list reconciliation table gives you this structure ready to use, including the section for supplements and over-the-counter products and a checklist to run before bringing the table to a professional.
Pair this with the broader appointment-preparation method in preparing for a medical appointment in 20 minutes if the reconciliation is happening ahead of a scheduled visit rather than a standalone pharmacy consultation.
A full medication list is sensitive personal health data, and typing every source into a general AI chat account leaves an extended record of your prescriptions in that account’s history. Use a tool with training turned off where available, and see what ChatGPT remembers, sees, and shares before reconciling a list in a shared device or account.
Reconciling for someone else
If you are building this list on behalf of a family member rather than yourself - a common situation for caregivers coordinating an ageing parent’s medications, as covered in coordinating an ageing parent’s care - the same transcribe-then-compare method applies, with one addition: note who collected each source and when, since a caregiver working from a photo of a pill bottle taken last month is a meaningfully different source than a pharmacy record checked yesterday.
Here is a medication list I collected on behalf of [person], along
with when and how I collected it: [paste sources and collection
notes]. Add a column noting who collected each entry and when. Do not
change any medication information based on this context - only add
the collection metadata.
Common pitfalls
- “Cleaning up” a label before transcribing it. Type it exactly as printed, unclear parts included - let the comparison step, not your own memory, decide what needs clarifying.
- Asking AI which source is correct. Comparing text is different from adjudicating medical accuracy; the second belongs to a pharmacist with access to the full prescribing record.
- Leaving out supplements and one-off prescriptions. These are exactly the entries most likely to be missing from at least one source, and most likely to matter for interaction checks.
- Resolving a discrepancy yourself before the appointment. A flagged mismatch is a question to bring, not a choice to make on your own between two sources.
- Skipping the “last confirmed” date. Without it, a pharmacist cannot tell whether a mismatch reflects a stale record or a genuine live conflict.
Do it before your next pharmacy visit or appointment
Gather every source, transcribe each one exactly, compare them for discrepancies, add the fields a pharmacist needs, and include anything that does not feel like “real” medication. Use the medication list reconciliation table to hold the comparison, and bring every flagged row to a pharmacist or clinician rather than deciding on your own which record to trust.
