Search for a pill identifier by picture app and you get a page of store listings, each promising to identify any pill from a photo. What none of them explains is what that promise involves, which makes it hard to tell a tool that reads the tablet from one that guesses at it. This post is that explanation, so that whichever app you open, you know what to expect from it and where it should stop.
A photo identifier worth trusting does four specific things. It reads the imprint code from the image, because the imprint is the only feature manufacturers are required to make unique. It uses shape and color to narrow the match rather than to make it. It matches against current product labeling, the kind manufacturers file with the FDA, because appearance changes whenever a manufacturer changes. And it reports doubt honestly, with a visible confidence signal and a clear “no match” when the tablet is unmarked or worn, instead of returning its best guess as if it were an answer. The pharmacist then confirms the read, free, in minutes.
The three things a photo can carry, and which one does the work
A photograph of a tablet holds three usable pieces of information: the imprint, the shape and the color. They are not equal.
Federal regulation, 21 CFR Part 206, requires solid oral human drug products, prescription and over-the-counter alike, to carry a code imprint that, in conjunction with the product’s size, shape and color, permits unique identification of the product and of the company that made or distributed it. The regulation puts the burden on the imprint. Size, shape and color are supporting context, and the rule says so in as many words.
The reason is practical. Color and shape are manufacturer choices with no standardized meaning. The same medicine is made by many companies in many colors, and one small white round shape is shared by an enormous number of unrelated products. A tool that says it identifies “by color and shape” is running a filter that returns a long list of candidates and then picking one, which is a guess wearing the clothes of an identification. A tool that reads the imprint first has one or a handful of candidates before shape and color are ever considered, and those two attributes then separate the remaining candidates cleanly.
So the first thing to look for is simple. Does the app read the code on the tablet, and does it show you what it read? If the result does not tell you the imprint it recognized, you cannot tell whether it recognized anything. The structure of those codes and how to read a worn one is covered in what pill imprint codes mean.
What the data behind a good identifier is
Reading the imprint is half the job. The other half is what it gets matched against, and here the important word is current.
A tablet’s appearance and imprint belong to a specific product from a specific labeler at a specific strength. When a manufacturer changes its formulation, when a new generic enters the market, or when a company is acquired and its codes are reassigned, the mapping from imprint to product changes with it. An identifier built on a static list compiled some years ago drifts out of date without anyone noticing, and the failure mode is quiet: a confident result for a code that now belongs to something else.
The reference that stays current is product labeling. DailyMed, run by the U.S. National Library of Medicine, holds the most recent labeling submitted to the FDA by companies for prescription and over-the-counter products, with well over a hundred thousand labels, though not a complete catalogue of everything on the market, and much of that labeling describes the product’s appearance and imprint. A good identifier’s matches trace back to that kind of labeling ecosystem, and the result it shows you should let you see the labeler and the product it matched, not just a drug name. That is also what lets a pharmacist confirm the read quickly: the same product name and labeler are on the dispensing record.
This is why a blog post is the wrong place to look up an imprint, and why this site does not publish tables of codes. A table is right on the day it is written and wrong at some unknown point afterward. Identification has to happen against live labeling, and then be confirmed by someone with the dispensing record.
Signs of honesty in a tool
The hardest thing for any identifier to do is say “I do not know”, and it is the thing that most distinguishes a useful one. The physical facts make it unavoidable. Imprints wear. Coatings chip. Tablets get snapped in half with the code on the missing piece. Some tablets never had a code, because supplements, compounded medicines, veterinary products and many products from outside the United States sit outside the imprint rule entirely; the categories are set out in what an unmarked tablet usually is. For all of these, the honest output is a partial match, a set of candidates, or no match, and a tool that never produces those outputs is not being careful, it is being confident on your behalf.
Three things to look for in the result screen, described as approaches rather than as any particular product:
- A confidence signal you can see. Whether it is a percentage, a band, or a plain statement that the match is strong or weak, the tool should tell you how sure it is, and a weak match should look different from a strong one.
- A clear no-match state. When the imprint cannot be read, or reads to nothing in current labeling, the tool should say exactly that and suggest what to do next: reshoot in better light, try the other face, or take the tablet to a pharmacist. It should not fall back on shape and color and present the result as if nothing had changed.
- No claim of certainty. Language like “guaranteed” or “always right” in a listing is a signal about the listing, not the tool. Identification from a photograph is a strong first read, and the tool that describes itself that way is the one describing itself accurately.
A fourth sign, less obvious: the tool should make it easy to take a better photograph. Most failed reads are lighting failures, since debossed codes on light tablets disappear under flash and appear in low side light. An identifier that coaches the shot is one that has been built by people who understand where the errors actually come from. The full technique is in how to identify a pill from a picture, and the light, background and focus are covered in how to photograph a pill for identification.
Where the tool ends and the pharmacist starts
A photo identifier answers one question: which product does this tablet appear to be? That is the question you have when a loose tablet turns up on the floor, when a refill looks different from last month’s, or when two bottles got mixed. It is a first read, produced in seconds, and its job is to turn a vague worry into a specific candidate.
Confirming that candidate is the pharmacist’s job, and it is the natural next step rather than a reluctant one. The pharmacist has the dispensing record, can check the product and labeler against what was actually supplied, and can answer the questions a tool cannot: whether this belongs to anyone in the household, and what to do with it if it does not. That check is free and takes a couple of minutes at the counter or on the phone. Anything about whether a medicine should be taken, in what amount or when, was never the tool’s question and is not the pharmacist’s alone either; it belongs to the prescriber and the label.
Time-critical situations sit outside the tool altogether. If a child or a pet may have swallowed something, contact Poison Help at 1-800-222-1222 in the United States, or emergency services, before trying to identify anything. The people who answer can work from a description.
It is worth knowing how the FDA frames software of this kind. Its oversight of device software functions concentrates on software that presents greater risk if it does not work as intended, such as software that diagnoses a condition or directs treatment, and it exercises enforcement discretion for lower-risk functions that help people manage their own health without providing specific treatment suggestions. That framing is a useful description of what a reference tool is for: it gives you information to take to a professional, and it does not tell you what to do with a medicine. No regulatory status is claimed here for any particular app, including this one; the point is that an identifier should behave like a reference and not like a diagnosis.
A checklist before you trust a pill identifier by picture app
Bring these to whichever app you are evaluating.
- It shows you the imprint it read, so you can check it against the tablet with your own eyes.
- It uses shape and color to narrow the match, and it says so, rather than presenting a color match as an identification.
- Its results name the product and the labeler, and trace to current labeling rather than to an undated list.
- It has a visible confidence signal and a real no-match state, and it reaches for them when the tablet is worn, broken or unmarked.
- Its listing does not promise certainty, and its results point you to a pharmacist to confirm.
- It helps you take a better photograph when the first one fails.
An app that meets those is a good first read on the tablet in your hand. RxVue was built to that description, and the fastest way to judge any identifier against it is to photograph a tablet you already know and see whether the tool tells you what you can see for yourself. If cost is the deciding question, free pill identifier app sets out what free tools still cover now the government’s own tool is gone.