Why carbs, specifically
Of the three macronutrients, carbohydrates convert to blood glucose the most and the fastest — most of a meal's carbs show up in your blood within an hour or two. Protein converts partially and slowly; fat barely converts at all but changes the timing of everything else. That is why mealtime insulin is dosed against carbs, using a personal insulin-to-carb ratio worked out with your care team — and why an accurate carb number is the input everything else depends on.
Get the count wrong and nothing downstream can save the meal: the right ratio applied to the wrong number is the wrong dose. That makes carb counting one of the few diabetes skills where pure information — not discipline, not restriction — directly moves your outcomes.
One framing note before the mechanics: carb counting is a T1 estimation skill, not a diet. Nothing here is about eating fewer carbs or "good" versus "bad" foods. It is about knowing what number to hand your dosing decision, whatever you choose to eat.
The units: grams, always grams
Carb counting is done in grams of carbohydrate. Older systems talked in "exchanges" or "servings" of 15 grams; some pump-era materials still do. Work in grams — it is what nutrition labels use, what your insulin-to-carb ratio is defined against, and what every food database speaks.
The reference points worth memorizing come faster than you would expect, because everyday eating is repetitive. A slice of standard sandwich bread runs about 15 grams. A cup of cooked white rice, about 45. A medium banana, about 27. A 12-ounce can of regular soda, about 39. Within a few weeks of honest counting, your top twenty foods become known quantities — see the food library for the usual suspects with full numbers.
Reading labels without getting played
Nutrition labels are the easy mode of carb counting, with three traps:
- The serving size is not the package. The bag of chips says 15 grams of carbs — per serving, and the bag holds three and a half servings. Always multiply. This single mistake probably causes more mystery highs than any other label issue.
- Total carbohydrate is the headline number. It already includes the sugar and fiber lines beneath it — do not add sugar on top of total carbs. Fiber is part of the total but behaves differently, which is the net carbs question, covered next.
- Sugar alcohols and "keto" labels play games. Products marketed as low-carb often subtract fiber and sugar alcohols aggressively to advertise a tiny net number. Sugar alcohols vary — some (like maltitol) still raise glucose meaningfully. If a "2 net carb" dessert reliably spikes you, believe your CGM, not the front of the box.
The net carbs question, T1 edition
Net carbs — total carbs minus fiber — makes intuitive sense: fiber is a carbohydrate you mostly do not digest, so it should not need insulin. In practice, most T1 care teams anchor dosing on total carbs and treat fiber as context, adjusting only when fiber is genuinely high. Two reasons: labels already vary enough without adding another subtraction estimated from imperfect data, and the fiber effect is gradual — a lentil bowl with 15 grams of fiber digests slower and flatter, which shows up as curve shape more than as fewer effective carbs.
The practical version: for most meals, count total carbs. For unusually high-fiber meals, expect a slower, gentler curve — and note what actually happens for next time. How to handle it in dosing terms is a fine question for your care team; the net carbs glossary entry has the fuller story.
Estimating without a label
This is the actual skill, because most real food has no label. In rough order of accuracy:
- A kitchen scale plus a database is the gold standard, and far less tedious than it sounds if you only do it for foods you eat weekly. Weighing your usual pasta portion twice teaches you a number you will use two hundred times.
- Measuring cups work for rice, cereal, pasta — the shapeless foods where eyes fail worst.
- Hand references (a fist of cooked rice ≈ a cup ≈ ~45g) travel everywhere and beat unaided guessing.
- Photo estimation — apps like CarbLens AI (disclosure: that one is ours) estimate carbs from a photo of the plate. Where this genuinely helps is mixed and restaurant plates, where eyeball error is worst; where it does not replace you is judgment — any estimate, human or AI, is a starting point you sanity-check, which is why CarbLens marks its numbers as editable estimates and warns you before you ever see a prefilled dose suggestion.
- Unaided eyeballing — everyone's default, and the studies are humbling: typical errors run 20%+ even among experienced adults with T1, and large meals get underestimated the most.
The pattern across all of these: accuracy is a per-food investment. You do not need to be good at estimating food in general. You need to be excellent at your own twenty meals.
Restaurant reality
Restaurants are where good counters go to spike: portions run 1.5–2× home size, fat is everywhere (delaying curves), and sauces hide sugar. The short version of surviving them — chains publish nutrition data, so look it up rather than guess; independents require anchor math (that naan is two slices of bread, call it 30); and the single most powerful trick is repetition: order the thing you have ordered before, because a meal you have watched land three times is not a guess anymore. Restaurant counting gets its own guide: how to count carbs at a restaurant.
Protein, fat, and the second-order stuff
Once carb counts are roughly right, the remaining surprises usually come from the other macros. A very large protein meal with few carbs (the steakhouse dinner) can produce a slow rise hours later as protein partially converts. High fat delays carb absorption — the pizza pattern — pushing the real peak past where mealtime insulin is strongest. Neither changes the carb count; both change the curve's shape and timing. Advanced dosing strategies exist for both and they are squarely care-team territory. Your contribution is the observation: this meal, this shape, every time.
From counting to knowing
Here is the honest endgame, and the reason obsessive per-meal math eventually relaxes: most people eat the same 15–25 meals on rotation. Every one of those meals only needs to be truly figured out once — counted carefully, dosed, observed, adjusted, observed again. After that it is a known quantity, and the mental load drops to nearly zero.
This is the deepest argument for logging meals alongside your glucose data, in whatever tool you will actually keep using. A log converts effort you are already spending into permanent knowledge: the burrito is 62 and needs respect, the lentil soup is 30 and gentle, mom's lasagna is 55 and runs late. Carb counting as a lifetime of arithmetic sounds exhausting. Carb counting as a finite project — learn your meals, keep the receipts — is genuinely finishable, and the numbers stop being the boss of the meal.
