Why does food logging take 23 minutes a day at first?

What did the trial actually measure?

Time on the record, from the record. In a 2019 trial by Harvey, Krukowski, Priest and West, published in Obesity, 142 adults kept an electronic food diary for 24 weeks, and the time they spent was read from the site's logs rather than asked. It came to 23.2 minutes a day in the first month, and 14.6 by month six.

The finding the authors put in their title was not the time. It was that how often people recorded, not how long they spent, went with the trial's outcome: "Log Often, Lose More". In the authors' analysis, the days with a record went with the outcome; the minutes spent did not, which is the sentence every page about food logging should start from.

Two cautions the trial itself carries. It measured an intervention with an electronic diary, not every tracker on the market, and the authors read the fall to 14.6 as experienced loggers getting fast; it does not report how many of the 142 had stopped by week ten. Another pair of trials answers that.

Where do the minutes go?

Into the entry, item by item. A meal of five foods is five searches, five lists of entries to pick from, five portions to set, and a correction when the entry was the wrong size. The arithmetic takes no time; the searching, picking and correcting are what the journalers describe, four times a day, which is why a month of it reads as a second job. The trial timed the recording as a whole, not each part of it, so how much of the 23 minutes each part took is not known.

The people who do it say so. Journalers studied by Cordeiro, Epstein and colleagues in 2015, 141 people and 5,526 forum posts, called the practice "too much effort", "time-consuming", "tedious", and named the database as a barrier in its own right: "multiple postings with all different information", "too many choices". In a 2019 study by Chen and colleagues, 80 percent of 43 adults rated MyFitnessPal easy to use and 20 percent said they would continue; the barriers were matching foods, portion size and time.

Easy and abandoned are not a contradiction. A tool can be easy to learn and still cost a quarter of an hour a day, and in two six-month trials fewer than half were still recording at week ten. The cost is highest in the first weeks, when the foods are new to the app and the habit is thin, which is where the barriers the journalers name do their damage.

Who gets fast, and who stops before they do?

The people who stay get fast: 14.6 minutes by month six is the time of someone who knows their foods and their app. The trouble is the ten weeks before that. In two six-month trials reported by Turner-McGrievy and colleagues in 2019, fewer than half the participants were still recording after week ten, whichever definition of adherence was used.

One more figure keeps the argument honest: with instructions on picking the right entry, 50 people recording four days in MyFitnessPal came out 1.3 percent high against the Belgian composition table, in a 2020 study by Evenepoel and colleagues. The database is not the problem. Picking from it, meal after meal, is.

What does one sentence cost instead?

A description, and a question when something is missing. "Lunch was 180 g of chicken breast, no oil, 150 g of cooked rice and a 125 g plain yoghurt" contains the weights and the fat, so little is left to ask, at most whether the chicken was weighed raw or cooked and which plain yoghurt; "lunch was chicken and rice" leaves the weights and the oil open, and IZeat's protocol has the assistant ask for them, in one message, before recording anything.

That is the whole cost, and this page prints no minutes figure for it: IZeat has measured no time saving and claims none. What was observed is a scene. On 7 September 2026 I corrected my lunch four times in fifteen minutes, a weight here, the cottage cheese down to 73 g there, and nothing was retyped: each correction was a sentence, and the record followed with a trace. I build IZeat with Seb, so that is a founder's day, not a customer's, and a description of one person's afternoon, not a measure of yours.

The effort that remains is real and worth naming: connecting the assistant once, saying a weight or a fat when you know it, answering a question when you do not, and reading the receipt back. What is gone is the part the trial's minutes went into, the search, the list of entries and the portion field, which is the part the journalers in the table named as the barrier.

How can you get through the first month, when the record is most likely to stop?

IZeat is the calorie tracker inside ChatGPT, Claude or any other assistant. Tell the assistant you already use what you ate instead of searching for foods and entering them one by one. The first month is where the time was highest in the 2019 trial, and in two other trials fewer than half were still recording by week ten; that is where the design aims: a record that is easier to keep while you work toward your goal, not one that is faster to type. Whether it is kept longer is the hypothesis, not a result we have measured.

Week one, the assistant asks more, because you have not yet said your weights and your fats: "how much rice, and what was the chicken cooked in?" You answer, and the receipt comes back with the rows, their sources and the day so far. Week three, the same breakfast is "like yesterday, without the egg": the assistant reads yesterday's from IZeat and records today's with the change. A snack is a sentence in the chat that is already open. A day you forgot is said the next day, with the day, and lands where it belongs. Nothing is scored, and a day with no record is shown as one.

• You spend the first weeks saying your foods, not searching for them. For a dish you have already recorded, the assistant can read that meal from IZeat and record it again with the change, instead of asking its weights a second time; the entry is the sentence you were going to say anyway.
• You record when the meal happens, where you already are. The chat is open; the snack is a sentence; the record follows the eating rather than a trip through an app.
• You keep going after a missed day, because nothing was lost. A day with no record is a day with no record; the period's average is taken over recorded days only, and there is no series to protect.

The limit, plainly: how long you keep this record is not something IZeat has measured, and the trials above measured other tools. What the design removes is the entry screen the minutes went into; whether that is enough for you is what your own second month will say.