How much typing does an AI food photo save if it still needs the ingredients?

Taking a lunch photo sounds easier than entering every ingredient. If the assistant needs the recipe and amounts afterward, I would want to know how much of that job it actually removed. A November 2025 study in Nutrients tested ChatGPT-5 on 195 dishes drawn from recipe images, SNAPMe and dietitian-prepared meals. Estimates improved as the researchers added nonvisual details and ingredient quantities. That is evidence about those inputs and that model, not a test of today's calorie apps. The reference values also came from different sources; these were not 195 meals all weighed under one protocol. For someone already keeping a food log, I would compare photo entry with their usual method on a few familiar meals. Include the time spent correcting the food name, portion and ingredients. That is a comparison I would try, not an outcome measured by the paper. A photo might be useful as a reminder even when its calorie estimate is too uncertain to use. I would also want an entry to show which amounts I supplied and which it guessed. A precise-looking total should not make me forget that the cooking oil was never specified. This is about the effort and uncertainty of logging, not advice to change what you eat. Would you rather correct a photo estimate, or save a meal you eat regularly and reuse that entry?