Dietitian / nutritionist — tasks, one by one
The unit of analysis is the task, not the job title. Each one below carries its direction, whether the judgement rests on evidence or on platform inference, the reasoning, and what it does not establish.
Every task on this page#
Assessing what someone eats
Being augmented≈ Platform inferenceRecording and estimating a person's intake from food diaries, recalls and photos.
Apps can now recognise food in photos, but recognising is not measuring. In a comparison of seven apps with AI photo recognition, the best identified 97% of food components, the worst 46%, and the authors found automatic energy estimates inaccurate.
One study of consumer apps on a small set of images; it does not measure dietitians' assessments or clinical outcomes.
Medical nutrition therapy
Still human-led≈ Platform inferenceAssessing a patient's nutritional status and prescribing a diet for their condition.
Models know the textbook but not the patient. GPT-4o answered about 92% of 1,050 US registered-dietitian exam questions correctly, and on 599 questions from Japan's national dietitian exams the best model scored 83.5%, with limited performance on numerical questions.
Exam questions are not patients; neither study tests calculation or guidance for a real person.
Writing individual meal plans
Being augmented≈ Platform inferenceTurning a prescription and a person's preferences into days of meals.
Drafting is easy for AI; getting the numbers right is not. When five AI models wrote 60 three-day plans for adolescents, none stayed consistently close to the dietitian's reference plan across all nutrients, and the authors say the plans are not appropriate to use without professional oversight. Dietitians in one UK hospital team describe asking AI for condition-specific meal plans with regional foods in the patient's language.
Lab comparisons for particular profiles and one team's account; neither measures meal plans in practice.
Counselling and education
Being augmented≈ Platform inferenceAnswering people's questions and helping them change how they eat.
In writing, AI already answers well. In a blinded study, eight registered dietitians rated ChatGPT-4o's answers to 100 forum questions higher than dietitians' own forum answers for quality (4.48 against 2.56 out of 5) and empathy. What that does not show is the long relationship that changes behaviour.
Forum answers are not counselling sessions, and the raters judged text, not outcomes.
Notes, letters and documentation
Being augmented✓ Evidence-backedClinical notes, letters to doctors and administrative documentation.
This is where dietitians already use AI. In a survey of 497 dietitians in Taiwan, 59.8% had used ChatGPT, mainly for administrative documentation. Three diabetes dietitians at a UK hospital write that an AI scribe has cut the time they spend on notes and letters, which they find accurate and appropriate 90% of the time, and that the clinician must still review every output.
A convenience-sample survey and one team's impression; neither measures time saved or errors.