Dietitian / nutritionist

United States. The statistics bureau projects employment of dietitians and nutritionists to grow 8 percent from 2025 to 2035, from 86,300 to 93,000, with about 5,700 openings a year, citing growing interest in the role of food and nutrition in wellness and preventive care. The page does not mention artificial intelligence. It counts jobs for one country.

A named person with standing publicly predicted something, on a date, in an attributable statement. It is recorded so that who said what, and when, stays checkable — and it never moves a task's assessment, because a prediction is not an observation. Its value arrives later: the record sits on the same page as the evidence about that occupation, so anyone reading the forecast reads the record of what happened next beside it. That is the reckoning; this site publishes no verdict on whether a forecast came true.

Japan. A food company's announcement that, after a trial with five food-service companies, it began selling on 11 May 2026 an AI service that automatically generates menus meeting conditions set by registered dietitians, from one to three meals a day for up to several months; standard diets only, for care homes, kindergartens and nurseries, at 5,467 yen per facility per month. It cites one pilot partner's 70% estimate. It is the company describing its own product.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

A cross-sectional study in which five AI models (ChatGPT-4o, Gemini 2.5 Pro, Claude 4.1, Bing Chat and Perplexity) generated 60 three-day diet plans for four standardized adolescent profiles, compared with a dietitian's reference plan for each. Micronutrients varied significantly between models and no model showed consistent proximity to the dietitian across all nutrients; the authors conclude AI-based dietary recommendations are not appropriate to use without professional oversight. Standardized profiles, not patients.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

A cross-sectional study in which 100 nutrition questions were taken from public online forums where registered dietitians had answered, each paired with a ChatGPT-4o answer, and eight licensed dietitians, blinded to the source, rated them. AI answers scored higher for quality (4.48 against 2.56 on a 5-point scale), empathy (4.62 against 3.21) and overall performance; dietitians' answers were more variable. Forum answers are not counselling sessions.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

Japan, one contract food-service operator that runs care-home, company and school catering. It says it took part in a trial of an AI menu planner in real food-service menus, that menu planning had been concentrated on a few staff, that the tool can generate menus meeting set conditions for up to several months, that a registered dietitian (管理栄養士) checks the menus, and that by its own estimate about 70% of menu-planning work could be cut. An operator's estimate from a pilot, not a measured result; the operator was a partner of the vendor.

Small-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.

Japan. Article 21(1) of the Health Promotion Act requires the operator of a specified feeding facility that a prefectural governor designates as needing special nutrition management to place a registered dietitian (管理栄養士) there; article 21(2) asks other specified feeding facilities only to make efforts to place a dietitian or registered dietitian. It names no technology; it requires a person in a role, not a way of planning menus.

Regulation, subsidy or public procurement is requiring or funding adoption — the mirror of a constraint. It shows adoption is being required, not that it has happened, so one mandate is never enough on its own; two independent ones are.

Japan. A study that tested GPT-4o, GPT-4o-mini and GPT-4o with retrieval on 599 publicly available multiple-choice questions from the 2022–2024 national examinations. All passed the 60% threshold; GPT-4o-RAG scored highest at 83.5%, did better in applied and clinical nutrition, but showed limited performance on numerical questions, and the authors say limitations in numerical reasoning and individualized guidance warrant further development. Exam questions, not patients.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

Taiwan. A cross-sectional survey of 497 dietitians found 59.8% had used ChatGPT, primarily for administrative documentation; younger dietitians with higher educational levels and hospital-based roles were more likely to use it. Self-reported use in one place; it does not measure time saved or accuracy.

Measured, large-scale use of a tool for real work, where the decision to use it was the worker's rather than an employer's. It is more than a capability record — the work is real, not a demo — and less than a deployment record, because no employer put it into production, required it, or built a process around it. Weighted `cautious`: `automating` means the machine can do the task AND there are adoption signs, and this is an adoption sign — but usage can be experimental, and much of the measurement comes from a party with a stake, so one record is never enough and two independent ones are. Note who is counting. Vendor telemetry sees this directly and sells the tool, so such a record names that stake in its scope; a statistics agency asking firms whether their workers use AI in tasks sees the same channel with no stake at all, and that is the better source where it exists.

England, one hospital team's first-hand account on the professional body's website. Three diabetes specialist dietitians say an AI scribe has reduced time spent on notes and letters, that 90% of the time the notes and letters are accurate and appropriate, and that the clinician must still review every output; they also describe asking AI for condition-specific meal plans with regional foods in the patient's language. An impression without a method, from one team.

Measured, large-scale use of a tool for real work, where the decision to use it was the worker's rather than an employer's. It is more than a capability record — the work is real, not a demo — and less than a deployment record, because no employer put it into production, required it, or built a process around it. Weighted `cautious`: `automating` means the machine can do the task AND there are adoption signs, and this is an adoption sign — but usage can be experimental, and much of the measurement comes from a party with a stake, so one record is never enough and two independent ones are. Note who is counting. Vendor telemetry sees this directly and sells the tool, so such a record names that stake in its scope; a statistics agency asking firms whether their workers use AI in tasks sees the same channel with no stake at all, and that is the better source where it exists.

A study that tested GPT-4o, Claude 3.5 Sonnet and Gemini 1.5 Pro on 1,050 US registered dietitian exam questions from an exam-preparation bank. GPT-4o scored 91.92% with zero-shot prompting and highest overall; the authors note results varied considerably with different prompts and question domains. Exam questions test knowledge, not the care of a patient.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

A study of nutrition apps, seven of which offered AI food-image recognition, tested on food images including mixed dishes. The most accurate identified 97% of food components (38 of 39) and the least accurate 46% (18 of 39); the authors found automatic energy estimations from AI food-image recognition inaccurate and say collaborating with dietitians is essential to improve the apps. Consumer apps on a small image set.

A demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.

If you are starting out, expect AI to draft your meal plans, education sheets and notes, and build what it gets wrong: the numbers for a specific patient, the diet for a condition, and a relationship that changes what someone eats.

Expect menu drafting and paperwork to speed up, especially in institutional food service. Medical nutrition therapy and the check on every AI-drafted plan stay with you.

Reshape the role

Check AI-drafted plans instead of writing them from scratch

Studies show AI meal plans deviate on nutrients, and menu tools are sold on the basis that a dietitian reviews their output.

Stay and strengthen

Stay in clinical nutrition where a licence is required

Japan requires registered dietitians in designated facilities, and models are weakest on patient-specific calculation.

Adjacent move

Move into validating nutrition software

Food-logging apps and menu planners need dietitians to check their food databases and energy estimates.

The evidence does not show that. AI passes exam question banks and writes fluent answers, but its meal plans deviate on nutrients and photo apps misjudge energy. Institutional menu drafting is being automated in Japan, where the law still requires registered dietitians in designated facilities. The US projects dietitian employment to grow 8 percent from 2025 to 2035.

We do not answer that with a number of years. Watch whether your employer buys menu-planning or note-writing AI, and what your licensing rules reserve for dietitians. Those tell you more than any date.

It can draft one, but studies found AI plans deviating from dietitians' plans on energy and nutrients, and the authors advise against using them without professional oversight — especially for children or medical diets.

At recognising foods, the best ones are; at estimating energy, a study found them inaccurate. Recognition ranged from 97% to 46% of food components across seven apps.