A shopper can ask AI which drink offers better value, compare two snacks, or find a meal that fits their budget.
“What should I eat?” is a broad question to put to AI. Add a budget, a few ingredients, and a goal, and the answer can become quite specific. That is part of its appeal, particularly when someone wants a meal idea or help comparing products without spending half an hour searching.
A MyFitnessPal-commissioned survey of 1,011 Australians, conducted in March 2026, found that 44 percent had used AI for coaching or guidance. The company released the findings while promoting its AI Coach feature. The survey gives an indication of reported behaviour, but it did not establish how accurate or useful the advice had been.
Some food questions lend themselves to straightforward help. A tool working from current nutrition information can compare amounts per serving or organise a recipe around ingredients already available. It can also explain what a label means in everyday language.
Other questions need more thought. The drink with more protein is not automatically the most suitable choice for the person asking. We would need to know what they want from it, how much they intend to consume, and what else they eat.
The ingredients themselves can encourage assumptions. UNSW Business School consumer psychologist Professor Nitika Garg discussed this after the Greek yoghurt and Biscoff biscuit “Japanese cheesecake” trend attracted attention in Australia. Viewers could associate Greek yoghurt with good nutrition and carry that judgement over to the whole dessert.
“People don’t always do the research,” said Garg.
Her commentary concerned social media influence, rather than a study of AI responses. It nevertheless raises a useful question for anyone asking AI to assess a recipe. Will the answer examine the ingredients and quantities, or repeat the description already attached to it online?
Food Standards Australia New Zealand distinguishes nutrition content claims from health claims. The first describes a nutrient or substance in a food and must meet relevant criteria. The second describes a relationship between food and a health effect and requires scientific support.
Those distinctions can be lost in a casual explanation. Finding an ingredient in a beverage does not establish that a serving delivers a particular benefit. The amount may matter, as may the conditions under which the claim is supported. An answer that omits those details can leave a shopper with a different understanding of the product.
For manufacturers and retailers, there is a practical reason to keep product information clear and current. If a shopper asks AI to compare products, accurate serving sizes, nutrition panels, and properly qualified claims provide a better starting point. Businesses using AI to draft their own explanations have the opportunity to check every claim before publishing.
An external tool’s answer is harder to control. A brand may have made a careful statement about its product, only for a summary elsewhere to present it as having a broader benefit. That makes it useful to understand what information is publicly available and whether it consistently describes the product.
AI-generated endorsements add another complication. The ASA’s guidance cautions against creating fake testimonials or experiences and makes clear that advertisers remain responsible for their advertising. A convincing presenter cannot supply evidence of a benefit simply by describing it.
There is much to like about getting a useful answer while standing in a supermarket aisle or planning dinner. The value is in the detail it helps us understand. If a food or drink is recommended for a health benefit, we should still be able to find what supports that recommendation before paying more for it.
