Screen guidance and AI guidance were written for different ages

The classroom note

A toddler who already has an opinion about a smart speaker's answers is not a hypothetical for guidance bodies; it is the situation their recommendations were written to anticipate before generative AI products existed to fill it. Two long-standing recommendations bound the early years case: the World Health Organization's 2019 guideline sets a maximum recommended amount of screen-based sedentary time for children under five, and the American Academy of Pediatrics' media guidance asks families to build an individual Family Media Plan rather than follow one fixed number. A conversational AI toy or app marketed at this age group has to be judged against both, and against a newer, AI-specific recommendation neither anticipated.

What the evidence says

The WHO guideline is a formal, peer-consensus document, developed, in its own words, using ‘the best available evidence, expert consensus and consideration of values and preferences, acceptability, feasibility, equity and resource implications,’ that caps sedentary screen time for children under five within a day that also specifies activity and sleep. It does not mention artificial intelligence or voice assistants; it addresses screen-based sedentary behaviour as a category. The AAP's page is a living clinical resource rather than a dated study: it reports children now average seven hours a day on entertainment media and recommends a plan revised ‘when necessary’ rather than a universal cutoff. Separately, UNESCO's 2023 guidance for generative AI in education proposes setting an age limit for independent conversations with generative AI platforms, the first of these three to name AI directly.

The implementation question

The practical gap is that none of the three sources was designed to answer the same question together: WHO caps sedentary screen time by age band without reference to what is on the screen; AAP asks families to individualise a plan without giving AI products a category of their own; UNESCO's proposed age floor for independent AI conversation is aimed chiefly at school-age use, not toddlers. A product built to talk back to a three-year-old sits at the intersection of all three without being squarely inside any of them.

What holds and what fails

What holds across all three is the underlying premise that early childhood is a period requiring deliberate limits set by adults, not a product's own settings. What fails is any claim, from a product or a review, that an AI companion for a young child has cleared paediatric or education-guidance scrutiny; none of these documents evaluates that category, so such a claim would go beyond what any of them establish, an editorial caution worth stating outright.

• Does this product fall inside an existing screen-time plan, or has no one decided that yet?
• Is a young child using it independently, or always alongside an adult who can redirect it?
• What is the product actually asking a very young child to say back to it?

Guidance written for screens in general and guidance written for AI in classrooms both existed before AI products aimed at toddlers did; a family filling that gap is, for now, doing so without a body of evidence built for the specific case.