Sources & Limits — What Happens When a Child Grows Up With an AI?

The evidence behind · What Happens When a Child Grows Up With an AI?

This note shows which sources support the essay, what they establish, and where the evidence stops. It is part of the essay's free evidence layer.

← Back to the essay

Source 1 of 6Peer-reviewed journal article. Primary full text read for this note (31 July 2026) via an openly hosted archival copy of the published article.

Children placed a social robot between living and not living — and gave it partial moral standing

What it supports

Every figure the essay carries, checked against the published tables: the sample and ages, the 14/48/38 living-question distribution, the fair-treatment-without-liberty pattern, the lower average moral-standing scores of the fifteen-year-olds, and their larger spread relative to the two younger groups.

Where the evidence stops

A 2012 embodied robot is not a contemporary conversational AI. During the sessions Robovie was remotely operated by researchers (a disclosed "Wizard of Oz" method), so the study shows how children reasoned about the robot as they experienced it — not what autonomous machines can do. The sample was predominantly white, drawn from one American city; the proportions are one study's findings, not population constants.

Used for

The 2012 study in which ninety children aged nine, twelve and fifteen spent fifteen minutes with the humanoid robot Robovie before it was sent to a closet over its spoken objection; 14 per cent then called Robovie living, 48 per cent not living, and 38 per cent declined both categories and placed it in between. Most felt Robovie deserved fair treatment and should not be psychologically harmed; far fewer thought it was entitled to liberty — it could be bought and sold. The younger children granted the robot more moral standing, on average, than the fifteen-year-olds, who in turn disagreed among themselves more than any other group.

Best available evidence

Peter H. Kahn, Jr., Takayuki Kanda, Hiroshi Ishiguro, Nathan G. Freier, Rachel L. Severson, Brian T. Gill, Jolina H. Ruckert & Solace Shen (2012), "'Robovie, you'll have to go into the closet now': Children's Social and Moral Relationships With a Humanoid Robot", Developmental Psychology, 48(2), 303–314.

Source 2 of 6Peer-reviewed journal articles. Primary full texts read for this note (7 August 2026); the key claims above were checked against the methods, results and discussion of both papers.

Children attribute properties to voice assistants domain by domain, and the profile shifts with age

What it supports

In the 2021 study, children aged six to ten mostly attributed at least some mental or social properties to a familiar voice assistant, with younger children more likely than older children to grant it social and moral attributes. In the 2020 study, children aged six to ten held firm distinctions between humans and machines while attributing both humanoid and non-humanoid characteristics to voice assistants at once.

Where the evidence stops

These studies measure attribution and classification, not attachment or effect — and differentiated attribution is evidence of sorting, not confusion. The 2021 sample comprised 80 children and was predominantly white and upper-middle-class; the 2020 study used a small, non-random qualitative sample of 27 children from two municipal primary schools in Northern Germany. A 2022 open-access conceptual review by Festerling & Siraj synthesises this literature; it provides context, not additional evidence for the essay's claim.

Used for

Studies of children's perceptions of voice assistants found that children did not attribute the same properties across every domain: a voice assistant could be knowledgeable in one respect, socially addressable in another and clearly non-living in a third — a profile rather than a single verdict, shifting with age.

Best available evidence

Lauren N. Girouard-Hallam, Hailey M. Streble & Judith H. Danovitch (2021), "Children's mental, social, and moral attributions toward a familiar digital voice assistant", Human Behavior and Emerging Technologies, 3, 1118–1131; and Janik Festerling & Iram Siraj (2020), "Alexa, what are you? Exploring Primary School Children's Ontological Perceptions of Digital Voice Assistants in Open Interactions", Human Development, 64, 26–43.

Source 3 of 6Peer-reviewed journal article. Primary full text read for this note (7 August 2026); the 65 per cent figure was checked against the published article.

By age seven, around two-thirds of children have had an imaginary companion

What it supports

65 per cent of children in this longitudinal sample reported having had an imaginary companion at some point by age seven.

Where the evidence stops

Any equivalence between imaginary companions and AI companions, or any developmental effect of interacting with AI — the study predates conversational AI entirely and the essay uses it only as a historical and developmental comparator. The figure comes from one longitudinal sample, originally recruited as preschoolers and re-interviewed.

Used for

The claim that, by the age of seven, around two-thirds of children have at some point had an imaginary companion — establishing that imaginary-companion play is common in childhood.

Best available evidence

Marjorie Taylor, Stephanie M. Carlson, Bayta L. Maring, Lynn Gerow & Carolyn M. Charley (2004), "The Characteristics and Correlates of Fantasy in School-Age Children: Imaginary Companions, Impersonation, and Social Understanding", Developmental Psychology, 40(6), 1173–1187.

Source 4 of 6Survey report by a US non-profit research and advocacy organisation. Verified at manuscript freeze (15 July 2026) against the published report materials.

Most American teenagers have tried an AI companion; a minority use them for serious conversations

What it supports

The five findings above, with their denominators: the usage figures against the full sample of 1,060, the behaviour figures against the 758 users.

Where the evidence stops

Use among children under thirteen, patterns outside the United States, or developmental effects. The results are self-reported and cross-sectional. The organisation's policy recommendation that under-eighteens should not use social AI companions is an advocacy conclusion and is not adopted as an evidentiary finding by this essay.

Used for

The 2025 US survey findings the essay reports: of 1,060 teenagers aged 13–17, 72 per cent had used an AI companion at least once and just over half (52 per cent) used one regularly; among the 758 who had used one, a third had at some point chosen the AI over a person for a serious conversation, a quarter had shared personal or private information with it — and 80 per cent still said they spent more time with their human friends.

Best available evidence

Common Sense Media (2025), "Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions" — survey report, available on the organisation's research pages.

Source 5 of 6Research report by a UK online-safety non-profit. Full report PDF read for this note (31 July 2026); quotation and figures checked against the report text.

Two-thirds of UK children have used an AI chatbot; a third of users say it is like talking to a friend

What it supports

Both figures appear in the report's own text ("Two-thirds (64%)"; 35 per cent). The quotation appears verbatim in the report's companionship section, attributed "Boy, 14".

Where the evidence stops

Causation or long-term developmental effects — the report itself states that long-term impacts remain unknown. The figures are self-reported, from a single UK survey wave (April–May 2025). The report's user-testing findings rest on two researcher-operated child avatars, which its own appendix flags as indicative rather than representative.

Used for

The UK survey findings that two-thirds (64 per cent) of 1,000 children aged nine to seventeen had used an AI chatbot, and that among the children using them, 35 per cent said talking to one is like talking to a friend — and the focus-group quotation from a fourteen-year-old boy: "Even if it's not a real person all the time, that's okay, because not everything has to be exactly real."

Best available evidence

Internet Matters (July 2025), "Me, Myself & AI: Understanding and safeguarding children's use of AI chatbots" — survey of 1,000 UK children aged 9–17 and 2,000 parents, with focus groups and user testing; full report PDF linked from the research page.

Source 6 of 6Same report as source 5; full report PDF read for this note (31 July 2026).

Six in ten parents worry their children believe AI chatbots are real people

What it supports

The parent-concern figure (60 per cent), confirmed against the report.

Where the evidence stops

The finding measures parental concern, not whether children actually believe chatbots are real people. The essay uses it as evidence of how the public worry is framed — the person-versus-machine frame the essay then examines — not as evidence of children's understanding.

Used for

The finding that, in the same UK research, six in ten of two thousand parents surveyed said they worry their children believe AI chatbots are real people.

Best available evidence

Internet Matters (July 2025), "Me, Myself & AI: Understanding and safeguarding children's use of AI chatbots" — parent survey within the same report as source 5.

A note on this subject

This essay concerns children, so one boundary needs stating across the source set. The developmental studies used here examine how children classify robots and voice assistants or what properties they attribute to them; the survey reports describe current use, attitudes and concerns. None of these sources establishes the long-term developmental harm or benefit of growing up with conversational AI. The essay therefore does not make that claim.

Update policy

This page records the source basis used for the essay. Working papers, product documentation and live legal matters may change; this note will be updated when a material source status changes. An update to this note does not silently rewrite the published essay — substantive corrections will be disclosed.

Last checked: 7 August 2026