Ten percent of Americans consider an AI a close friend

Last updated
Ten percent of Americans, mostly young people, consider an AI a close friend.

A related friendship measure has reached 10%, but wording, population and observation window differ from this prediction.

At a glance

  • Assessment: Emerging
  • Confidence in assessment: 80%
  • Outcome: unresolved
  • Timing: pending
  • Evidence: proxy
  • Predicted timing: July 2027
  • Primary source: July 2027: The Cheap Remote Worker; AI 2027 archived scenario

What AI 2027 Predicted

Ten percent of Americans, mostly young people, consider an AI a close friend. The source is July 2027: The Cheap Remote Worker in the AI 2027 scenario.

Archived scenario version and section-page references. The linked live scenario may change; these definitions use the archived version identified here.

How We Track This

Primary operational population: US adults, explicitly labeled as a limitation relative to “Americans.” Require a representative weighted survey of at least 1,000 adults using the exact close-friend concept, fielded within July 2027, and independent corroboration with comparable wording within 60 days. Threshold: weighted point estimate at least 10%; publish uncertainty and age breakdown. Ask directly whether the respondent considers any AI a close friend.

These operational thresholds specify this tracker’s test. Where the scenario uses qualitative language, they are interpretations rather than numbers quoted from the original.

Current Evidence

YouGov’s July 1 report found that 10% of surveyed US adult citizens described an ongoing personal friendship with an AI chatbot. Fieldwork was June 26-29, 2026, with 1,126 respondents. This is useful precursor evidence; it is not the same as the close-friend question for the July 2027 window.

Counterevidence & Limitations

Adult-only measurement cannot directly establish the all-ages proportion. Chatbot usage, emotional attachment or companionship are not synonymous with close friendship. Report sampling and nonresponse uncertainty; do not infer causality or wellbeing effects.

Missing comparable survey evidence leaves the result unresolved.

What Would Change Our Assessment

Representative comparable surveys remain below 10%, or the result depends on leading wording or a selected AI-user sample.

Evidence satisfying the full test would change the assessment; earlier observations inform the trajectory but do not resolve this test’s specified observation window. Partial proxies are recorded with their scope.

Update History

DateUpdate
2026-09-06A related friendship measure has reached 10%, but wording, population and observation window differ from this prediction. Relevant primary evidence and limitations are described above.