AI companion statistics and research in 2026
Chatbot use is mainstream. Using one specifically for companionship is not. Here is the clearest current evidence on adoption, emotional use, loneliness, and safety—without turning correlations into causal claims.
of U.S. adults report ever using an AI chatbot
Pew, 2026
of U.S. adults report chatbot use for companionship
Pew, 2026
of U.S. teens report chatbot use for emotional support or advice
Pew, 2026
participants in a four-week randomized chatbot study
MIT / OpenAI, 2025
01 · Adoption
Broad chatbot adoption is much larger than companionship use
A Pew Research Center survey of 5,119 U.S. adults found that 49% ever use an AI chatbot and 24% use one daily. But only 10% reported using chatbots for emotional support or advice, and 4% for companionship.
Among teens, Pew found 64% use chatbots, while 16% had used them for casual conversation and 12% for emotional support or advice. Those numbers matter because “uses ChatGPT” and “has an AI companion” are not interchangeable claims.
02 · Wellbeing
Short-term relief is plausible; durable benefit is still unproven
A peer-reviewed Journal of Consumer Research article reported that AI companion conversations reduced momentary loneliness in its experiments. Feeling heard helped explain the effect. The study supports a careful claim—some conversations can provide immediate relief—not the stronger claim that companions cure loneliness over time.
In a separate four-week randomized study of 981 people and more than 300,000 messages, assigned voice mode and conversation type did not produce significant overall differences in loneliness, real-world social interaction, emotional dependence, or problematic use. Participants who chose to use the chatbot more showed worse outcomes, but that association does not establish which caused which.
The practical reading is deliberately modest: an AI conversation may help someone feel better in the moment, while the long-term effects likely depend on the person, product design, intensity of use, and whether AI supplements or substitutes for human connection.
03 · Safety
Teen use, privacy, and product incentives deserve scrutiny
Common Sense Media found that roughly a third of teen users had chosen an AI companion over a person for an important or serious conversation, and about a quarter had shared personal information. Its survey also found discomfort with some companion responses. These are self-reports, but they identify concrete design and privacy questions.
In September 2025, the U.S. Federal Trade Commission opened an inquiry into seven chatbot providers. The agency requested information about monetization, data handling, safety testing, disclosures, and protections for children and teens. An inquiry is not a finding that every product caused harm; it is evidence that these questions are material enough for formal regulatory review.
AI companions should not be presented as therapists, crisis services, or substitutes for professional care. Anyone in immediate danger should contact local emergency services or a qualified crisis resource.
04 · Source table
What each source can—and cannot—support
| Primary source | Sample | Supported finding | Important limit |
|---|---|---|---|
| Pew Research Center — U.S. adults | 5,119 adults; Feb. 17–23, 2026 | 49% ever use chatbots; 24% use them daily; 10% use them for emotional support or advice; 4% for companionship. | General chatbot use, not an AI-companion-only sample. |
| Pew Research Center — U.S. teens | 1,458 teens ages 13–17; Sept.–Oct. 2025 | 64% use chatbots; 16% have used them for casual conversation; 12% for emotional support or advice. | Self-reported use; the question groups several chatbot products together. |
| Common Sense Media / NORC — teens | National U.S. teen survey; published July 2025 | 72% had used an AI companion at least once and 52% used one at least a few times a month. | Its definition includes general-purpose chatbots used socially, so it is broader than dedicated companion apps. |
| OpenAI / MIT Media Lab affective-use studies | Nearly 40 million interactions plus a four-week RCT with 981 participants | Assigned conversation type and modality produced no significant overall differences in the four measured psychosocial outcomes. Heavier voluntary use was associated with worse outcomes. | The heavier-use result is correlational and should not be read as proof that use caused those outcomes. |
| Journal of Consumer Research — loneliness experiments | Five studies; journal article published June 2025 | AI companion conversations reduced momentary loneliness in the tested settings, with feeling heard an important mechanism. | Momentary relief is not evidence of a durable reduction in baseline loneliness or a reason to replace human support. |
Methodology & disclosure
How this brief was assembled
- Cutoff: Sources available through July 15, 2026.
- Source standard: Primary surveys, study records, peer-reviewed papers, author or institutional study pages, and government notices. Vendor market forecasts were excluded.
- Claim standard: Survey findings are labeled as self-report; observational associations are not described as causal; immediate changes are not described as durable outcomes.
- Scope: The strongest population estimates here are U.S.-based. They should not be generalized globally without local evidence.
- Commercial disclosure: Ami publishes this brief and also builds an AI companion product. The source links and downloadable table are provided so readers can audit every headline claim independently.
- Corrections: Send a primary source and the disputed claim to vibe@withami.ai.