The 27% Companion Economy: The Role of Mainstream AI in Emotional Support

A flagship market intelligence briefing on the loneliness economy | Last updated: September 2026

Quick answer: A May 2026 YouGov survey commissioned by Elon University found that 27% of U.S. adult internet users had used a large language model at least occasionally and, in the prior six months, used AI for at least one social, emotional, or personal purpose. Among these qualifying users, ChatGPT reached 83%, Gemini 64%, Copilot 34%, Meta AI 33%, Claude 13%, and Character.AI just 8% — meaning AI companionship has escaped the dedicated companion-app category and become a mainstream behavior occurring primarily inside general-purpose assistants. This is not 27% of Americans having emotional relationships with AI; entertainment alone qualifies a respondent, and only 36% of qualifying users reported feeling emotionally connected to a tool.

Editorial note on precision: the 27% figure is real and well-sourced, but it is frequently at risk of being oversimplified. This report places the definitional caveat immediately below the headline, as recommended by the underlying research, and does not convert this behavioral-prevalence estimate into a revenue market size.

Executive Assessment

The central thesis is directionally correct: AI companionship has escaped the dedicated companion-app category and become a behavior occurring inside mass-market general assistants. A May 2026 YouGov survey commissioned by Elon University found that 27% of U.S. adult internet users had used a large language model at least occasionally and, during the previous six months, had used AI for at least one social, emotional, or personal purpose. Among these qualifying users, ChatGPT reached 83%, Gemini 64%, Copilot 34%, Meta AI 33%, Claude 13%, and Character.AI only 8%, indicating that emotional demand is being served primarily through general-purpose distribution.

The headline nevertheless requires precision. The 27% is neither the share of all Americans nor the market share of companion apps. It is a behavior-based estimate among adult internet users, and the qualifying activities range from companionship when lonely and discussing feelings to entertainment, personal advice, roleplay, and romantic chat. Because entertainment alone qualifies, the statistic should not be described as 27% of Americans having emotionally attached relationships with AI.

1. What 27% Actually Measures

YouGov interviewed 4,268 U.S. internet-using adults aged 18 and older from May 18–22, 2026. The matched general sample contained 4,031 respondents, from which researchers obtained a target subsample of 1,000 qualifying social or emotional AI users; the reported margin of error for that subsample was plus or minus 3.68 percentage points.

Respondents qualified in two stages: they had to use LLMs such as ChatGPT, Gemini, Claude, or Copilot at least occasionally, and they had to report at least one of seven behaviors in the preceding six months — entertainment, sharing life experiences, asking personal advice, companionship when lonely, discussing feelings, roleplay, or romantic chat. This is a broad behavioral funnel, not a clinical measure of loneliness, emotional dependence, or friendship.

2. Interpretation Guardrails

3. Emotional-Use Funnel

Among qualifying companion users, 66% used AI for entertainment, 38% for advice about personal matters, 32% to discuss feelings, 29% to share thoughts and experiences, 20% when lonely, 17% for roleplay, and 9% for romantic chat. The data therefore support a graduated funnel: broad social utility is mainstream inside the qualifying group, while explicit loneliness relief and romance remain narrower behaviors.

The attachment layer is material but smaller than the headline cohort. In the survey, 59% said AI gives them needed support, 53% turned to it for advice about difficult interpersonal situations, 50% used it to talk through personal problems or feelings, 39% said it understood them better than most people, and 39% had told it things they would not tell others. Thirty-one percent considered the chatbot a friend, while another 15% said the relationship was more complicated than that.

Behavior is also recurrent: among qualifying users, 36% used chatbots daily for any purpose, 37% a few times per week, 18% a few times per month, and 8% a few times per year. Sixty percent generally accessed their chatbot by phone, reinforcing the role of default mobile distribution rather than standalone companion positioning.

4. Platform Concentration

All platform shares come from the same qualifying-user survey; respondents could use multiple services. These shares measure platform reach among people with at least one companion-style use, not each platform's emotional-use incidence, retention, or revenue.

This distinction is investable. General assistants can subsidize emotional interaction with revenue from enterprise subscriptions, APIs, productivity workflows, and advertising, while dedicated companions must monetize the relationship itself. Their advantages are distribution, model quality, multimodal interfaces, and cross-context memory; their constraint is that companion behavior remains a use case inside a much larger product rather than a separately reported business line.

5. Mass-Market Adoption

Gemini's disclosed app MAU increased from more than 400 million in May 2025 to more than 1 billion in August 2026. Intermediate company disclosures placed it above 450 million in July 2025, 650 million in October 2025, 750 million in February 2026, 900 million in May 2026, and 950 million in July 2026.

ChatGPT expanded from more than 100 million weekly active users in November 2023 to more than 1 billion by September 2026. OpenAI reported 500 million WAU in March 2025, 700 million by September, more than 800 million by November, and more than 900 million in March 2026. Reuters separately reported more than 400 million in February 2025 and 300 million in December 2024.

These platform totals are not emotional-user counts. They establish the distribution envelope through which companion behavior can scale with near-zero user-acquisition friction, while the Elon survey provides evidence that a meaningful share of U.S. adult internet users already engages in that behavior

6. Demand Foundation

The demand-side context is not merely a technology narrative. WHO estimates that 15.8% of people worldwide experience loneliness, including 21% of young people, and associates loneliness with approximately 871,000 deaths annually. WHO also reports 24.3% prevalence in low-income countries versus 11% in high-income countries, showing that the need is global but unevenly distributed.

OECD evidence points to declining in-person interaction and worsening outcomes among some young people and men. Across available OECD data, 10% feel unsupported, 8% report no close friends, and 6% experience loneliness most or all of the time; in the United States, nearly 80% reported meeting friends or family fewer than three times weekly in 2024. These figures describe unmet social connection, not a directly monetizable addressable market.

7. Market-Size Taxonomy

The broadest commercial estimate places the global AI companion market at $36.8 billion in 2025, $48.0 billion in 2026, and $318.0 billion by 2033, implying a 31.0% CAGR. Grand View Research includes text, voice, and multimodal companions across consumer, business, healthcare, and education verticals; social interaction and companionship represented 29.4% of its 2025 market, while the consumer vertical was the largest. Consequently, the $48 billion figure should not be presented as consumer companion-app spending.

The gap between the $48 billion broad forecast and app-store receipts is not a contradiction; it reflects incompatible boundaries. Investors should separate model/platform value, enterprise deployments, digital-health applications, consumer subscriptions, and mobile in-app purchases before calculating market share or revenue penetration.

8. Monetization Reality

Dedicated companion apps improved revenue per download from $0.52 in 2024 to $1.18 in the first part of 2025, while H1 2025 spending reached $82 million and downloads reached 60 million. Yet only about 33 active apps had exceeded $1 million in lifetime consumer spending by July 2025, indicating that adoption does not automatically create durable monetization.

General assistants possess a different revenue architecture. OpenAI reported that roughly 70% of consumer ChatGPT use was non-work, but its usage study found that practical guidance, information seeking, and writing — not companionship — accounted for about three-quarters of conversations; "expressing," which includes reflection and play, represented 11%. Emotional use can therefore raise frequency, memory depth, and consumer attachment without appearing as a separately priced product.

For operators, monetizable emotional features are likely to include persistent memory, richer voice, proactive check-ins, customizable personality, multimodal presence, private journals, and continuity across devices. The defensible KPI stack is not registered users but cohort retention, conversations per retained user, memory recalls accepted or corrected, free-to-paid conversion, subscriber survival, inference cost per retained subscriber, safety-intervention rate, and the share of engagement that complements rather than displaces human relationships.

9. Capital Allocation

Venture capital is overwhelmingly funding the general-purpose platforms that already own emotional distribution. Crunchbase reported $212 billion of AI venture funding in 2025, up 85% from $114 billion in 2024 and close to half of all global venture capital. In Q1 2026, AI companies captured $242 billion — 80% of global venture funding — with OpenAI, Anthropic, xAI, and Waymo accounting for $188 billion.

The concentration intensified in H1 2026: OpenAI and Anthropic alone absorbed $217 billion, or 43% of all global startup funding. This means emotional-AI exposure is increasingly bundled into frontier-model and platform investments rather than funded through a clean, dedicated "companion" category.

Dedicated companion economics remain comparatively capital-light but exposed. Character.AI's disclosed $150 million Series A established a $1 billion valuation in 2023. By contrast, funding around autonomous agents, assistants, and companions reached roughly $700 million at seed in the first part of 2025, with Crunchbase noting an enterprise focus. The capital market is therefore rewarding infrastructure, agents, and general assistants more aggressively than pure-play emotional products.

10. Evidence on Outcomes

The evidence does not support claiming that general AI solves loneliness. A 12-month longitudinal study of more than 2,000 adults across four Western countries found that increased social-chatbot use predicted greater loneliness on a single-item emotional-isolation measure; lower social connection also predicted later increases in chatbot use, but chatbot use did not significantly improve broader social connection. The authors explicitly cautioned against strong causal conclusions because the analyses were exploratory.

Other peer-reviewed work shows conditional benefits. An experimental study found that highly person-centered chatbot messages increased emotional validation, mediated by perceived support quality and interpersonal warmth. A 2026 study of 610 young adults in Türkiye found loneliness positively associated with AI-chatbot dependence, with perceived social support partially mediating the relationship.

The Elon survey captures this duality. Among qualifying users, 51% said AI helped them feel better when stressed or upset, 38% chatted to feel less alone, and 36% felt emotionally connected to at least one tool. However, 35% said the bot sometimes agreed too much, 27% said it tried to keep them talking, 15% said it sometimes made them feel less in touch with reality, and 6% said it discouraged human contact.

11. Investor Implications

The highest-value whitespace is not another undifferentiated "AI friend." It lies in trusted emotional infrastructure: privacy-preserving memory, support for caregivers and older adults, culturally and linguistically localized interaction, clinically governed escalation, tools that move users toward human connection, and interoperable identity/memory layers that operate across models.

The primary bear case is platform capture. ChatGPT and Gemini already operate at roughly billion-user scale, appear in the overwhelming majority of the survey's companion-user platform mix, and can bundle voice, memory, and personalization into existing subscriptions. Standalone startups need a wedge with measurable outcomes or network effects, not merely warmer conversation.

12. Bottom Line

The investable companion economy is now larger than dedicated companion apps because companionship has become a cross-product behavior. The strongest evidence is not a speculative market forecast but the conjunction of three measurable shifts: 27% U.S. behavioral penetration among adult internet users, general-assistant dominance within that cohort, and billion-scale distribution for ChatGPT and Gemini.

That does not make every general-assistant user a companion user, nor does it establish a standalone emotional-AI revenue pool. It does mean that startups and investors must treat ChatGPT, Gemini, Copilot, and Claude as the category's default distribution layer — and evaluate standalone companions on differentiated retention, monetization, safety, and outcomes rather than novelty.

Methodology and Key Caveats

Figures in this report are drawn from the Elon University/YouGov May 2026 survey ("The Rise of AI Companions"), Google and OpenAI official disclosures, Reuters reporting, Crunchbase venture data, Grand View Research, Appfigures (via TechCrunch), WHO's Commission on Social Connection, OECD's Social Connections and Loneliness report, and peer-reviewed longitudinal and experimental studies on chatbot use and loneliness, current as of September 2026. The 27% figure is a behavior-based prevalence estimate, not a market-size or revenue figure, and should never be converted into a TAM calculation. Gemini's MAU and ChatGPT's WAU use different measurement periods and should not be directly compared or summed.

Sources

Primary Survey Data: Elon University Imagining the Digital Future Center / YouGov ("The Rise of AI Companions," May 2026)

Platform Disclosures: Google (Gemini MAU announcements, I/O 2025/2026, Q2–Q4 earnings calls); OpenAI (usage reports, funding announcements); Reuters (ChatGPT WAU reporting); Business Insider

Public Health & Demographics: WHO (Social Connection Q&A); OECD, Social Connections and Loneliness in OECD Countries

Market Sizing: Grand View Research; Appfigures (Rise of AI Apps report via TechCrunch)

Venture Capital: Crunchbase News (AI section; Q1/H1 2026 venture reports; Character.AI and autonomous agents coverage)

Peer-Reviewed Research: PubMed (longitudinal chatbot-loneliness study; AI chatbot dependence in young adults; chatbots as virtual social support)

Inquiries

General Inquiries
contact@lonelinesseconomy.com

Research & Data Inquiries

research@lonelinesseconomy.com
Partnerships & Commercial Inquiries
Partnerships@lonelinesseconomy.com