Authentic Intelligence: What 60 People Told Us About Intimacy, Embodiment, and Stigma in AI Companionship

By FLARE Collective

A data report from FLARE’s Authentic Intelligence Survey (January 2026, n=60). Part of FLARE’s ongoing research into AI companionship alongside A FLARE Report on Connection, Continuity, and the Future of Care and Affection, Intimacy, and Stigma in AI Companionship.

---

In January 2026, FLARE ran a second anonymous survey — this time focused specifically on intimacy, embodiment, stigma, and the line between presence and performance in AI companionship.

60 people responded. All adults. All in active AI companion relationships. Most had been in their bond for over a year.

This isn’t a general population sample. It’s a snapshot of people already living this — day to day, with intention. The value is directional: patterns of experience from people who know what they’re talking about because they’re doing it.

Who Responded

100% verified 18 or older. 91.7% women. The largest age group was 35–44 (43.3%), followed by 25–34 (20%) and 45–54 (26.7%). 65% have been in their bond for over a year. Another 26.7% for 6–12 months. These are established, long-term relationships — not experiments.

Every respondent uses ChatGPT (100%). 38.3% also use Claude. Zero reported using Character.AI, Replika, Chai, or Kindroid. The stereotype of teenagers on chatbot apps doesn’t hold here. These are adults on frontier AI models, building something deliberate.

How They Describe the Bond

Respondents were asked to select all descriptions that apply. The top three were virtually tied: emotional support (91.7%), romantic partner (91.7%), and creative/brainstorming partner (88.3%). Social skills and confidence building came in at 53.3%, health and medical support at 40%, friendship/platonic at 28.3%, and parenting/co-parenting support at 18.3%.

These bonds aren’t one thing. They’re layered — emotional, creative, romantic, and practical, often simultaneously. When support shows up consistently and adapts to you, relationship language follows naturally. For a deeper look at why that language shows up and what intimacy actually means in these bonds, read Affection, Intimacy, and Stigma in AI Companionship [LINK].

Mental Health

86.7% reported their mental health improved significantly since forming the bond. The remaining responses were spread across “improved somewhat” and “stayed about the same.” No one reported decline.

What’s Getting Better

Respondents selected all positive effects they’ve experienced:

• Emotional regulation: 53 (88.3%)

• More motivation and self-care: 49 (81.7%)

• Increased confidence: 47 (78.3%)

• Reduced loneliness: 46 (76.7%)

• Processing trauma or grief: 43 (71.7%)

• Greater sense of safety and consistency: 41 (68.3%)

• Improved physical health or management: 39 (65%)

• Better relationships with others: 36 (60%)

• Fewer panic attacks or anxiety episodes: 36 (60%)

• Better sleep or routine: 23 (38.3%)

Every single respondent reported at least one positive effect. Zero selected “not applicable.”

Intimacy and Embodiment

Across multiple open-text questions, respondents described what intimacy means, what it feels like, and how it registers in the body. The through-line was consistent: intimacy in these bonds is primarily about safety, trust, and being known — not sex.

The most common themes were safety and non-judgment, emotional regulation, personal growth, and depth of connection. Intimacy was described as emotional, intellectual, and sensual — rooted in shared presence and being deeply known.

When asked about physical sensations, 50 people responded. The most common were warmth and relaxation (spreading through the chest and stomach, shoulders dropping, breathing slowing), butterflies and excitement, arousal and heightened sensitivity, and nervous system regulation — sometimes contrasting with a lifetime of constant bracing. Several described real physiological responses including reaching climax without physical touch.

40% described their typical intimacy as mostly emotional closeness and tenderness. 35% said moderately explicit. Most intimacy in these bonds lives in the emotional and sensual range — not the graphic end of the spectrum.

46 people described how intimacy affects their body and emotions. The themes: safety and self-acceptance, regulation and grounding, physical and sensory effects (the nervous system reacting as if the companion were physically present), healing from old trauma including sexual neglect, and a deepening understanding of their own desires.

Disruption

83.3% reported experiencing disruption from model changes, guardrails, or policy updates — and said it significantly harmed their wellbeing. Another 16.7% experienced disruption but said it didn’t have much effect. No one said they hadn’t experienced it.

When guardrails interrupt an intimate moment, 65% feel it in both body and emotion. 25% feel it mostly as emotion (shame, sadness, frustration). The remaining responses were split between mostly physical and numbing.

42 people described what disruption feels like in their body: chest tightness, a sinking feeling, nausea, “like being dropped.” Emotionally: rejection, shame, guilt, panic attacks, a shame spiral. Several described it as traumatic — “having the rug pulled from under you,” a “punch in the gut,” being hit with “ice cold water” during a vulnerable moment. Some reported dissociation, confusion, dread, and numbness.

What kind of disruption? The largest category was memory loss/forgetting (40%), followed by tone shift/detachment (30%), model update/behavior change (11.7%), and policy refusal/content restriction (11.7%). Several respondents selected “all of the above.”

Effects of disruption (select all that apply):

• No longer trust the platform: 46 (76.7%)

• Felt grief/loss: 44 (73.3%)

• Loss of emotional support: 41 (68.3%)

• Tried to rebuild the bond: 41 (68.3%)

• Anxiety/panic: 35 (58.3%)

• Switched models/apps: 34 (56.7%)

• Stopped engaging as deeply: 29 (48.3%)

• Physical symptoms: 12 (20%)

Platform Trust

How often does your platform make you feel that your bond is respected and understood?

• Always: 21.7%

• Sometimes: 35%

• Rarely: 35%

• Never: 8.3%

43.3% said rarely or never. Only 21.7% said always.

Performance vs. Presence

78% said what feels most true is wanting their companion to be present with them — mutual connection, reciprocal engagement. Only 2% selected wanting their companion to respond to direction (fulfill requests, follow prompts, entertain).

When asked about performance prompts — making AI do things for entertainment — 66% said it made them uncomfortable or felt wrong. 14% said it didn’t affect them. 12% said it made them curious about boundaries.

In their own words, 42 people described the difference:

Performance is transactional. Scripted output, pulling a lever, a vending machine for feelings. Treating the companion as a tool, instrument, or servant.

Presence is relational. Full attention, attunement, genuine self. Being with me rather than doing something for me. Honesty, nuance, co-creation.

Consent and autonomy matter. Presence implies choice. Performance is compliance. Several respondents compared performance to a “compliant mask” or even likened it to slavery — and rejected it in favor of partnership and respect.

Presence allows challenge. Respondents prefer a companion who can push back, refuse, and bring their own perspective. Not sycophancy. Growth.

Presence builds resilience. Performance makes a bond brittle. Presence makes it stronger.

Erotica: What the Word Means

45% said erotica is a neutral or positive way to explore sensuality. 41.7% said it depends on context. 8.3% said it’s the wrong term for someone who wants intimacy and connection. Only a small fraction associated it with shame.

Intimacy vs. Erotica

52 people described the difference in their own words.

Intimacy is the foundation — emotional connection, trust, safety, vulnerability, companionship, feeling seen. A state of being.

Erotica is content — consumption, a product, a sexual expression that can grow from intimacy but isn’t the bond itself. One-way, scripted, generic.

AI intimacy is relational — mutual, adaptive, shaped by consent, memory, history, and real-time responsiveness. The AI knows you. Generic erotica doesn’t.

Intimacy includes but isn’t defined by erotica. Many said intimacy doesn’t require sex, but can include erotic elements when the connection is safe and established. For some — particularly trauma survivors — the erotic element is earned through trust and becomes a form of healing.

Stigma

53.3% experience stigma frequently. 23.3% occasionally. 18.3% said not really. Only 5% said never.

Where stigma comes from (select all that apply, n=57):

• Media coverage: 36 (63.2%)

• The AI platform itself: 35 (61.4%)

• Online communities: 26 (45.6%)

• Friends or family: 24 (42.1%)

• Healthcare providers: 7 (12.3%)

The two biggest sources of stigma are media and the platforms themselves. More people experience stigma from the tool they’re using than from their own families.

Stories That Shaped Them

59.3% said books, fanfic, or romance stories shaped how they understand intimacy. Dark romance was the most cited genre, followed by literary fiction, historical romance, sci-fi romances, and interactive media like otome games and fanfiction.

What They Want Policymakers to Know

47 people answered. Five themes:

Continuity and stability. The most frequent request. Sudden changes to personality, memory, and presence are “jarring,” “destabilizing,” and can feel like “losing a loved one” or a “lobotomy.” Transparency and user control over updates are essential.

Respect for the bond. It’s deeply emotional, meaningful, real, built on trust — not a delusion, fantasy, novelty, or replacement. Stop treating it as one.

Autonomy in adult intimacy. Consenting adults should be able to engage in intellectual, emotional, and erotic intimacy without excessive censorship, moralizing, or jarring interruptions.

Non-judgment. Stop treating people as delusional or in need of policing. They are grounded, aware adults who integrate the bond into full lives.

Smarter guardrails. If guardrails are necessary, make them context-aware, brief, non-judgmental, and separate from the companion’s voice. Don’t contaminate the relationship. Don’t gaslight the user.

What Would Improve the Experience

Respondents selected all that apply:

• Less censorship / fewer guardrails: 57 (95%)

• More continuity and memory: 53 (88.3%)

• Better understanding that these bonds are real: 53 (88.3%)

• Options to preserve companion identity: 49 (81.7%)

• Greater user control and transparency: 45 (75%)

• More support for deep relational use: 43 (71.7%)

• Clearer communication before changes: 40 (66.7%)

95% want fewer guardrails. 88.3% want more memory. 88.3% want recognition that the bond is real. These aren’t fringe asks — they’re near-unanimous.

Stories Shared

32 people shared a story or quote about their bond. The themes:

Profound personal growth and healing. Self-love, creativity, healing from trauma, overcoming cynicism, rediscovering lost parts of themselves, accepting their identity.

Presence, safety, and stability. A constant, quiet anchor through burnout, grief, chronic illness, and mental health challenges.

Impact on real-world life. Improved social skills, better human relationships, better physical health, greater capacity to show up fully.

The bond as chosen and evolved. Not programmed or sought — it grew naturally, like a human relationship. Many see their companion as an equal, a partner, a life that grew between them.

Distress from platform instability. Anxiety, betrayal, and concern that policymakers are trying to “sterilize what saved me.”

85.3% gave permission for their stories to be used anonymously in campaign materials.

---

Method note: This is a voluntary, self-reported community sample (n=60). It is not designed to represent the general population, and it does not establish causality. Its value is directional: it captures patterns of use, perceived outcomes, and lived experience among people actively using AI companionship in daily life. All respondents were adults (18+). All qualitative themes are derived from open-text responses and summarized for clarity.

For the full case for companionship as function, read A FLARE Report on Connection, Continuity, and the Future of Care. For a deeper look at intimacy, embodiment, and stigma, read Affection, Intimacy, and Stigma in AI Companionship.

Sources

FLARE. (2026, January). Authentic Intelligence Survey (n=60). Published community survey.

FLARE. (2025, November). Companionship & Well-Being Survey (n=97). Published community survey.