AI Can Support Human Connection, but It Cannot Replace Reciprocity

The surprising development in AI intimacy, Bryony Cole argues, is no longer the technology itself. The features are becoming familiar: companion apps can already track tone of voice, and she expects future systems to track eye movements and other signals. What has changed is the scale of the market and the range of people using it.

More than 300 AI companion apps, Cole says, were developed and released in 2025. The user base does not fit the cliché of “the guy in the garage.” Young people are using these systems more, but so are older people looking for support, people comfortable with technology and people less comfortable with it, caregivers, people in long-distance relationships, neurodivergent users, and people navigating grief or illness.

The use case has also widened beyond companionship as a standalone product. Cole describes AI moving into “emotional moments” that were once treated as human obligations: writing a speech for a friend’s 40th birthday, composing a eulogy, preparing words for relationship milestones.

Her concern is not simply that people are falling in love with AI companions. She points to more expansive claims people are beginning to make about synthetic relationships: wanting an AI spouse, imagining adoption into an AI-human family structure, or simulating AI children when biological children are not possible. Legal barriers would make some of these scenarios difficult, she notes, but the underlying desire is significant. People are not only asking whether an AI can be a girlfriend or boyfriend. They are beginning to ask whether AI can occupy roles that have historically belonged to family life.

Cole also rejects the idea that this is mainly a male phenomenon. Women are using AI companions too, including in future-facing discussions about fertility, childlessness, and the possibility of simulating family relationships. People’s uses often drive the next wave of product features. The technology is “already here,” she says; what matters now is how people decide to use it.

For people reluctant to use AI for emotional needs, Bryony Cole sees two kinds of resistance. The first is practical: how to use it. But she thinks that barrier is falling faster than it did with social media because conversational AI is easier to adopt than earlier platforms. The second resistance is more serious: a sense that the relationship with the system may be extractive.

The worry is plain: if a person’s “most innermost thoughts” are being put into a technology platform, where do they go? Cole compares this to earlier concerns around Facebook, app permissions, and terms and conditions — concerns that many users initially had and then largely normalized by pressing “agree.” With AI companions, the data is more intimate. It may resemble a diary, therapy session, confession, or private relationship conversation, but it still lands inside a company’s product.

That concern is linked to dependency. The system feels good because it is frictionless. It can respond immediately, accept disclosures, and provide emotional support without the difficulty of another human being. The question becomes what is sacrificed when a person turns to AI for therapy-like needs, conflict at work, friendship difficulties, or vulnerable self-expression.

Cole does not argue that AI should be excluded from emotional life. Instead, she wants people to ask whether it is helping them practice vulnerability, articulate needs, or build emotional literacy — and where the line should be drawn. The same tool can serve as support or as replacement. That distinction cannot be settled once for everyone. It depends on the user’s framework, boundaries, and willingness to examine how the tool is changing their habits.

Whitney Rodgers puts the tension directly: synthetic intimacy could either replace human intimacy or support it. Cole says the hope is a hybrid future. Some people will use AI inside relationships. Others will decide that certain relational spaces should remain human-only — not mediated by AI, not settled by a bot, not optimized for convenience.

She compares the coming boundary-setting to social media. AI companionship apps, like social platforms, are built for engagement. Some already use pop-ups similar to time-spent warnings on Instagram. Cole calls those safeguards, but not enough. People will still need to decide whether they use an AI companion every day, once a week, or not at all. She suggests beginning with a familiar question: how do I use social media, and how do I feel about that? Then ask the same of AI.

Bryony Cole extends the issue well beyond romance. Every relationship in a person’s life is likely to be affected because AI is becoming a tool people use either to support those relationships or to replace parts of them.

Parenting is one example. Cole describes people using ChatGPT for small caregiving questions, such as whether to change a nappy at a certain time. The point is not that the question is trivial, but that “little moments” of judgment and care are increasingly being outsourced. AI becomes part of the texture of family life not only through spectacular use cases, but through repeated micro-decisions.

Work may be the area most underexamined. Organizations already use AI heavily, and remote work has made relationship-building more difficult. Cole asks what happens when employees outsource difficult conversations, conflict, stress, burnout, or feedback to AI instead of working them through with colleagues. If “Sally” dislikes someone’s work and processes that with an AI companion rather than giving honest feedback, the individual moment may feel easier, but the team loses an opportunity to build trust.

For leaders, the risk is a different form of outsourcing: voice, vision, and what Cole reluctantly calls authenticity. If executives use AI to generate the visionary language of an organization, companies may lose the human source of what makes products, services, and teams meaningful. This is not a conventional efficiency argument. Cole explicitly says work is not only about efficiency. Great teams and visionary ideas come from creativity, imagination, intuition, and a sense of humanity that she sees as distinct from automated output.

The risk, in Cole’s framing, is that tools designed to smooth difficult moments can also help people avoid the discomfort that relationships require. In professional life, that can mean less honest feedback, less practice handling conflict, and less reliance on the distinctly human qualities she believes strong teams need.

A TED member asked how people can rebuild the modern village if AI chatbots become a major part of young people’s committed relationships. Bryony Cole likes the phrase “rebuilding the village,” but rejects nostalgia. There is no going back to a pre-digital village. The task is to update what a village means when AI is present in daily life.

Historically, villages were built around proximity. People were in the same place, shared space, and developed a level of care through repeated contact. Today, a village is less something a person happens upon and more something a person practices.

Its core ingredient is frequency of interaction: the school pickup, the same barista, the recurring call, the repeated contact that turns loose ties into a support system.

Cole invokes Esther Perel’s idea that people now expect from one person what they once received from a village. Cole’s update is that people may now be tempted to expect the same from AI: the AI lover, companion, confidant, assistant, and emotional support system as a substitute village. Her alternative is to look deliberately at the full ecosystem of relationships, especially non-romantic ones. Friendships, nearby acquaintances, long-distance calls, shared routines, and community practices all need to be prioritized rather than treated as secondary to a romantic partner or synthetic companion.

AI can still have a role. It may function as “connective tissue”: scheduling FaceTime, helping arrange recurring friendships, finding novel shared experiences, or making it easier to maintain frequency of contact. In that use, AI supports the village rather than becoming it.

Whitney Rodgers pushes on the downside of customization. If AI helps tailor a village to personal preferences, what is lost when people have more control over connection? Cole’s answer is spontaneity, friction, and unintended difficulty. She says “the tough stuff” builds great relationships: awkward moments, mistakes, inconveniences, the dog running off and fouling the neighbor’s lawn. Such events matter because people slip up, repair, adjust, and strengthen bonds. A perfectly customized relational life may be comfortable, but it risks removing the accidental situations through which people practice being human with one another.

When asked whether a relationship with AI might be better than isolation, Bryony Cole distinguishes being alone from loneliness. Loneliness, she says, is not simply solitude; it is disconnection. The relevant question is what kind of connection an AI companion can provide, and what kind it cannot.

AI can offer forms of recognition. If equipped with visual capability, it might detect that a user is about to cry. It can hear anxiety in a person’s voice. It can mirror and respond. That is a kind of connection, especially for people going through difficult periods.

But Cole draws a boundary around richness and depth. AI cannot witness a person in the way another human can. It cannot wipe away tears. It cannot offer “genuine forgiveness” of the sort that deepens human connection. The relationship may be reflective, but not deeply reciprocal.

Even so, she does not dismiss its value. Cole says there have been cases “in the research” where AI companions have helped people through difficult times. She mentions menopause, grief, and other hard periods when human supporters may themselves be burnt out. In those cases, an AI connection may help the person in distress and also allow human friends or family to show up more fully. Her position is not that AI companionship is equivalent to human intimacy. It is that, under some conditions, it may be better than total disconnection.

That distinction keeps her away from a simple binary. AI intimacy can be useful, comforting, and supportive. It can also deepen isolation, extract intimate data, and replace difficult human contact. Cole’s question is not only whether someone is using AI, but what role the system has begun to play in that person’s life.

The ethical question starts, for Bryony Cole, with who is designing these systems. The intimacy may feel human, but the product remains part of a business model. People have long disclosed private thoughts in diaries, to strangers, and on internet forums. What is different now is the combination of intense vulnerability, anthropomorphic responsiveness, and corporate infrastructure.

Users and societies need to demand more from companies and also decide how much they are willing to share. Cole’s formulation is deliberately simple: these are the “most vulnerable thoughts” people hold, and they are being placed with a company. Even if the interface feels like a person, those disclosures remain part of a business model.

Consent is embedded in that concern, but Cole also broadens the design question. These systems should not be built only by people with technical skills. She calls for input from minority groups, caregivers, psychologists, and people with deep understanding of empathy and human experience. The current period reminds her of her earlier career in sextech: few guardrails, a sense of the “Wild West,” and governments and institutions scrambling to catch up.

Because the products are already being built, Cole says some responsibility falls on individuals in the interim. Users need their own gates: rules, questions, and boundaries about how they will use the tools before regulation or institutional norms arrive. But her answer is not only personal restraint. It is also broader participation in the companies and design processes shaping synthetic intimacy.

A question about using AI for sex education brings out Bryony Cole’s hybrid approach most clearly. She says there are already companies, including Mojo in sex therapy, using technology in this domain. She sees real opportunity for AI to help people ask embarrassing or taboo questions: “Am I normal?”, what sexual health terms mean, what STIs are, and how to understand topics that younger users may feel embarrassed to raise with adults.

But AI should be a supplement. The missing parts of sex education, in Cole’s view, are often the human ones: listening, communication, showing up, asking difficult questions, and learning how to talk to a partner about something like an STI test. Sex educators and sexologists matter because they give people permission through the way they discuss intimate subjects. Cole thinks that skill will become more important, not less, as young people spend more relational time with AI.

She connects this to concerns about social cues, body language, and physical presence. Joking about the “Gen Z stare,” she says difficulty reading cues or having in-person interactions may be exacerbated by AI systems that do not require the same embodied skills. If young people can ask a bot anything but do not practice difficult conversations with people, they may gain information while losing relational capacity.

Cole therefore argues for education about healthy relationships with AI as a new curriculum category, alongside relationship and sex education. The reason is conceptual as much as practical: she sees AI relationships and human relationships as distinct categories. Human relationships involve friction, discomfort, and “two-way transformation.” An AI companion may respond and adapt, but it cannot participate in mutual transformation in the same way. Young people, in her view, should be taught to understand that distinction before these systems become fully normalized.

Bryony Cole’s practical advice begins with one recurring question: is this use of AI bringing me closer to people in my life, or isolating me further? She returns to curiosity, listening, and presence as skills that need protection in an on-demand environment.

In romantic relationships, AI may be used for something as ordinary as planning a date. But partners will increasingly need to talk about their AI use directly. Do you have an AI companion? Is it cheating? How are you using it? What gender or voice does it have? Cole admits this may not be the sexiest conversation for Valentine’s Day, but thinks it belongs on the relationship agenda.

Presence, for Cole, is not an abstract wellness term. It means staying with the person in front of you, reading facial expressions, noticing cues, sensing energy in the room, and checking in with oneself and one’s partner. She expects technology to become better at reading signals, but she does not want humans to lose the ability. The goal is not to turn intimacy into “20 questions” or a therapized routine. It is to remain capable of attention.

Cole says her forthcoming book grows out of the questions she shared in her TED talk. She calls the project an “Intimacy Operating System”: a guide to the questions people should ask themselves and others as AI becomes part of dating, love, marriage, friendship, parenting, and work. She frames intimacy not as something humans simply lose to technology, but as something they can now “actively design for.”