The conversation has stalled. You’re sitting across from someone you matched with three days ago, and for a moment the restaurant feels very quiet. So you excuse yourself, walk to the bathroom, and type into your phone: What are good questions to ask on a first date to build a deeper connection?
Within seconds, you have a list. It sounds reasonable — thoughtful, even. The large language model (LLM) that generated it, like your own little dating coach, has probably drawn on Gottman research, attachment theory, life style blogs and several thousand Reddit threads, blended into advice that feels personalized but is, in fact, statistical: the most likely useful-sounding response, averaged across everything the model was trained on.
What Does a Dating Coach Do?
A traditional dating coach is a person — sometimes a therapist, sometimes not — who helps you identify patterns in how you approach relationships, gives you feedback on how you come across, and holds you accountable for the changes you say you want to make. The good ones observe you in real time: how you talk about yourself, how you describe what you want, where your story about past relationships gets stuck.
As of spring 2026, LLMs like ChatGPT and Claude have stepped into this role for millions of people. They’re accessible, anonymous, available at 2 a.m., and they don’t charge by the hour. People upload screenshots of text conversations for analysis. They ask how to respond to mixed signals. Red flags.They draft messages, decode silences, and rehearse difficult conversations — all through a chat window.
This is not inherently a problem. But it is worth understanding what’s happening under the hood.
AI Dating Advice: How LLMs Generate Relationship Guidance
A large language model is not an expert. It is as of April 2026 a prediction engine trained on vast quantities of text: books, articles, forums, therapy transcripts, pop psychology blogs, peer-reviewed research, and everything in between. When you ask it for relationship advice, it generates the response that is statistically most plausible given all of that training data.
The result often sounds good. It may even cite specific research — including the work of Drs. John and Julie Gottman, whose findings on what makes relationships succeed or fail are among the most well-established in the field. But here is the thing worth knowing: an LLM doesn’t distinguish between a concept grounded in four decades of longitudinal research, like the Gottman research, and a concept it absorbed from a lifestyle blog. It produces a blend, imagine a smoothie of the scientific and the commonplace. The person reading the advice has no reliable way to tell which is which.
Researchers sometimes describe this as a tendency toward the median. LLMs are trained on the full distribution of human text, and their outputs gravitate toward what is most commonly said. In relationship advice, this means that precise, research-based concepts — like Gottman’s specific framework for bids for connection, or the measurable dynamics of trust and 5:1 ratio — can get diluted into vaguer formulations: “bids for attention,” “be honest about your feelings,” “add a little humour to the conversation”. These sound right. They might not be wrong. But are they specific enough to be actionable in the way that research-tested Gottman-frameworks are?
This is what we might call concept contamination — the blurring of precise, evidence-based ideas with well-meaning but generic advice. It matters because the difference between “communicate better” and “turn toward your partner’s bids for connection 86 percent of the time” is the difference between a fortune cookie and a finding.
What the Research Actually Says About Dating
So what does the research say? In Eight Dates, Drs. John and Julie Gottman address the algorithm question directly, drawing on a study by psychologist Samantha Joel at the University of Utah. Joel’s team measured over 100 variables — self-esteem, goals, values, loneliness, what people said they wanted in a partner — and tried to predict whether two people would feel romantic attraction after a short date. Nothing worked. None of the variables predicted attraction.
In Eight Dates, the Gottmans outline four skills for intimate conversation that they consider foundational:
- Put words on what you feel. Not “I’m fine” or “I don’t know” — but the actual, specific feeling. Anxious. Hopeful. Overwhelmed. The Gottmans provide an emotional vocabulary checklist, because most of us have a smaller feeling-vocabulary than we realize.
- Ask open questions. Not yes-or-no questions, but the kind that invite the other person to say more than a sentence. What was that like for you? rather than Did you enjoy it?
- Make exploratory statements. Prompts that help a partner open up: Tell me more about that. Help me understand what that meant to you.
- Express tolerance, empathy, and understanding. Not agreement — understanding. The goal is not to fix or to win, but to make the other person feel heard.
These skills sound simple. They are not. They require what the Gottmans call attunement — the practice of actively tuning yourself to another person’s emotional frequency. It is a form of attention that cannot be faked, outsourced, or automated.
Discussion about this post