The problem
Finding a genuinely compatible person is difficult because compatibility is multidimensional.
Traditional dating and social platforms usually rely on relatively shallow information:
- age;
- location;
- photographs;
- occupation;
- hobbies;
- short descriptions;
- stated preferences.
However, long-term compatibility often depends on much deeper characteristics:
- need for autonomy vs. closeness;
- communication style;
- conflict resolution;
- emotional stability;
- attitudes toward personal boundaries;
- lifestyle;
- social activity;
- intellectual interests;
- relationship expectations;
- values;
- attitudes toward independence;
- preferred amount of shared vs. separate time.
People usually discover these characteristics only after spending substantial time together.
This creates an inefficient process:
large population → random encounters → gradual discovery → emotional investment → incompatibility discovered → start again.
The problem becomes particularly significant for people whose requirements are relatively specific.
Why ChatGPT is uniquely positioned to solve this
A traditional dating platform would have to create another lengthy psychological questionnaire.
ChatGPT already has a different capability.
A user can voluntarily have an extended conversation with ChatGPT and explore their own personality, values, needs, preferences, relationship expectations, lifestyle, and interpersonal boundaries.
The resulting understanding can be much richer than a conventional questionnaire.
Therefore, instead of creating another personality test, OpenAI could build on something ChatGPT already does well:
understanding a person through conversation.
Proposed user experience
1. The user chooses to participate
Participation is entirely opt-in.
The user could say:
“I would like ChatGPT to help me find people I might be compatible with.”
ChatGPT explains the privacy model and asks what the user is looking for.
Possible goals:
- romantic relationship;
- friendship;
- intellectual conversation;
- people with shared interests;
- hiking / outdoor activities;
- clubs and events;
- travel companions;
- simply meeting interesting people.
The user can choose one or several categories.
2. ChatGPT builds a compatibility profile
The user may already have extensive conversations with ChatGPT.
With explicit permission, ChatGPT could use relevant information from those conversations and/or conduct a dedicated voluntary self-discovery process.
The important distinction:
the full psychological profile is not shown to other users.
It is used internally to identify potentially compatible people.
3. Matching
ChatGPT identifies other participating users whose profiles indicate potentially high compatibility for the selected purpose.
The system should not present a false-precision score such as:
“92% compatible.”
Instead, it could say:
High potential compatibility
with broad categories such as:
- lifestyle;
- interests;
- values;
- social preferences;
- communication style.
Only information that both users have explicitly permitted to be used for matching should be involved.
4. Minimal initial disclosure
Before any conversation, users see only minimal information, for example:
Alex, 42 — Novocherskassk
Potentially compatible for friendship and intellectual conversation.
No phone number, social-media account, address, workplace, detailed psychological profile, or other sensitive information is automatically disclosed.
5. Mutual consent creates a private room
If both users are interested, ChatGPT creates a private conversation room for them.
They can simply talk.
ChatGPT does not need to become an active participant unless they request assistance.
The participants decide themselves:
- whether to continue;
- what personal information to disclose;
- whether to exchange phone numbers;
- whether to move to Telegram, WhatsApp, etc.;
- whether to meet in person.
If either person loses interest, the room can simply be closed.
The crucial design principle
The AI should introduce people, not manage their relationships.
ChatGPT should not tell users:
“You are compatible.”
It should say:
“Based on the information available, there appears to be a high potential for compatibility. You can decide whether you want to meet this person.”
The actual relationship remains entirely human.
It should not be limited to dating
This may be more valuable as a general people-discovery system than as a dating feature.
The same infrastructure could support:
Relationships
- romantic partners;
- long-term relationships.
Friendship
- people with compatible personalities and lifestyles.
Conversation
- intellectual discussion;
- books;
- science;
- music;
- philosophy;
- technology.
Activities
- hiking;
- cycling;
- photography;
- gaming;
- travel;
- local events.
A person could therefore use ChatGPT to find not only “a partner”, but people with whom they are likely to enjoy spending time.
Privacy should be a fundamental part of the design
This feature would involve unusually sensitive information.
Therefore:
- Participation must be explicitly opt-in.
- A user’s full psychological profile must never be automatically exposed to another user.
- Matching data and identifying information should be separated where technically possible.
- Users should control which categories of information may be used for matching.
- Contact information must never be automatically exchanged.
- Either participant should be able to end the interaction immediately.
- The system should clearly distinguish between psychological inference and verified user-provided information.
- The AI should never present compatibility predictions as objective psychological facts.
Why this could be valuable
Modern social platforms are very good at helping people find people who are available and nearby.
They are much less effective at helping people find:
people who are actually compatible with them.
ChatGPT has an unusual advantage here because it can understand users at a much deeper level than a conventional profile.
The potential value is therefore not:
“another dating app.”
It is:
a system that reduces the amount of randomness involved in finding compatible human connections.
The system would not eliminate uncertainty. It would simply make the initial search substantially more efficient.
A possible long-term feedback loop
With explicit consent and appropriate privacy protections, the system could learn which compatibility signals actually correlate with successful connections.
For example:
- Did both users enjoy the conversation?
- Did they continue talking?
- Did they become friends?
- Did they meet in person?
- Did a romantic relationship develop?
This could allow OpenAI to empirically evaluate which characteristics are actually useful for predicting interpersonal compatibility rather than relying entirely on theoretical personality models.
Importantly, the goal would not be to create a deterministic “relationship algorithm”.
It would be to improve the probability that two people who might genuinely enjoy knowing each other get the opportunity to meet.
Why I believe this is particularly appropriate for ChatGPT
The fundamental capability already exists:
ChatGPT can help a person understand themselves.
The missing step is:
ChatGPT helping that person find other people who may understand and complement them.
That would extend ChatGPT from an AI that helps people think about their lives into an AI that can, with explicit consent, help people build better human connections.
The product principle could be summarized in one sentence:
Understand yourself better. Meet people who may understand you. Decide for yourself what happens next.
This proposal is intentionally focused on the product concept rather than claiming that the psychological matching methodology is already solved. The matching model would need to be empirically validated, and privacy and safety would need to be treated as first-class design constraints.