A mirror, with the method in view.

TraitSpring helps adult women reflect on relationships with men. The web experience adapts the owner’s original Chinese mini-program: profile, conversation, confirmed facts, rule-based calculations and a personal report.

Your information comes first.

Fifteen short profile sections cover your circumstances, resources, responsibilities, preferences and real experiences. Optional questions may be left blank. Unknown and “prefer not to say” do not become zero. Your own money, family resources, partner preferences and actual candidates stay separate.

The AI conversation is a place to add context. Before generating, review the profile. If you changed a numerical detail in chat, update the corresponding field too: only the confirmed profile feeds the program’s score. Editing a profile or continuing the conversation clears the old report so it cannot silently appear current.

What the score means.

The source product uses nine factors. Its female weights are subjective appearance 30%, age 15%, personal net assets 10%, income 10%, family net assets 10%, education 10%, height 5%, income stability 5% and available family support 5%. These are product choices, not measured laws of attraction. The original male rule assigns more weight to financial inputs; equal numbers do not establish mutual interest.

Each factor uses the original versioned scoring anchors. Missing fields contribute an interval instead of an invented number. Negative net assets remain visible as negative values in the profile and use the scoring table’s minimum of zero. Weight, parents’ occupations and desired partner conditions never add points. Appearance can use an optional self-rating or an explicitly requested AI photo impression; neither is an objective beauty measurement.

The source project cites Xu (2025) for factors discussed in a Chinese survey experiment. It does not validate these exact product percentages, age curves or monetary anchors. This model has not been validated for Western dating. It cannot measure your worth, predict a relationship, give a percentile or calculate the wealth of someone you can obtain.

Currency is explicit.

You can enter USD, EUR, GBP, CAD, AUD or CNY. Money is kept in your stated currency and tax basis. To use the original CNY scoring table, we fetch dated European Central Bank reference rates. This is nominal currency conversion, not a purchasing-power or dating-market adjustment. If a valid recent rate cannot be obtained, the affected financial inputs remain unscored.

What AI does—and does not do.

OpenAI GPT-5.6 Terra writes the conversation and report. Calculations remain programmatic. Report quotations must exist in the material you supplied; an exact quote is still not proof that the model interpreted it correctly. The report separates stated needs, observations and hypotheses. It is neither therapy nor a clinical or psychological assessment.

The optional photo feature discusses visible presentation. Photos are not used to identify you, infer your personality or set a partner’s financial ceiling. OpenAI GPT Image 2.5 Flare can create an optional illustration from your report’s written portrait brief. Your uploaded photo is not used for this illustration. It is fictional: no real matching database, future prediction or biometric inference is involved.

Your choices stay yours.

Review every conclusion against your experience. You can edit, regenerate, print or delete your browser copy. Free AI access currently requires ChatGPT sign-in and has daily limits: 50 conversation requests, 4 report requests, 3 photo requests and 2 illustration requests per signed-in account. Failed provider calls may consume an allowance; there is no payment charge by TraitSpring in this release. A site-wide capacity limit also applies.

For sensitive or urgent situations, a reflection tool cannot replace personal support from a qualified professional. Read our privacy details before sharing a story or a photo.

AI report version: mirror-openai-20261026-v1. Scoring version: product-index-20261025-v3. Updated October 26, 2026.