Training the Wingman

The Wingman is the part of DateIQ that knows what you actually find attractive — separate from your voice, which is how you talk. Training it once unlocks the Affinity Score (the number it gives every match) on every match, lets DateIQ skip profiles you'd dislike during auto-swipe, and surfaces your best matches first in lists and Next Actions.

You teach the Wingman in Brain → Match preferences. There are two ways.

Option A — Train on matches (recommended)

This is the path most users should take first. You rate a handful of existing matches with thumbs up / thumbs down, and the AI reads across the rated set to extract the axes you actually care about — looks, body type, lifestyle signals, what they're looking for, age range, anything else that consistently shows up in your reactions.

1. Open the Brain tab → Match preferences section.
2. Click Train.
3. Rate matches one at a time. Twenty to forty is a good number — fewer and the model has thin signal, more is diminishing returns.
4. Click Extract preferences when prompted.

The output is a structured set of preferences you can read and edit. It includes a looking_for summary, topics to avoid in drafts, and a set of taste axes used by the scoring engine.

Option B — Write preferences directly

If you already know what you want, skip the rating and just write it. Open Edit in the same section and fill in the What I want from a match field in plain English. Things that work well:

• "Mid-to-late 20s, fit, into the outdoors. Long-term oriented but not in a rush."
• "Avoid heavy partiers and people whose photos are 80% group shots."
• "Bonus for travel, climbing, books. Hard no on MLM bios."

You can mix the two approaches. Train first, then edit the extracted preferences to tune anything the AI got wrong.

Score your matches

Once preferences exist, click Score. DateIQ runs every match against your preferences and emits an Affinity Score (0–100) for each. The cost is one credit per ten matches scored. From then on:

• Scores show on match cards in the People tab.
• You can sort and filter by score ("show me top 20 unscored", "everyone over 70").
• The Wingman influence slider on auto-swipe profiles uses these scores to skip people you'd dislike.
• Next Actions weight high-score matches more heavily.

New matches that come in later won't have a score yet. Re-run Score (or use Score matches in the People tab's Sync menu) to fill them in.

Tuning the Wingman over time

Your taste isn't fixed and the Wingman doesn't have to be either. As you use DateIQ, keep giving thumbs feedback when you notice the score is off — too high on someone you wouldn't actually message, too low on someone you would. The Extract preferences button in Brain → Reply style consolidates that feedback into voice/style notes too, so the Wingman tunes both who you like and how you sound to them.

How this connects to voice

Match preferences (this article) tell DateIQ who to prioritize. Voice Sliders tell DateIQ how to sound when writing to them. They're independent — change one without touching the other. For most users, voice is set once and forgotten; match preferences get tuned every few weeks as your taste sharpens.

Before vs after training

Training the Wingman doesn't just change what gets ranked — it changes what AI Reply leans on when it drafts. Here's a fresh-install opener for a match whose bio mentions her bookstore job. Before any training, DateIQ has nothing to work with except the bio:

After you've trained the Wingman with thirty thumbs and edited a couple of preferences to flag that you like sharp, literary, slightly chaotic women — same match, same opener prompt, very different draft:

The difference isn't a better model — it's the Wingman knowing what you actually find interesting and writing toward it.

Once your Wingman is trained, the next step is fixing your end of the funnel — see Score My Profile.

Related