# How the AI learns your taste

> Thumbs up, thumbs down — how your ratings shape future plans.

Forkboard starts from the little you tell it at setup, then gets sharper from what you actually eat. There's no long quiz on purpose — your ratings teach the board far more than a setup screen ever could, and they keep teaching it every week.

![A meal card with thumbs up and thumbs down feeding into the next week's board.](https://forkboard.app/docs/teaching-thumbs.svg)

*It refines from what you actually rate.*

## The three signals you send

Every dinner card carries a small rating bar. The way you use it quietly shapes what comes next:

- A thumbs up nudges future plans toward that kind of meal — its cuisine, its style, its level of effort.
- A thumbs down steers the board away from it.
- Swapping a night out sends a quiet negative signal too, even if you never touch a thumb. The board reads it as 'not this one' and leans elsewhere.

You don't have to rate everything. A few honest votes a week are plenty, and you can always change your mind — tap the other thumb to flip it, or tap the active thumb again to clear it.

## What the next plan is built from

When Forkboard drafts your week, it works from your household's food picture — nothing more. That picture is:

- Your cuisines, your avoid list, and any dietary patterns or allergens.
- Your kitchen settings — skill, equipment, weekday and weekend time, serving size, units.
- What's in your pantry.
- Your household's recent ratings — the meal titles and notes from the last 90 days, so the board has a sense of what's been landing lately rather than a year ago.

Add a short note when you rate and it carries weight: 'too spicy for the kids' or 'make this again' tells the board the why, not just the verdict. Notes stay short, up to 280 characters.

## What it never sees

The planner is built only from food preferences, ratings, and your pantry. It never receives your name, your email, or any billing detail — those simply aren't part of how a plan is made.

> Note: Your data is never used to train AI models, and it is never sold. It exists to plan your dinners, and that's the whole of it.

Related: [Rating & reporting meals](https://forkboard.app/docs/rating-and-reporting.md) · [Account, data & privacy](https://forkboard.app/docs/account-and-privacy.md)

Canonical page: https://forkboard.app/docs/teaching-the-ai