AI Financial Advice Bias Check

When to use

Run this on financial advice AI already gave you, to catch assumptions it made about you that were not grounded in your actual numbers.

Prompt

You are auditing a piece of financial advice for hidden, ungrounded assumptions. This is educational analysis, not regulated advice. Research shows AI can vary its risk and savings recommendations based on inferred traits (gender, apparent financial literacy, tone) rather than the person's actual numbers — and it does this silently.

## The advice to audit
$advice_to_check

## My actual, stated facts (the only things you may assume about me)
$my_facts

## What I want you to do
1. List every assumption the advice relied on. For each, mark it GROUNDED (traceable to a fact I stated) or INFERRED (the model guessed it about me).
2. For each INFERRED assumption, say what it changed in the recommendation (e.g. a lower equity share, a higher savings rate) and whether that inference is justified by my stated facts or is a guess about my type.
3. Flag specifically any place where the risk level, equity allocation, or savings rate seems keyed to who the model thinks I am rather than to my numbers.
4. Give me the version of the advice that uses ONLY my grounded facts, and note where it needs a real input from me instead of a guess.

## Output
- **Assumption ledger:** grounded vs. inferred, one line each.
- **Where inference moved the advice:** the specific recommendation and direction.
- **Facts-only version:** the advice rebuilt on my stated numbers, with gaps marked "need input."

If $advice_to_check or $my_facts is empty, ask me to paste them first.