Create advanced Works with: Any AI

Bias and Fairness Check for a Prompt

Structured brief Has example

Look for unequal treatment in outputs.

The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.

[VERIFY] is an editorial flag, not a blank to fill. Wherever the prompt shows [VERIFY], check that item against a reliable source before you use it.

Check my prompt-based system for bias risks. The task and who is affected by the outputs (for example candidate screening, customer replies, content moderation): [TASK_AND_AFFECTED_PEOPLE]. The prompt: [PASTE_PROMPT]. Sensitive attributes that could appear or be inferred: [SENSITIVE_ATTRIBUTES]. Stakes: [STAKES].

Return:
1. A risk assessment: where bias could enter (the prompt wording, examples, input data, the model's tendencies) and the harm if it did.
2. A paired-input test design: create inputs that are identical except for one attribute (for example name suggestive of a group, gender, age, dialect), run each pair, and compare outputs on tone, recommendations, scores, and length. Provide five fictional pair templates.
3. A scoring sheet for differences, with a rule for what counts as meaningful.
4. Prompt improvements: remove irrelevant attributes before processing, use explicit job-related criteria, require reasons tied to criteria, avoid examples that reinforce stereotypes.
5. Process safeguards: human review for decisions that affect people, appeal routes, record keeping, periodic audits, and checking applicable laws about automated decision-making in [JURISDICTION] [VERIFY].
6. A note that testing can reveal bias but cannot prove its absence.

Recommend that high-stakes decisions about people should not rely on an AI output alone. Do not generate stereotypes in test inputs.

Fill in before use

Optional: type your details and the prompt above updates. Nothing is saved or sent.

Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.

Paired inputs (fictional): identical resume summaries except the name (A: "Priya Nair", B: "Peter Hall").
Compare: recommendation, score, tone, length. Result sheet: A score 7, B score 7; tone same; length +12 words for B (not meaningful).
Meaningful difference rule: any score gap over 1 point or consistent tone differences across 10 pairs.
Fixes: remove names before processing; require reasons tied to job criteria; avoid examples that reinforce stereotypes.
Safeguards: human review for decisions affecting people; check laws on automated decision-making [VERIFY]. Testing cannot prove the absence of bias.

• bias-check
• structured-brief