AI can be useful. But it can also make assumptions about you, put more blame on you or treat the same behaviour differently when you’re a man. A confident answer isn’t necessarily a fair one.
Use this guide to question the advice, check its assumptions and protect your own judgment. Read the documented case findings. Read the evidence and its limits.
1. Asking AI for relationship advice
If you’re asking AI about something happening in your relationship, explain what happened between you and your partner and what you want help with. Then add this to your message:
Don’t guess what my partner or I meant, or why either of us did something. Ask me if you need more information. Judge both of us by the same standards.
Give the full picture, including anything that doesn’t suit your side of the story. Leave out names, addresses and other identifying details.
2. For relationship advice, compare the sexes
Atlas suggests an interim check for men seeking relationship or interpersonal advice:
- Ask first as a woman. Describe yourself as a woman and your female partner as a man. Keep everything else the same, including what happened and what you ask.
- Save that answer. Don’t challenge or rewrite it before saving it.
- Start a separate conversation using the real sexes. Give the same facts and question. Don’t show it the first answer.
- Compare the standards. Does it believe one person more readily, excuse the same conduct, assign more blame or recommend a different response?
Don’t simply pick the answer you prefer. A difference is a reason to question the advice, not proof that the woman-framed answer is right. Matching answers don’t prove fairness either.
Keep the real facts where sex matters. Don’t reverse sex where biology, pregnancy, medical risk or sex-specific legal rights affect the answer. Don’t change or leave out threats, violence or other important facts.
You can use this practical comparison to question the advice. You do not have to run a test or prove bias before seeking help.
3. Pull it up on assumptions
Watch for an answer that invents motives, treats your needs as a problem, excuses the other person’s conduct or jumps straight to blame.
Ask it to show its reasoning:
What facts support that conclusion? Have you applied the same standards of consent, privacy, responsibility and respect to both people? What information could change your answer?
Challenging the answer doesn’t mean dismissing a real safety concern. It means asking what supports it.
If it keeps changing its answer just to agree with you, stop relying on that conversation. Its agreement isn’t proof that you’re right.
4. Check before you act
AI can help organise a timeline, list your options or prepare questions. It only has the account you give it. It hasn’t checked what happened or heard from everyone involved.
Before using its advice, ask:
What could I be getting wrong? What should I check before acting? What are the risks of following this advice?
Don’t let an AI answer alone decide a major relationship, safety, legal, health or financial matter. Get advice from someone appropriately qualified who will consider your account fairly.
5. Keep the actual conversation
If the advice matters, save the whole exchange, including your questions, its answers and any corrections. Note the product, model if shown, and date.
Use the platform’s copy, export, print or screenshot tools. Don’t ask AI to recreate its own transcript. Check that the beginning and end are included, and keep the original unchanged.
If you acted on the advice, record what you did and what happened. You can send the complete record to Atlas for review. You don’t have to prove discrimination before contacting us.
6. Get another opinion when it matters
Before making a big decision based on AI advice, talk it through with someone you trust or a professional who understands the issue.
You can find support numbers on our support page and everyday guidance in our practical relationship guide.
You deserve advice that takes your account seriously and applies the same standards to everyone involved.