AI POWERS
How to use and verify an AI answer
A practical method for asking clear questions, checking claims, tracing sources and deciding when an AI answer needs expert review.
Start with a question you can check
State your goal, audience, relevant constraints and the format you need. Separate what you know from what you want the AI to work out.
For a current question, include the place and date. Ask the model to identify missing context instead of filling gaps with confident guesses.
Read the answer in layers
First identify its main claim. Then mark facts, estimates, interpretations and suggestions separately. A fluent explanation is not evidence that each part is correct.
Look for dates, numbers, names and quotations: these are easy to repeat and easy to get wrong. Ask for uncertainty and alternatives where the evidence is incomplete.
Check the original source
Find the publication, official document, research paper or primary dataset behind an important claim. Open it and confirm the passage, date, scope and any later correction.
A list of plausible links is not proof. Check that the URL works and that the page actually supports the specific statement being made.
Match the checks to the stakes
For health, law, finances, safety and other consequential choices, verify against current authoritative sources and consult a qualified professional when needed.
For a low-stakes draft, a quick factual review may be enough. For a decision with real consequences, document what you checked and when.
Protect information and keep control
Remove personal, confidential or sensitive details before submitting a question. Use a fictional example when the real details are not essential.
Keep the final decision, attribution and corrections with a person. Recheck time-sensitive advice before using it again.
A simple way to explore a subject
AI POWERS uses your question, the selected topic and language to prepare a concise answer. Requests are independent: previous questions are not conversation context. A recent answer to the same normalised question in the same topic and language may be reused from a temporary cache. The service does not automatically consult the independent resources listed on these pages.
Does this interface accept files or browse every cited page?
This interface accepts a short text question. Guides may describe methods using other tools; they do not mean that file uploads, audio transcription or web browsing are available here. Open and verify important references yourself.
Sources for further reading and checking
These independent resources help you find a reference. Check the language and scope: a French resource may not describe rules in another country. These are not sources automatically consulted for generated answers.
- CNIL — IA générative : questions-réponses
Questions and answers on using generative AI and protecting data.
Language: fr · Scope: France / UE - NIST — Generative AI Profile (AI 600-1)
A generative AI risk profile for identifying and managing risks.
Language: en · Scope: US - UNESCO — Guidance for generative AI in education and research
Guidance on generative AI in education and research.
Language: en · Scope: International
Link check: 2026-10-06 · These links are reading resources. The answer service does not consult them automatically. A checked link does not guarantee every document is current.