How to Check an AI Answer Before You Share It

An AI answer can look ready to forward: tidy bullets, precise numbers and a convincing explanation. That is exactly when it is worth slowing down.

OpenAI describes hallucinations as plausible but false statements generated by language models, and says ChatGPT can still hallucinate. Useful writing is not proof that the facts are right.

1. Find the claims that matter

Mark the parts that could cause trouble if wrong: prices, deadlines, statistics, quotes, product claims, and advice on health, money or legal matters. A clumsy caption needs an edit. A wrong application deadline can cost someone an opportunity.

1. Open the source, not just the citation

Ask the AI for the original source, then open it yourself. Check that the page supports the exact claim. For features, use official docs. For prices, the pricing page. For deadlines, the organisation's notice. A real link can still point to the wrong page. If you can't find support, remove the claim or say it's unverified.

1. Check the date, region and plan

If an AI says a tool is free, check what "free" means: trial or ongoing? Which features? Available in India? Payment method needed? A correct answer from last year may be wrong today. Note when you checked.

1. Recalculate numbers and dates

Redo any percentage, total or conversion. Verify the weekday and timezone for dates. 9 AM in another country is not automatically 9 AM IST.

1. Ask for uncertainty, not reassurance

Instead of "Are you sure?", try: "List the factual claims in this answer. For each, give the original source and date. Separate confirmed facts from assumptions, and say what you could not verify." Treat it as a checking aid, not a certificate.

Before you hit send:

• Did I open the source for the claims that matter?
• Does it apply to this date, location and plan?
• Did I check the numbers and timezone?
• Did I remove unsupported claims or label uncertainty?

Use AI to get a first draft faster. Check the facts before that draft becomes someone else's decision.

Source: https://openai.com/index/why-language-models-hallucinate/