How to quickly fact‑check any AI answer in under one minute

Ask the AI for its source

When the AI (which is a large language model, or LLM – a complex computer program designed to understand and generate human-like text) provides a factual claim, immediately follow up by asking where it got that information. This is your first line of defence against AI hallucinations, which is when the AI confidently makes up facts or details that aren't true.

On screen, you’ll see the AI’s answer to your original question. After reading it, type your follow-up question in the text box at the bottom of the chat window and press Enter (or tap the send button, usually an aeroplane icon). The AI will then start typing its response, ideally listing specific sources like website links or publications. If it says it can't provide a direct source, it might explain that its knowledge comes from its vast "training data," which means you'll need to do more checking yourself.

You'll know it worked when the AI either provides specific links (URLs) or names official reports or organisations, or explains why it can't provide a direct source.

Open the source link in a new tab

If the AI provides a specific website link (URL), your next step is to open it in your web browser. This allows you to directly check the original information and see if it supports the AI's claim.

You'll see the AI's chat response, which should now include one or more clickable links. Click on the first link. If it's not clickable, carefully copy the entire web address (URL) – usually by highlighting it with your mouse or finger, then right-clicking (or long-pressing) and selecting 'Copy'. Then, open a new tab in your web browser, click into the address bar at the top, right-click (or long-press) and choose 'Paste', then press Enter. The webpage should load.

If the link is broken, leads to a completely different topic, or the content on the page doesn't match the AI's claim, then the AI's source is unreliable, and you'll need to treat its answer with caution.

You'll know it worked when you see the actual source content on a new browser tab and can visually compare the information there to the AI's original claim.

Cross-check with a second independent source

Relying on just one source, even an official one, isn't enough to be completely sure. A quick cross-check with a second, independent source significantly increases your confidence. This helps confirm the information and ensures the original source wasn't outdated or misinterpreted.

First, open a fresh tab in your web browser. In the search bar, type a short, specific phrase from the AI’s original claim, then add site:.gov or site:.edu to your search. This tells the search engine (like Google) to only show results from government (.gov) or educational (.edu) websites, which are generally considered highly reliable. For broader news, you could use site:.news for major news outlets.

You'll see a list of search results filtered to only show websites from the domain you specified. Click on one or two of the most relevant results to quickly scan their content. If you don't find relevant .gov or .edu results, try a more general search, but be more critical of the sources you find.

Look for another government department's statistics or a reputable university report on the topic.

You'll know it worked when you find at least one other reputable source that either confirms or contradicts the AI's claim and its first source.

Compare information and assess reliability

Now it's time to bring all three pieces of information together: the AI's original answer, its provided source, and your independently found second source. This comparison helps you make an informed decision about how reliable the AI's answer truly is.

Arrange your browser tabs or notes so you can easily see all three: your AI chat, the first source you opened, and the second source you found through your search. Carefully read the key facts from each and compare them side-by-side. Look for consistency in numbers, dates, and core statements. If all three align closely, you can be fairly confident in the AI's answer. If there are differences, note them.

If some numbers are slightly different (e.g., 5.1% versus 5.2%), consider whether the difference matters for your purpose. If they're wildly off, or one source contradicts the others entirely, the AI's initial answer should be treated as unreliable.

You'll know it worked when you have a clear picture of whether the AI's information is trustworthy (all sources align) or needs more investigation (discrepancies exist).

Refine your query or dig deeper (if needed)

If you found discrepancies in Step 4, or if the AI couldn't provide clear sources, don't give up! You can either go back to the AI with more specific questions or expand your independent research.

Return to your AI chat window. Based on what you found, politely ask the AI to clarify or correct its information. You can even provide the conflicting data you found. Alternatively, if the AI seems stuck or unable to help, switch back to your web browser and broaden your search terms, looking for explanations for any differing figures or more detailed reports. This iterative process is how real research is done.

The AI will likely start typing a new response, either acknowledging the discrepancy and correcting itself, or providing a new explanation. If you're doing a web search, new articles or reports might appear that shed light on the differences (e.g., provisional versus final figures, or different reporting periods). Remember, the AI can help draft the email, but it can't make real-world decisions for you – it's a tool to assist.

You'll know it worked when you either receive a corrected or more nuanced answer from the AI, or you find enough independent sources to confidently understand the information yourself, even if it meant adjusting the AI's initial claim.