14/09/2026
AI can sound very confident when it’s completely wrong.
Another simple AI tip from the Cre8ive team: make AI show you where it got its information.
AI “hallucinations” happen when tools such as ChatGPT or Claude generate information that sounds plausible, but isn't actually true.
The problem? The answer often doesn't look wrong.
It can be beautifully written, confidently stated and include very specific figures.
So when we're using AI for research where accuracy matters, we give it much tighter instructions.
Instead of:
❌ “What are the latest statistics on X?”
Try:
✅ “Answer using only information from [SPECIFIC WEBSITES/SOURCES]. Cite the exact source and link for every statistic or figure. If the information isn't available from those sources, say ‘not found’. Don't fill the gaps from memory.”
For example, if we wanted New Zealand population statistics, we could tell AI to use Stats NZ rather than simply asking it to find the answer.
Why?
🔎 You control the quality of the sources.
You're not leaving AI to decide which websites should be trusted.
🔗 You can check the answer yourself.
If AI tells you “37% of New Zealand businesses…”, you should be able to click through to the original source and confirm it actually says 37%.
And this is important:
⚠️ A citation isn't proof that the answer is correct.
AI can misinterpret a source, attribute information incorrectly or even provide a citation that doesn't support the claim.
For anything important, open the source and check it.
A good rule for using AI for research:
👉 Don't just ask for the answer. Ask for the evidence behind the answer.
AI can dramatically speed up research. It shouldn't replace verification.