Elton Jones uses 10 prompts to improve ChatGPT accuracy
Updated
Updated · Tom's Guide · Apr 30
Elton Jones uses 10 prompts to improve ChatGPT accuracy
6 articles · Updated · Tom's Guide · Apr 30
In a Tom's Guide article, he says the prompts target hallucinations, vague answers, overconfidence, wordiness, poor structure and weak reasoning.
His suggested instructions include asking for uncertainty flags, clarifying questions, contrarian views, practical action plans, concise bullet points and "I don't know" instead of guesses.
Jones argues reusable prompt templates can make ChatGPT more useful for personal and professional tasks by producing more tailored, structured and actionable responses despite the chatbot's known limitations.
If we must constantly prompt AI not to lie, can we ever truly trust its answers as factual?
As AI still hallucinates in 2026, are advanced systems making these 'prompt hacks' obsolete?