{"id":10533,"date":"2025-11-04T16:39:36","date_gmt":"2025-11-04T22:39:36","guid":{"rendered":"https:\/\/attentionmedia.io\/?p=10533"},"modified":"2025-11-04T16:39:36","modified_gmt":"2025-11-04T22:39:36","slug":"how-to-get-genai-to-say-it-doesnt-know","status":"publish","type":"post","link":"https:\/\/attentionmedia.io\/?p=10533","title":{"rendered":"How to get genAI to say it doesn\u2019t know"},"content":{"rendered":"
\"\"<\/div>\n

What would you call an assistant who invented answers if they didn\u2019t know something?<\/p>\n

Most people would call them \u201cFired.\u201d<\/p>\n

Despite that, we don\u2019t mind when AI does it. We expect it to always have an answer, but we need AI that says, \u201cI don\u2019t know.\u201d That helps you trust the results, use the tool more effectively and avoid wasting time on hallucinations or overconfident guesses.\u00a0<\/p>\n

Here are ways to get ChatGPT, specifically, to be clear about its limitations, and questions you can use with any genAI.<\/p>\n

Getting ChatGPT to say if it doesn\u2019t know<\/h2>\n

To reduce dead ends and increase trust in the output, you can guide your session with a few smart habits:<\/p>\n

Ask for candor upfront<\/strong>: Use a prompt like: \u201cIf you don\u2019t know something, say so and explain why.\u201d<\/em><\/p>\n

Challenge vague responses<\/strong>: When answers feel fuzzy, follow up with: \u201cAre you certain about this, or are you guessing?\u201d<\/em><\/p>\n

Dig deeper: Key questions to ask when vetting AI tools for marketing<\/a><\/em><\/strong><\/p>\n

Request the reasoning<\/strong>: If ChatGPT says \u201cI don\u2019t know,\u201d it can also explain whether that\u2019s due to a lack of data, policy restrictions, or technical limitations.<\/p>\n

Reward honesty<\/strong>: If you get a clear \u201cI can\u2019t answer that,\u201d reinforce it with feedback like: \u201cThanks \u2014 that\u2019s what I needed.\u201d<\/em><\/p>\n

These small steps reinforce that accuracy matters more than filling in blanks.<\/p>\n

Pro tip:<\/strong> To apply this across an entire session, add a direction at the start like: \u201cOnly respond if you can verify the answer. If not, say you don\u2019t know and explain why.\u201d<\/em><\/p>\n

(Note: ChatGPT won\u2019t remember this across sessions, so you\u2019ll need to repeat the direction each time.)<\/p>\n

How to evaluate other AI tools for transparency<\/h2>\n

Generative AI platforms\u2019 priorities are: Be helpful, be safe and be accurate \u2014 in that order. That\u2019s why most of them will always generate something, even if it\u2019s a hallucination. The one notable exception is Google Gemini, which has \u201cI don\u2019t know,\u201d baked into it.\u00a0<\/p>\n

If it can\u2019t find a verifiable answer to a prompt, it explicitly says so. It then explains why, providing context for the failure and determining whether the information is likely private, too recent, undocumented or simply not publicly available\/documented.<\/p>\n

Here are questions you can use to test whether other tools will prioritize accuracy over performance:<\/p>\n

Capability testing<\/h3>\n