Want better answers from AI? Tell it to ask questions | MarTech

"GenAI chatbots' priorities are first to be helpful, second to be harmless and third to be accurate. So it always provides an answer - even if it means making up the answer (aka, hallucinating). Both OpenAI and Gemini are built to keep things moving. When they see a prompt that's a little vague, they'll often pick the most common one and move on-unless you tell it not to. To get more control, try adding a line like:"

"If you want this to apply across a more extended conversation, start your session with something like: "For this session, don't assume anything. Always ask for clarification first if a prompt isn't clear." That kind of strong opening can keep the instruction top-of-mind for the model-at least for a while. Keep in mind that Gemini doesn't offer true memory or session settings, so you may need to repeat the request later."

Generative AI often behaves like an overly enthusiastic intern: it acts fast, assumes understanding, and can produce off-target answers. GenAI chatbots prioritize being helpful, harmless, and then accurate, which can lead them to provide answers even when unsure, sometimes fabricating information. Both OpenAI and Gemini fill gaps using training data and common patterns unless instructed otherwise. Explicit prompts that require clarification, session-level instructions to avoid assumptions, and periodic repetition of those instructions can reduce misinterpretation. Gemini emphasizes speed and lacks persistent session memory, so reminders may be necessary to maintain clarification behavior.

Read at MarTech