Prompting for Accuracy: Fewer Made Up Answers
Use grounding, honest uncertainty and quote checks to cut down on confident but false answers.
A taste of a lesson
I asked for statistics on remote work productivity with sources. Two of the links did not work. Did it make them up?
Quite possibly. If your assistant was not using web search, it may have generated plausible looking references from patterns, not real pages. Even with search, links can be misquoted. So treat every reference as unverified until you open it. Better approach: find two or three real reports yourself, paste the relevant sections, and ask the model to summarise only from them, quoting each figure. Try it: which source would you trust most on this topic, and why?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Explain in plain words why models produce made up answers
- Ground answers in pasted sources and allow honest uncertainty
- Rewrite questions that contain false premises
- Verify citations and key facts outside the chat
Lesson plan
- 1 Why fluent does not mean true Understand how prediction produces plausible fabrications. Start
- 2 Grounding in sources Make answers come from material you provide. Start
- 3 Permission to be unsure Invite honest uncertainty instead of forced confidence. Start
- 4 Questions that mislead the model Spot and rewrite questions with built in false assumptions. Start
- 5 Citations and search Handle references from assistants with and without web search safely. Start
- 6 A personal verification habit Decide what to check, how and when, based on stakes. Start
Try asking
About this tutor
Language models sometimes produce fluent answers that are simply wrong: invented facts, fake citations, misremembered numbers. No prompt removes this risk, but good habits reduce it and make errors easier to catch. This tutor teaches practical techniques for everyone, from new users to experienced ones: giving the model sources to work from, explicitly allowing 'I do not know', asking for quotes before answers, separating facts from inferences, avoiding questions with false premises, and knowing when to verify outside the chat. You also learn why a tool with web search can still misread its sources.
Reviews
4.7
3 ratingsSample
- Viktor S.Sample
Sensible and not preachy. Grounding plus quotes is now how I use the assistant for policy questions at work.
- Amara O.Sample
Clear for a beginner like me. I check citations every time now, and I found two invented ones last week.
- Leila B.Sample
The false premise lesson was eye opening. I had been asking 'why' questions about things that never happened and getting confident stories back.
About the teacher
Prompt workflows for heavy users: chaining, standing instructions, long documents and reasoning models
9 tutors 269 lessons taught Sample
I work with people who already use AI assistants every day and want more dependable results. My background is in operations and process design, which taught me to treat a prompt like a small procedure: inputs, steps, checks and a clear output. I teach chaining, reusable instructions, long document work and how to test whether a prompt change actually helped....
See Darius's profile and tutorsMore like this
Other tutors on the same or nearby topics.