AI Fundamentals
Catch AI when it’s confidently wrong .
AI models use the same confident tone for things they know and things they’ve guessed. This short trail gives you the one follow-up question that pulls those apart.
No credit card required. Cancel anytime.
Why not just ask "are you sure?"
"Are you sure?" is the natural follow-up, and it’s the wrong one. The model will apologize and change its answer whether or not the original was wrong. Sometimes the new answer is right, sometimes it’s a fresh hallucination. You won’t be able to tell which.
The follow-up this trail teaches works differently. Instead of asking the model to evaluate its own confidence overall, you ask it to *separate* what it verified from what it reconstructed. Models can do this better than they can self-evaluate. Because you’re asking for two lists instead of one judgment.
No new tool, no paid fact-checking subscription. Two prompts in any chatbot (ChatGPT, Claude, Gemini, Copilot) in under 5 minutes. Once you have the pattern you keep it forever.
Questions
- Is my data private?
- Yes. Your prompts and AI responses are deletable, exportable, and only visible to you. We never expose individual prompts to admins, teammates, or Joring employees.
- Does this work for citations and quotes specifically?
- Yes. Citation hallucination is the most-cited failure mode this technique catches. After the model gives you a list of citations, run the confidence-check prompt and it will typically admit which references it reconstructed from training data vs. which it can verify.
- Which AI tools does this work in?
- All of them. ChatGPT, Claude, Gemini, Copilot, Perplexity, Llama-based assistants. The technique is about how you prompt, not which model you’re prompting.
- Doesn’t the confidence answer itself just hallucinate?
- Sometimes it does, yes. The technique gets you a much better signal than not asking. And it pairs well with external verification (look up the specific facts in a real source). Treat the confidence breakdown as a triage tool, not a final truth.
- How is this different from a fact-checking tool?
- Fact-checking tools require you to know in advance what to check. This technique surfaces what to check. The model itself flags which parts it’s uncertain about, so you know where to focus external verification.
- What happens after the 14-day trial?
- Continue on Pro to keep real-time coaching, guided trails, and your exportable AI record. Or stay on Free with limited features. The trail progress you’ve already made doesn’t disappear.
Use AI as a fast collaborator, not a confident oracle.
No credit card required. Cancel anytime.