Verify before you trust

This course remembers completed sections and your last course page in this browser on this device. RayAI does not receive this progress or anything you write. Unfinished answers and written responses are not saved, so an interrupted activity restarts when you return.

MODULE 5 | 35 minutes | FAST TRACK CORE + GO DEEPER

Verify before you trust

Core idea: fluent is not the same as true

AI output can be clear, detailed, and wrong. Verification means comparing a important factual claim against evidence or a reliable method that is sufficiently independent of the original generation process.

The verification ladder

1. Fit: Does the output actually follow the task, audience, boundaries, and format?

2. Grounding: Does each important factual claim match the supplied source?

3. Currency: Does a changeable fact come from a current authoritative source?

4. Calculation: Can numbers be recomputed with a calculator, spreadsheet, code, or another reliable method?

5. Completeness: What is missing, ambiguous, or outside the evidence?

6. Fairness and consequence: Who could be affected, and what review level does that require?

7. Decision: Use, revise, escalate, or reject.

Confidence is not evidence

A model can sound certain when it is wrong and cautious when it is right. A percentage such as '90% confident' is not automatically calibrated. Prefer source quality, direct evidence, reproducible calculation, and clearly stated uncertainty.

VERIFICATION PROMPT

Citations and links

• Open every important citation. Confirm that the source exists, supports the claim, and is current enough for the purpose.

• Prefer primary and authoritative sources for laws, policies, standards, research, product documentation, and current facts.

• Distinguish a quotation from a paraphrase, and do not attribute words that the source did not say.

• If the AI tool cannot browse or access the source, provide the source yourself or verify outside the tool.

Verify in proportion to consequence

Not every sentence needs the same checking effort. Verify claims that could change a decision, create a commitment, affect another person, involve money or safety, or be published as fact. For low-stakes brainstorming, a quick sense-check may be enough. For consequential work, use independent authoritative evidence and qualified review.

CLAIM CHECK

See it: verification changes the event plan

The draft claims that the venue holds 100 people and is wheelchair accessible. Neither fact appears in the planning notes.

CHECK

Spot the failure

An AI answer cites “City Accessibility Rule 2026, Section 14,” but the link does not work and you cannot find the rule. What would you do before relying on this claim?

Try it: practice the skill

What to do

1. Choose the task suggested for this module, or use another simple, low-risk task.
2. Open ChatGPT or another AI assistant in a separate tab or window.
3. Enter your prompt and send it. Do not include private or sensitive information.
4. Read the AI answer.
5. For this module, choose one factual statement in the answer that matters and check it with a suitable source or reliable method.
6. Return to this course and complete the short notes below.
   If you cannot use an AI assistant now, read the worked example and write what you would try later.