Everyday AI5–10 minutes to learn

Research With AI Without Getting Misled

Use AI to research, not to guess.

Gives you a prompt for your own AI tool

WHAT YOU'LL ACHIEVE

Use AI to structure a research task, explore relevant questions, evaluate sources and produce a useful summary without treating AI-generated claims as automatically reliable.

Read the guidance, then copy the ready-made prompt to use with your AI tool.

FIELD NOTE / RESEARCH METHOD

THE TOOL: REUSABLE RESEARCH PROMPT

Give the question a structure.

Help me research the following topic:
[INSERT TOPIC OR QUESTION]

Identify the main questions I should investigate.
Explain the important concepts in plain English.
Suggest primary or authoritative sources and provide the exact title, publisher, date and direct link when available.
For each important factual claim, explain which source supports it and what the source actually says.
Do not invent citations, quotations, dates or links. Say when you cannot verify a source or claim.
Clearly distinguish established facts, interpretations and uncertainties.

A SHORT SOURCE-VERIFICATION METHOD

Check the claim, not just the citation.

Choose one specific, important claim and ask the AI for the source that directly supports it.
Open the original source yourself. Check the author or publisher, publication date, method and relevant passage—not just the AI's summary.
Compare the source with the exact claim: figures, population, time period, scope and caveats. Check whether it reports a measured result or only a forecast or opinion.
Look for an independent primary or authoritative source. Multiple pages repeating the same original claim are not independent confirmation.
Keep a link and the passage you checked. If the source does not support the claim or you cannot verify it, label it uncertain or leave it out.

WORKED EXAMPLE

An AI claims onboarding is “40% faster.”

After copying the prompt, add that sentence as the topic and ask for the original study or report behind the figure. Then open the cited source and check:

Does the source actually report 40%, and is it measuring onboarding time rather than a different outcome?
Which organisation, roles and tasks were included, and what time period and baseline were used?
Was the figure measured in a real deployment, estimated by a survey, or projected by a vendor? What limitations does the source state?
Can an independent source confirm the result, or is every citation repeating the same report?

If the evidence supports only one organisation's estimate, rewrite the claim with that limitation instead of presenting “40% faster” as a general fact. If you cannot find the original evidence, do not rely on the figure.