Compare convincing claim formats without mistaking confidence, prestige, or repetition for support.
Plan for this page
Turn an abstract idea into an observable decision.
You will leave with
Build a transferable source-checking rubric that separates the exact claim, source role, method, corroboration, scope, and uncertainty.
Time
18 min
Before you begin
For each decision, choose the action that best strengthens the claim-to-evidence chain. Feedback explains the decision pattern without ranking the learner.
Do this now
Predict your result before selecting the first option.
What should I trust when different AI claims look convincing?
For each decision, choose the action that best strengthens the claim-to-evidence chain. Feedback explains the decision pattern without ranking the learner.
Interactive model
Decision, consequence, result
Selected layerA provider says its new model is “twice as capable.” What do you inspect first?
The announcement links to one benchmark chart but does not define capable.
Read the complete text alternative
A provider says its new model is “twice as capable.” What do you inspect first?The announcement links to one benchmark chart but does not define capable.
A news article quotes a study saying AI raises productivity by 40%. What is the strongest next step?The article does not show the study population or the work measured.
Five posts repeat the same screenshot as proof of a policy change. How do you count the evidence?Every post links back to one anonymous account.
01
A provider says its new model is “twice as capable.” What do you inspect first?
The announcement links to one benchmark chart but does not define capable.
02
A news article quotes a study saying AI raises productivity by 40%. What is the strongest next step?
The article does not show the study population or the work measured.
03
Five posts repeat the same screenshot as proof of a policy change. How do you count the evidence?
Every post links back to one anonymous account.
Complete the decisions to reveal the model
Transfer artifact
My source-checking rubric
Write the four checks you will use before relying on an important AI claim in a new context.
Safety boundary
Use the fictional cases here for practice. For consequential health, legal, financial, safety, or public claims, consult current primary sources and qualified people.
Method note
This deterministic model rewards traceability and proportionate verification. It cannot determine whether a real claim is true and does not simulate a live research environment.
Text alternative
Read each source situation as a linear list. Compare the claim, origin, evidence, missing context, and verification action without relying on color or motion.
AI Risk Management FrameworkU.S. National Institute of Standards and Technology
Supports contextual mapping, measurement, management, monitoring, governance, accountability, and human oversight; it is voluntary guidance, not proof that a particular system is safe.
Stanford AI Index Report 2026Stanford Institute for Human-Centered AI
Provides dated evidence on AI capability, adoption, work, policy, and impacts; chapter methods and definitions must be checked before relying on individual claims.
inAi