Phase 2 · Question · Check · Decide

AI, Misinformation & Deepfakes

Learn what generative AI can and cannot tell you, how misleading media can be used, and ways to verify an urgent claim before acting.

Prefer to learn by doing?

Try the same lesson as a hands-on, step-by-step activity in a practice phone. It takes a few minutes and nothing is saved.

Open the interactive lesson
Start here

AI can sound sure and still be wrong.

Chatbots generate responses based on patterns and can make up facts, sources, or details. Images, audio, and video can also be edited or generated. There is no single visual clue that reliably proves media is fake or real. Check important claims with independent sources and use judgment.

Keep private details out of prompts: Do not enter passwords, financial or health information, private messages, identity documents, or details that identify another person into an AI tool. Check the tool’s privacy terms and settings before use; services handle data differently.

A practical verification routine

Treat a claim as a starting point to check

  1. Identify the specific claim and who is making it. An AI-generated answer is not itself a source.
  2. Check the date and look for original documents, named experts, or a relevant official announcement.
  3. Compare at least two independent, reputable sources. Similar-looking copies of one claim are not independent confirmation.
  4. For urgent requests, contact the person or organization using a known number or saved route—not a number or link in the message.
  5. If still uncertain, do not forward, pay, install, or share private information while deciding.
Four fictional case files

When convincing media or confident wording is not enough

1. A chatbot invents a local service

A chatbot confidently lists a phone number for a community grant office, but the office cannot be found in an official directory.

Why AI answers need checking

Why: Fluent language and confident tone do not show that a detail is accurate or current.

How it happens: A generative system may produce a plausible answer from learned patterns even when it lacks reliable information about the specific question.

What may be targeted: Time, trust, visits to incorrect pages, and potentially personal information if a fabricated contact route is followed.

Safer response: Do not rely on a chatbot for a phone number, deadline, legal or medical claim, or payment instruction. Verify through an official directory or organization reached independently.

2. A familiar voice asks for urgent money

A voicemail sounds like a relative and says they are in trouble, need secrecy, and want money sent immediately.

Why a familiar voice is not enough proof

Why: The request combines urgency, secrecy, and money. Audio can be edited or generated, and caller ID can be misleading.

How it is used: A scammer may use a public recording or a short clip to imitate a voice and prevent the listener from checking.

What may be targeted: Money, payment-account details, family information, and additional contacts to impersonate.

Safer response: Pause. Call the family member on a number you already know, or contact another trusted relative. Agree on a family check phrase if useful; do not use only the number or instructions in the urgent message.

3. A dramatic image appears with a breaking claim

A shared image supposedly shows a recent local emergency, but the caption gives no source, location, or date.

Check context instead of guessing from appearance

Why: Missing context means the image could be old, from another place, miscaptioned, edited, or generated. A visual inspection alone cannot settle it.

How it is used: Misleading images can provoke fear, anger, donations, clicks, or fast sharing before facts are known.

What may be targeted: Attention, trust, money, private details entered on linked pages, or public opinion.

Safer response: Look for a dated report from relevant local authorities or reputable news sources reached independently. Avoid resharing until the source and context are clear.

4. A chatbot asks for a private document

A learner is invited to upload an identity card and medical letter so an AI tool can “personalize” its advice.

Why convenience does not remove privacy risk

Why: These documents contain sensitive details that are not needed for general practice or ordinary questions.

How it is used: Depending on the service and settings, submitted information may be stored, reviewed, or processed under terms that differ between providers.

What may be targeted: Identity numbers, addresses, signatures, health details, account identifiers, and information about other people.

Safer response: Do not upload the documents. Ask a general question without identifying details, review the service’s privacy controls, and use a qualified professional or official service for sensitive matters.

Practice · Verify an urgent voice message

A familiar-sounding caller urgently asks for money and secrecy

What is the safest next step?
Take-away checklist

Pause before trusting or sharing AI-generated content

  1. Could this answer, image, audio, or video be wrong, old, edited, or missing context?
  2. Can I find an original and independent source for the claim?
  3. For urgent money or safety requests, have I verified through a number I already trust?
  4. Have I kept passwords, health, finance, and identity details out of the tool?
  5. If I cannot verify it, I can wait and choose not to forward or act.

Try it for real

These links open the real website in your own web browser, in a new tab. Nothing here asks you to sign in or buy anything. Do the interactive practice at the top first, then try these.

How to come back to this lesson

  1. Look for the tabs at the top of your browser, or the tabs button on a phone, and choose the Connected Pathways tab.
  2. Or close the new tab or page, and you will land right back here.
  3. If you get lost, type this address into your browser: https://connectedpathways.app/learn/phase2-ai-misinformation