Deepfake detection workshop & training kit

A free, ready-to-run 45-minute workshop that teaches any team how to spot AI-generated images, cloned voices, deepfake video and edited documents — and what to do before acting on a file. Use it for staff training, classrooms, newsroom onboarding or creative studios.

Agenda (45 minutes)

1. 0–10 min · The synthetic media landscape. How image, video and voice generators work, and the real scams they power: fake invoices, CEO voice calls, romance profiles, staged insurance photos.
2. 10–20 min · Passive forensics. What you can see without the creator's help: lighting and geometry, hands and text, error level analysis, camera metadata and sensor noise.
3. 20–28 min · Provenance. C2PA Content Credentials, watermarks like Google SynthID, and why "no watermark" doesn't mean "real".
4. 28–40 min · Hands-on lab. Everyone takes the Spot the AI quiz, then checks a sample file with the free checker and reads the heatmap together.
5. 40–45 min · Your team's rule. Agree one verification rule for your work (see checklist below).

Slide outline for facilitators

1. Title: "Can you still trust what you see and hear?"
2. Three real-world scam patterns relevant to your audience.
3. How generators make images, video and voices (one slide each).
4. Seven visual tells — see how to spot AI images.
5. Voice clones and the call-back rule — see detecting voice clones.
6. Deepfake video tells — see Sora and Kling guide.
7. Provenance vs detection — see SynthID vs forensics.
8. Live demo: upload a file, read the verdict and heatmap.
9. Quiz results and discussion.
10. Your team checklist.

Hands-on exercises

• Image lab: Compare a camera photo with a generated portrait. List three differences before running a check.
• Voice lab: Play a short synthetic voice sample. Discuss what would make you call back on a known number.
• Document lab: Review an invoice with changed bank details. Follow the steps in detecting fake invoices.
• Role-play: A "manager" requests an urgent payment by voice note. The team must verify before acting.

Team verification checklist

1. Pause when a file creates urgency, money movement or reputational risk.
2. Check the source: who sent it, through which channel, is that normal?
3. Run a forensic check and save the result.
4. Confirm through a second, known channel (call back, in person).
5. Keep the original file and its fingerprint (SHA-256) if it may become evidence.
6. Escalate amber or red results to a named person.

Variants by audience

• Finance & HR: invoices, payslips, voice-note payment requests, fake candidates.
• Newsrooms: verifying user-submitted video and images before publishing.
• Creative studios: proving your own work is human-made — see VerifyAI for creative agencies.
• Schools & universities: media literacy, and handling AI-writing accusations fairly without relying on a single detector score.

FAQ

Want a facilitated session for your team? Email support@goverifyai.com.