Google's secret detection method

Hey AI Enthusiast,

Happy New Year.

Google just launched video verification in the Gemini app.

You can now check if a video was created or edited with Google AI.

Upload a video. Ask: "Was this generated using Google AI?"

Gemini scans for SynthID watermarks embedded in audio and visual tracks. Tells you exactly which segments contain AI-generated elements.

Example: "SynthID detected within the audio between 10-20 secs. No SynthID detected in the visuals."

This is infrastructure for a world where most content is AI-generated.

And it's the perfect way to start 2026.

But first, today's prompt and a look back at what we covered together...

🔥 Prompt of the Day 🔥

AI Marketing 90-Day Roadmap: Use Claude or ChatGPT

Act as an AI transformation specialist. Create one 90-day implementation roadmap for adding AI to [MARKETING AREA] that delivers results fast.

Essential Details:

• Marketing Area: [CONTENT/ADS/EMAIL/SOCIAL/ANALYTICS]

• Current State: [WHAT YOU'RE DOING NOW]

• Goal State: [WHAT YOU WANT IN 90 DAYS]

• Budget: [TOTAL AMOUNT AVAILABLE]

• Team Skill Level: [BEGINNER/INTERMEDIATE/ADVANCED]

• Top Priority: [#1 THING TO ACHIEVE]

Create one roadmap including:

Days 1-30: Quick Wins

• Which AI tool to start with

• First 3 tasks to automate

• Team training basics

• Success metrics to track

Days 31-60: Build Systems

• Full tool integration

• Workflow changes needed

• Advanced training

• Quality checks

Days 61-90: Scale Up

• Optimize what's working

• Expand to more tasks

• Measure ROI

• Plan next phase

Also Include:

• Weekly team check-ins

• How to handle resistance

• Monthly progress reports

Get AI working in your marketing in 90 days.

✅ Thank You for 2025 ✅

2025 is done.

We covered 250+ major AI announcements together.

OpenAI GPT-5. Google's Gemini 3. Claude Sonnet 4.5. Anthropic's Claude 4 Models. Meta acquiring Manus for $2 billion. OpenAI buying Neptune.

We watched AI move from tools to infrastructure.

Google turned Chrome into a research assistant. OpenAI made ChatGPT an app platform. Salesforce embedded AI agents inside Slack. Notion agents automated 20-hour tasks in 20 minutes.

We saw AI video go from broken physics to Sora 2. AI image generation become 4x faster with ChatGPT Image 1.5. AI coding reach production reliability with Claude dominating benchmarks.

We tracked the consolidation. The acquisitions. The partnerships. Meta opening doors to competitors while racing Llama 5. Cognition acquiring AI coding startups. Grammarly acquiring AI email clients.

And we watched regulation start to matter. FTC demanding AI companion safety explanations. Privacy policies changing. The shift from "move fast" to "move carefully."

Every week. 250+ announcements. Zero hype.

Just what actually shipped.

And you were here for it.

Thank You

Thank you for opening these emails. For replying with your thoughts. For sharing what's working and what's not.

You don't just read about AI. You're building with it. Testing it. Breaking it. Figuring out what actually matters.

That's what separates this newsletter from the noise.

We're not here for demos. We're here for what ships. What works. What breaks. What costs. What scales.

2025 gave us 250+ moments worth tracking.

2026 starts today.

Did You Know?

Your smartphone's AI intentionally drains battery faster when you're near phone stores, using location data and usage patterns to encourage upgrades at optimal profit moments.

🗞️ Breaking AI News 🗞️

How SynthID Watermarking Actually Works

Google launched video verification in the Gemini app today.

You can now check if a video was created or edited with Google AI.

The Process

Upload a video to Gemini (up to 100 MB, 90 seconds long).

Ask: "Was this generated using Google AI?"

Gemini scans for SynthID watermarks embedded in both audio and visual tracks.

It returns a response with context and specifies which segments contain AI-generated elements.

Example response: "SynthID detected within the audio between 10-20 secs. No SynthID detected in the visuals."

What SynthID Is

SynthID is an imperceptible watermark.

It's embedded directly into AI-generated content during creation. Not added afterward.

Audio and visual tracks each get their own watermark.

The watermark survives:

• Compression

• Editing

• Format changes

• Minor modifications

It's designed to be undetectable to humans but readable by verification systems.

Why This Matters

Deepfakes are getting better. AI-generated content is everywhere.

Without verification, you can't trust what you see.

Courts need provenance for evidence. News organizations need to verify sources. Platforms need to label synthetic media.

Google's building the infrastructure for that verification.

This isn't just about Google AI. SynthID is part of the C2PA (Coalition for Content Provenance and Authenticity) standard.

Other companies are implementing similar systems:

• OpenAI with content credentials

• Adobe with Content Authenticity Initiative

• Microsoft with media provenance

By end of 2026, most major AI platforms will have verification built in.

The Limitations

SynthID only detects content created with Google AI.

If a video was made with OpenAI, Runway, or another tool, SynthID won't detect it.

This is platform-specific verification, not universal detection.

Also: SynthID can be removed with enough effort. It's not unbreakable.

But it raises the bar. Casual manipulation becomes harder.

What Happens Next

Verification becomes table stakes.

Every AI content platform will need detection and watermarking.

Platforms without it lose trust. Content without provenance gets flagged.

2026 is the year content verification becomes standard infrastructure.

The question shifts from "Can we detect AI content?" to "Why isn't this content verified?"

Over to You...

What was your biggest AI win in 2025?

Hit reply and share.

To shipping more AI in 2026,

Jeff J. Hunter 
Founder, AI Persona Method | TheTip.ai

NEW: “AI Money Group” to Learn to Make Money with AI