1 / 18
←→ navigate N notes F fullscreen

Speaker Notes

AI Creators
How Influencers Are Building Empires with AI

Meet someone who turned AI tools into a career

Have You Noticed?

Your favorite content looks different now.

Guest Profile

Meet Chelsea Gong

"I sat down with Chelsea recently and asked: walk me through everything — the tools, the process, the numbers. And that's exactly what I'm going to share with you today."

Chelsea's Toolkit

Tried many. Kept three.

Tried & Moved On

Midjourney, HeyGen, ChatGPT, Gemini — tested in 2024, didn't fit the workflow

Claude AI

Text engine. Generates scripts, titles, keywords, and show notes from Chelsea's ideas

Nano Banana

Visual engine. Creates thumbnail art from keywords — the visual identity of the podcast

NotebookLM

Audio/Video engine. Produces podcast episodes and English-language summaries

Chelsea's Idea
→
Claude AI: Text + Titles + Keywords
→
Nano Banana: Thumbnail Art
→
NotebookLM: Audio/Video + English Summary
→
Upload & Publish

Daily updates. Three tools. One focused pipeline.

Interview Segment

"Why did it take so long to launch?"

I'm very demanding about quality, and my personal time is limited. I have to produce something I'm genuinely satisfied with in very little time — otherwise I simply won't ship. That's why I spent all of 2024 testing tools, but didn't launch Season 1 until mid-2025.

Aesthetics = Views

Chelsea's data: thumbnails make or break your content

  • Same podcast episode, different thumbnail:
  • Mediocre thumbnail: ~100 views
  • Beautiful, crafted thumbnail: ~5,000 views
  • That's a 50x difference from the visual alone
  • AI generates the image — but your taste determines what works

Whether a thumbnail looks good or not directly determines your views. Aesthetics aren't optional — they're the gateway.

Interview Segment

"How do you produce 5 episodes a week?"

The batching strategy: plan the season, execute on Sundays

Phase 1 — Season Planning

Not a professional creator — just stories she wants to tell. Plans 75+ episode long-form series in one creative session

Phase 2 — Sunday Execution

Every Sunday, 1 hour to produce the entire next week: 5 episodes, thumbnails, titles, everything

Season Brainstorm
→
Pick 5 Topics
→
Claude AI → Scripts
→
Nano Banana → Art
→
NotebookLM → Audio/Video
→
Upload All 5

Separate creative thinking from mechanical production. Plan the season once. Execute every Sunday. 1 hour → 5 episodes. That's the power of batching + AI.

Chelsea's Pipeline in Detail

From idea to published episode — how long each step actually takes

Choose Topic
~30 sec (pre-planned)
→
Claude AI: Script + Titles
~30 sec
→
Nano Banana: Thumbnail
~2–3 min
→
NotebookLM: Audio/Video
~5 min
→
Review & Upload
~2 min
■ Human decision (taste & judgment) ■ AI execution (generation & conversion)

Audio Episode

~10 minutes total from idea to published

Video Episode

~12 minutes total from idea to published

Weekly Batch (Sunday)

1 hour → 5 episodes

Interview Segment

"I do zero editing."

0 post-production. 0 trimming. Straight from AI to upload.

Zero editing, zero post-production. NotebookLM's output goes directly to the platform. Traditional podcast post-production takes at least 3 hours per episode — especially editing and audio quality. I refuse.

The Experiment

2024: Tested many tools. Most failed on usability and quality — couldn't meet her standards

The Decision

Traditional production: 6 hours per episode. AI production: 10 minutes. Zero editing, zero compromise.

The trick: invest in input quality (great scripts, sharp topics) so the AI output needs no fixing. Sometimes the best optimization is removing a step entirely.

Interview Segment

"How did you grow so fast?"

Product-Market Fit — the same content, wildly different results

小宇宙 (Xiaoyuzhou)

10,000

subscribers in 6 months
Zero paid promotion
100% organic growth via algorithm + word of mouth

Spotify

30

subscribers over the same period
Same content, same quality
Wrong audience, wrong algorithm

The same product performs 300x differently depending on where you put it. Platform selection isn't a distribution detail — it's a strategic decision that determines whether your product finds its audience at all.

Why Platform Fit Matters More Than Tools

The universal lesson from Chelsea's 300x gap

The Right Questions

Where does your target audience actually hang out? What format do they prefer? What does the algorithm reward?

The Startup Principle

Distribution before product. The best content in the world fails if nobody sees it.

These platform questions matter 100x more than which AI tool you use. Tool selection is a detail. Platform selection is a strategy.

What's Next

Cross-Platform Automation

One creative input → 5+ platform-native outputs

One Script
→

小宇宙

Chinese audio podcast — current stronghold

Spotify / Apple

English audio via NotebookLM translation

Douyin / TikTok

Short video clips auto-extracted from episodes

YouTube / Bilibili

Long-form video with AI-generated visuals

Blog / Newsletter

Written version auto-generated from script

The future competitive advantage isn't knowing one platform — it's knowing how to be everywhere simultaneously. One creative decision, automated everywhere.

What's Next

The Efficiency Frontier

When production cost → 0, taste becomes the only bottleneck

2023

Write + record + edit + publish
4–8 hours per episode

2025 (Chelsea)

Script + generate + upload
10–12 minutes per episode

2026–2027?

Choose topic + approve output
~2 minutes per episode?

The Economics

Chelsea's 4 monetization models for AI-powered podcasting

1. Digital Goods

Templates, prompt packs, curated resource lists — low effort, high margin with AI assistance

2. Platform Ads

小宇宙's built-in ad program (similar to YouTube AdSense) — revenue shares once you hit subscriber thresholds

3. Sponsorships

Brands approach you when audience is large enough — at 10K subscribers, Chelsea is entering this territory

4. Audience Tipping

Listeners send tips during/after episodes — common on Chinese podcast platforms

When your production cost is 10 minutes of time and $0 in tools (free tiers), even small revenue per episode is profitable. AI didn't just change how content is made — it changed the economics of who can afford to make it.

Your Turn: The Student Creator

How to Start

  1. Pick a niche you actually care about
  2. Master 2–3 AI tools deeply — depth beats breadth
  3. Develop your taste & voice first — then use AI to amplify
  4. Start small: one platform, consistent schedule
  5. Build one thing per week. Show it to someone. Get feedback.

Chelsea's Recommended Stack

NotebookLM (Free) — One-Stop Solution

Chelsea's #1 recommendation. Audio & video podcast generation, English summaries, all from one tool. Start here.

Claude AI (Free Tier)

Scripts, titles, keywords — high quality text that needs zero editing

Nano Banana

Keyword-driven thumbnail art — beautiful visuals in 2–3 minutes

The Ethics Check

The EARTH Framework — responsible AI content creation

  • Ethical — Does your content treat people fairly? Are you transparent about AI use?
  • Adaptive — Are you learning from audience feedback, or just optimizing for clicks?
  • Responsible — Do you own the consequences of what you publish, even if AI made it?
  • Transparent — Can people tell what's AI-generated? Should they be able to?
  • Harmonious — Are you contributing positively to the ecosystem?

Your audience can usually tell when something is fake. Authenticity is security. And the same tools that let you create amazing content let bad actors create convincing deepfakes. Use the power responsibly.

Key Takeaways

  1. AI is a force multiplier, not a talent replacement — it amplifies what you're already good at
  2. The feedback loop is everything — Create → Measure → Adjust → Repeat
  3. Authenticity is your moat — AI can't replicate genuine human connection
  4. Depth beats breadth — master 2–3 tools, not 20
  5. Start NOW — the learning curve gets steeper every month
  6. Ethics aren't optional — think about your impact before you post

Questions & Discussion

"Would you follow an AI-generated influencer?
Why or why not?"

"If you could use AI to build any kind of content, what would it be?"

"What's one AI tool you're going to try this week?"