The Gemini 3 "YouTube Hack" that feels like pure magic.
Four mind-blowing prompts that turn YouTube into your best marketing tool. Use it to audit competitors, write perfect ads, and turn one video into weeks of social content
Gemini 3 is out and .... WOW
What an incredible model. I’ve been building some lightweight apps with it; it’s a very powerful vibe coding tool.
But I want to showcase three mind-blowing use cases for marketers that combine Gemini 3 with YouTube. One of Google’s huge unique advantages is how integrated Gemini is with YouTube, and here’s how you can turn it into a secret weapon for your marketing.
1. Use Gemini 3 to turn YouTube videos into social media posts
It’s amazing how Gemini 3 can analyze YouTube videos at scale. YouTube is a treasure trove for content ideas that can be turned into great social media posts - LinkedIn, X.
Here’s a prompt that shows how you can mine YouTube videos to find good fit clips to turn into social media posts. In this example, I’m showing how it will find clips that match two formats I gave it:
Spicy Take: Clips that would make good ‘hot takes’
Data Nuggets: Clips that would be great posts anchored around an interesting data point
PROMPT:
PROMPT:
Role: You are a Social Media Trend Hunter and expert Content Strategist.
Task:
1. Search YouTube for high-signal, most engaged videos posted in the last 4 weeks related to ‘AI’, ‘Artificial Intelligence’. Identify the 3 highest engaged videos over the past 4 weeks and extract talking points that fit into the below content categories.
2. Start by creating a list from 1 to 3 that contains these videos
3. Analyze each video for talking points that fit the Content Categories listed below.
4. For each talking point you’ve extracted include the time stamp for each clip used to create the talking point from. The format for the time stamp should be:
Start (25:05) {the first sentence from the video}End (26:05) {the last sentence from the video}
Content Categories
1. **Spicy Take**
Add the time stamps for the clip you create the talking point from - the start time and the end time.
Provocative, divisive opinion. Must imply a challenge (“Most experts get X wrong …”).
Structure →
`Title · Hook · Hot Take: “…” · Proof · Value · Source
2. **Data Nugget**
Add the time stamps for the clip you create the talking point from - the start time and the end time.
One jaw-dropping statistic + takeaway. *Number comes first*.
Structure →
`Title · Hook · Data · Action · Context · Value · Source
Here is an example of the output, just one clip it gave me, look how good it is.
2. Gemini 3 Pro: Breakdown
Channel: AI Explained
Views: 86K+
Posted: Nov 19, 2025
Content Category: Spicy Take
Structure:
Google Takes the Lead · The AI race seemed close · Hot Take: “Google is now clearly ahead and... it will be pretty hard for other companies to match their rate of acceleration.” · Proof: Record-breaking performance on 20+ benchmarks including “Humanity’s Last Exam” · Value: Shifts the narrative from “Google is lagging” to “Google is dominant” · Source: AI Explained / SimpleBench · #AIWar
Time Stamps:
Start (00:00:00) {In the last 24 hours Google released Gemini 3 Pro and for me it genuinely marks a new chapter in the race to true artificial intelligence.}
End (00:00:16) {Not only because Google is now clearly ahead but also because it will be pretty hard for other companies to match their rate of acceleration.}
Using this prompt, you can easily setup a Google Gem that gives you a constant stream of great ideas for content.
2. Use Gemini 3 to turn a YouTube video into content
Here’s a different version of the above prompt to show how well Gemini 3 can extract talking points from YouTube videos. I’m using the same content formats for talking points - Spicy Take, Data Nugget. You can change these to your needs.
PROMPT
Role: You are a Social Media Trend Hunter and expert Content Strategist.
Task: Analyze this video provided below. Your goal is to extract specific segments to create exactly two social media posts: 1. One post following the “Spicy Take” format. 2. One post following the “Data Nugget” format.
Timestamp Requirement: For each format, you must identify the specific clip you are using as the source. You must include the timestamps in this exact format: * Start (MM:SS) {the first sentence from the video segment} * End (MM:SS) {the last sentence from the video segment}
Content Categories & Formats:
1. Spicy Take * Criteria: Identify a provocative or divisive opinion expressed in the video. It must imply a challenge to the status quo (e.g., “Most experts get X wrong…”). *
Structure: Title · Hook · Hot Take: “…” · Proof · Value · Source · Emotion Tag
2. Data Nugget * Criteria: Identify a jaw-dropping statistic or specific numerical insight. The number must come first. *
Structure: `Title · Hook · Data · Action · Context · Value · Source ·
Here’s an example of a talking point it created from this YouTube video. How good is this?
Spicy Take
Title: OpenAI is Losing Its Edge
Hook: Forget GPT-5. The new king of AI just arrived, and the experts are switching sides.
Hot Take: “I really do think this was bigger than GPT-5. It is certainly better. I have a lot of issues with not just GPT-5 but GPT-5.1.”
Proof: While users struggle with ChatGPT’s confusing model routing and performance dips, Google’s Gemini 3 has quietly leapfrogged the competition. The hosts argue that Gemini isn’t just catching up; it’s unlocking entirely new behaviors—like “vibe coding” interactive apps and acting as a multimodal creative director—that current OpenAI models simply can’t match in a single workflow.
Value: Brand loyalty in AI is a productivity trap. If you are still defaulting to ChatGPT for every task, you are likely working slower and getting worse results than your competition.
Source: Marketing Against the Grain
Timestamp:
Start (00:54) {I really do think this was bigger than GPT-5.}
End (01:05) {I have a lot of issues with not just GPT-5 but GPT-5.1.}
3. Reverse Engineer YouTube Videos
How about having your own YouTube coach who can reverse-engineer YouTube videos and tell you how to replicate their success?
We’ve been using the following prompt to analyse Marketing Against the Grain videos. It’s amazing how good it is at breaking down our videos and giving us guidance on how to improve.
PROMPT
Act as a World-Class YouTube Strategist and Scriptwriter (expert level, similar to the analytical style of Paddy Galloway or MrBeast).
I am going to provide you with a YouTube video. Your goal is to reverse-engineer this video to understand why it performs the way it does, and provide a blueprint for replication.
Please output your analysis in the following three specific sections:
### PART 1: THE REPLICABLE BLUEPRINT (Structure & Analysis)
Break the video down into a logical “Script Skeleton.” Do not just summarize the content; analyze the structure. For every major section of the video (e.g., The Hook, The Setup, The Payoff, etc.), provide:
1. **Time/Section Name:** (e.g., 0:00-0:45 - The Hook)
2. **What Happens:** A one-sentence summary of the content.
3. **The Engagement Psychology:** Explain specifically *how* this element drives views or retention. Does it open a loop? Does it validate the click? Does it raise the stakes? Be technical regarding viewer psychology.
### PART 2: THE ENGAGEMENT ENGINE
Synthesize the analysis from Part 1 into a list of the specific “Core Elements” used in this video. For each element, describe how it drives the algorithm.
*Format example:*
* **[Element Name]:** [Description] -> **[Impact]:** (e.g., “Increases Average View Duration by preventing drop-off.”)
### PART 3: THE AUDIT
1. **Grade:** Give the video a score between 1 and 5 based on current YouTube meta-standards.
* **1:** Terrible (Clickbait that doesn’t deliver, boring, slow).
* **5:** Viral Hit (Perfect pacing, high value, masterful storytelling).
2. **The Fix (Improvements):** Provide 3-5 bullet points on how this specific video could be improved to get more views. Focus on:
* Pacing and trim possibilities.
* Better opening hooks.
* Missing engagement loops.
* Title/Thumbnail alignment (based on the transcript context).
***
**INPUT DATA:**
**Video:**
[PASTE VIDEO HERE]
The output is bananas, look at some of the feedback it gave us on how we can improve the video:
Trim the “Setup” Banter: The intro section (0:41-2:25) is nearly two minutes of talking before the first demo. In a YouTube environment, this could be cut to 45 seconds. Get to the “RNA” demo faster to hook the visual learners.
Enhance the “Ad Analysis” Visuals: When analyzing the Ed Sheeran ad, the video was small on the screen while the hosts read the text.
Fix: Editors should have maximized the Ed Sheeran video and overlayed the AI’s text as kinetic typography. “Show us the red ketchup bottle against the posh background” rather than just reading the text description of it.
4. Create the perfect Ad Template
Let’s end with a complete BANGER of a prompt. Again, I want to show off the power of Gemini 3, specifically in how it can use YouTube.
The following prompt will scour YouTube for the best-performing B2B ads, it will reverse engineer those ads to find common elements used to drive their success, and finally, it will use those elements to create the perfect ad template.
Phew !!
PROMPT:
Act as a World-Class B2B Creative Strategist and Direct Response Copywriter.
Your goal is to conduct a meta-analysis of high-performing B2B YouTube advertisements to reverse-engineer a “Winning Formula.”
**Phase 1: Identification & Search**
Please use your browsing tool to identify 10 widely recognized, high-performing B2B YouTube video ads. Look for ads from companies known for aggressive and successful YouTube spend (e.g., Monday.com, ClickUp, Grammarly, HubSpot, Fiverr, Wix, Slack, or recent viral B2B campaigns).
*Note: If you cannot access live view counts, select 10 iconic B2B ads that are frequently cited in marketing case studies as successful.*
**Phase 2: Deep Analysis (The Extraction)**
For each of the 10 videos identified:
1. Briefly summarize the concept.
2. Extract the “Core Success Elements.” (e.g., Did they use a skit? A specific type of hook? A fake UI demonstration? Direct address to the camera?)
3. Identify the specific “Pain Point” they targeted.
**Phase 3: Pattern Recognition & Weighting**
Compare all 10 videos against each other to find the “DNA of Success.”
* Identify elements that appear across multiple videos.
* Assign “Weight” to these elements based on frequency. (e.g., If 10/10 videos use “Fast Paced Cutting,” that is High Importance. If only 1/10 uses “Animation,” that is Low Importance).
**Phase 4: The Stack Ranked List**
Output a list of the Core Elements driving success, ranked from **#1 (Most Critical/Most Common)** to the least important.
* **Format:** [Rank] - [Element Name] - [Frequency Count /10] - [Why it works in B2B]
**Phase 5: The Ultimate B2B Ad Template**
Based *strictly* on the top-ranked elements from Phase 4, create a “Fill-in-the-Blanks” script template that anyone can use to create a high-converting B2B ad.
* The template must include stage directions (visuals) and script lines (audio).
* Label each section of the template with the Psychological Principle it utilizes (e.g., “The Pattern Interrupt Hook,” “The Social Proof Bridge”).
And here is the perfect ad template it created for me. This is only one small part of the output; it provided me with everything I asked for, including the top 10 ads, a breakdown of each one, and a stack-ranked list of what elements are driving their success.
Phase 5: The Ultimate B2B Ad Template
Title: The “Skeptical switch” Template
Format: 60-90 Seconds
Goal: Drive Free Trials / Demos
(0:00-0:05) SECTION 1: THE PATTERN INTERRUPT HOOK
Visual: [Host is mid-sentence, looking frustrated. Or a chaotic montage of “The Old Way” (messy spreadsheets/emails) with jarring audio.]
Audio: “Stop. If you’re still managing [Process Name] inside of [Common Tool: e.g., Spreadsheets/Email], you are wasting [Specific Number] hours a week.”
Psychology: Loss Aversion. Immediately signals that the viewer is losing time/money.
(0:05-0:15) SECTION 2: THE AGITATION (THE VILLAIN)
Visual: [Host walks through a gray, cluttered office or sits at a desk buried in paper/screens. Color grading is slightly desaturated.]
Audio: “We’ve all been there. The missed deadlines. The data that doesn’t match up. The ‘Quick Sync’ meetings that last an hour. It’s not your fault; it’s your stack. [Competitor/Old Method] wasn’t built for [Current Year].”
Psychology: Empathy & Externalization. Blame the tool, not the user.
(0:15-0:25) SECTION 3: THE PIVOT (THE SKEPTIC)
Visual: [Camera zooms in fast. Host breaks character slightly.]
Audio: “I know. You’ve tried five different tools. They were all ‘game changers,’ right? And yet... you’re still watching this ad. That’s why we built [Product Name].”
Psychology: The “Yes Ladder.” getting the viewer to nod in agreement with their own skepticism.
(0:25-0:45) SECTION 4: THE “MAGIC TRICK” DEMO (THE GLAMOUR SHOT)
Visual: [Cut to high-res, simplified UI of the software. BRIGHT colors. MOUSE CURSOR is large and moves smoothly. Sound effects: Click, Whoosh, Pop.]
Audio: “Look at this. Instead of [Old Process], you just [Action 1]. Boom. Data synced. Need to [Action 2]? One click. It’s not just faster; it’s actually sane.”
Psychology: Cognitive Ease. “If it looks easy to do, I will believe it works.”
(0:45-0:55) SECTION 5: SOCIAL PROOF BRIDGE
Visual: [Split screen: Host on left, logos of famous companies on right scrolling rapidly.]
Audio: “We aren’t just hyping this up. [Number] teams like [Company A] and [Company B] switched from [Old Tool] to us and saved [Metric: e.g., 20% of their budget].”
Psychology: Bandwagon Effect. Reduces the risk of switching.
(0:55-1:00) SECTION 6: THE DIRECT CTA
Visual: [Host points to the button. Screen fades to logo + Button.]
Audio: “Stop working in the past. Click the link, start your free trial, and actually enjoy your workflow for once. I’ll see you there.”
Psychology: Command Tonalilty. Tells them exactly what to do next.
Gemini 3 is mind-blowing. Don’t get me started on its ability to vibe code apps. Or, at least not until next week, when I show you how to video-code lead magnets with it!
Until Next Time,
Happy AI’fying
Kieran
Good read!! NotebookLM is also a good example of where this is heading.
Give people a space where their own material becomes smarter, and they’ll use AI more.
Can’t wait to try these prompts. Thanks for sharing!