StarZero All shows Try the podcast clipper →

Apr 2, 2025 · 42m · startup-ideas

ChatGPT 4o replaces your $10,000 Ad Agency (Full Tutorial)

Jacob Pozel · 24m spoken Greg Isenberg · 10m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this masterclass tutorial, host Greg Isenberg and digital marketer Jacob Pozel demonstrate how to utilize ChatGPT 4o's multimodal capabilities to rapidly design high-converting, professional visual advertisements. Through structured prompt breakdowns, case studies, and a live campaign build, the video provides actionable workflows for scaling digital marketing creative in minutes.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 29.8% of the talking time here. How this is scored →

Greg as informed peer 4.3 Guest teaching 5.3 Guest disagreement 0.4 Greg pushing back 1.6
05100:0015:0030:001:03–5:53 · Greg as informed peer 5/10 Episode Roadmap and Framework Overview Jacob lays out the multimodal framework of ChatGPT-4o and how image references replace strict prompting. Greg adds value by connecting the tool to performance marketing principles like multivariate persona testing and ROAS optimization.5:54–9:07 · Greg as informed peer 2/10 Text-Only Prompting and Internal Brand Knowledge Jacob schools Greg on why pretraining recognizes major brands like Nike or Ridge Wallet natively while smaller brands require reference images. Greg openly acknowledges he did not know this distinction.9:11–13:31 · Greg as informed peer 4/10 Prompt Discovery and Chat Reset Workflow Jacob demonstrates prompt discovery via Sora's explore feed and explains the critical technique of opening fresh chats to prevent hallucination loops. Greg relates this workflow to narrowing time-to-value for non-technical creators.13:34–16:19 · Greg as informed peer 4/10 Ad Template Remixing and Performance Ethics Greg challenges the ethics of cloning competitor ad layouts, asking if ripping high-spend designs feels dirty. Jacob pushes back gently, framing it as the standard industry workflow that agencies already practice via templates.16:21–20:01 · Greg as informed peer 5/10 Seasonal Adaptations, Newsjacking, and Ad Research Tools The discussion covers seasonal variations, rapid newsjacking, and ad research tools like Creative OS and Foreplay. Greg contributes actionable marketing angles like micro-holidays and birthday triggers.20:01–30:47 · Greg as informed peer 6/10 Live Demonstration: Conceptualizing the LCA Campaign Greg intervenes when Jacob suggests a newspaper layout for LCA, asserting firm brand ownership by redefining the agency as an elite luxury SWAT team fitting a vintage Rolex aesthetic. Jacob adapts the prompt accordingly.30:48–41:25 · Greg as informed peer 4/10 Live Demonstration: Iterating in a Fresh Chat Jacob explains model temperature and iterative prompting in a new conversation window after the initial output generated an unwanted Rolex watch. Greg notices a hallucinated client logo (Disney) during the output review.1:03–5:53 · Guest teaching 5/10 Episode Roadmap and Framework Overview Jacob lays out the multimodal framework of ChatGPT-4o and how image references replace strict prompting. Greg adds value by connecting the tool to performance marketing principles like multivariate persona testing and ROAS optimization.5:54–9:07 · Guest teaching 7/10 Text-Only Prompting and Internal Brand Knowledge Jacob schools Greg on why pretraining recognizes major brands like Nike or Ridge Wallet natively while smaller brands require reference images. Greg openly acknowledges he did not know this distinction.9:11–13:31 · Guest teaching 6/10 Prompt Discovery and Chat Reset Workflow Jacob demonstrates prompt discovery via Sora's explore feed and explains the critical technique of opening fresh chats to prevent hallucination loops. Greg relates this workflow to narrowing time-to-value for non-technical creators.13:34–16:19 · Guest teaching 5/10 Ad Template Remixing and Performance Ethics Greg challenges the ethics of cloning competitor ad layouts, asking if ripping high-spend designs feels dirty. Jacob pushes back gently, framing it as the standard industry workflow that agencies already practice via templates.16:21–20:01 · Guest teaching 5/10 Seasonal Adaptations, Newsjacking, and Ad Research Tools The discussion covers seasonal variations, rapid newsjacking, and ad research tools like Creative OS and Foreplay. Greg contributes actionable marketing angles like micro-holidays and birthday triggers.20:01–30:47 · Guest teaching 4/10 Live Demonstration: Conceptualizing the LCA Campaign Greg intervenes when Jacob suggests a newspaper layout for LCA, asserting firm brand ownership by redefining the agency as an elite luxury SWAT team fitting a vintage Rolex aesthetic. Jacob adapts the prompt accordingly.30:48–41:25 · Guest teaching 5/10 Live Demonstration: Iterating in a Fresh Chat Jacob explains model temperature and iterative prompting in a new conversation window after the initial output generated an unwanted Rolex watch. Greg notices a hallucinated client logo (Disney) during the output review.1:03–5:53 · Guest disagreement 0/10 Episode Roadmap and Framework Overview Jacob lays out the multimodal framework of ChatGPT-4o and how image references replace strict prompting. Greg adds value by connecting the tool to performance marketing principles like multivariate persona testing and ROAS optimization.5:54–9:07 · Guest disagreement 0/10 Text-Only Prompting and Internal Brand Knowledge Jacob schools Greg on why pretraining recognizes major brands like Nike or Ridge Wallet natively while smaller brands require reference images. Greg openly acknowledges he did not know this distinction.9:11–13:31 · Guest disagreement 0/10 Prompt Discovery and Chat Reset Workflow Jacob demonstrates prompt discovery via Sora's explore feed and explains the critical technique of opening fresh chats to prevent hallucination loops. Greg relates this workflow to narrowing time-to-value for non-technical creators.13:34–16:19 · Guest disagreement 2/10 Ad Template Remixing and Performance Ethics Greg challenges the ethics of cloning competitor ad layouts, asking if ripping high-spend designs feels dirty. Jacob pushes back gently, framing it as the standard industry workflow that agencies already practice via templates.16:21–20:01 · Guest disagreement 0/10 Seasonal Adaptations, Newsjacking, and Ad Research Tools The discussion covers seasonal variations, rapid newsjacking, and ad research tools like Creative OS and Foreplay. Greg contributes actionable marketing angles like micro-holidays and birthday triggers.20:01–30:47 · Guest disagreement 1/10 Live Demonstration: Conceptualizing the LCA Campaign Greg intervenes when Jacob suggests a newspaper layout for LCA, asserting firm brand ownership by redefining the agency as an elite luxury SWAT team fitting a vintage Rolex aesthetic. Jacob adapts the prompt accordingly.30:48–41:25 · Guest disagreement 0/10 Live Demonstration: Iterating in a Fresh Chat Jacob explains model temperature and iterative prompting in a new conversation window after the initial output generated an unwanted Rolex watch. Greg notices a hallucinated client logo (Disney) during the output review.1:03–5:53 · Greg pushing back 0/10 Episode Roadmap and Framework Overview Jacob lays out the multimodal framework of ChatGPT-4o and how image references replace strict prompting. Greg adds value by connecting the tool to performance marketing principles like multivariate persona testing and ROAS optimization.5:54–9:07 · Greg pushing back 0/10 Text-Only Prompting and Internal Brand Knowledge Jacob schools Greg on why pretraining recognizes major brands like Nike or Ridge Wallet natively while smaller brands require reference images. Greg openly acknowledges he did not know this distinction.9:11–13:31 · Greg pushing back 1/10 Prompt Discovery and Chat Reset Workflow Jacob demonstrates prompt discovery via Sora's explore feed and explains the critical technique of opening fresh chats to prevent hallucination loops. Greg relates this workflow to narrowing time-to-value for non-technical creators.13:34–16:19 · Greg pushing back 5/10 Ad Template Remixing and Performance Ethics Greg challenges the ethics of cloning competitor ad layouts, asking if ripping high-spend designs feels dirty. Jacob pushes back gently, framing it as the standard industry workflow that agencies already practice via templates.16:21–20:01 · Greg pushing back 0/10 Seasonal Adaptations, Newsjacking, and Ad Research Tools The discussion covers seasonal variations, rapid newsjacking, and ad research tools like Creative OS and Foreplay. Greg contributes actionable marketing angles like micro-holidays and birthday triggers.20:01–30:47 · Greg pushing back 4/10 Live Demonstration: Conceptualizing the LCA Campaign Greg intervenes when Jacob suggests a newspaper layout for LCA, asserting firm brand ownership by redefining the agency as an elite luxury SWAT team fitting a vintage Rolex aesthetic. Jacob adapts the prompt accordingly.30:48–41:25 · Greg pushing back 1/10 Live Demonstration: Iterating in a Fresh Chat Jacob explains model temperature and iterative prompting in a new conversation window after the initial output generated an unwanted Rolex watch. Greg notices a hallucinated client logo (Disney) during the output review.

speaking balance: gold is Greg, purple is the guest (3 minute bins)

0:00 · Greg 57.8% · guest 42.2%0:00 · Greg 57.8% · guest 42.2%3:00 · Greg 28.8% · guest 71.2%3:00 · Greg 28.8% · guest 71.2%6:00 · Greg 6.2% · guest 93.8%6:00 · Greg 6.2% · guest 93.8%9:00 · Greg 30% · guest 70%9:00 · Greg 30% · guest 70%12:00 · Greg 19.8% · guest 80.2%12:00 · Greg 19.8% · guest 80.2%15:00 · Greg 41% · guest 59%15:00 · Greg 41% · guest 59%18:00 · Greg 45.8% · guest 54.2%18:00 · Greg 45.8% · guest 54.2%21:00 · Greg 30.8% · guest 69.2%21:00 · Greg 30.8% · guest 69.2%24:00 · Greg 30.4% · guest 69.6%24:00 · Greg 30.4% · guest 69.6%27:00 · Greg 14.6% · guest 85.4%27:00 · Greg 14.6% · guest 85.4%30:00 · Greg 17% · guest 83%30:00 · Greg 17% · guest 83%33:00 · Greg 21.8% · guest 78.2%33:00 · Greg 21.8% · guest 78.2%36:00 · Greg 26.1% · guest 73.9%36:00 · Greg 26.1% · guest 73.9%39:00 · Greg 41.3% · guest 58.7%39:00 · Greg 41.3% · guest 58.7%42:00 · Greg 93.7% · guest 6.3%42:00 · Greg 93.7% · guest 6.3%
Sharpest disagreement ▶ 15:25 Jacob defends ad cloning practices

Jacob counters Greg's moral hesitation about copying competitor ads by noting that entire creative agencies and templating platforms already operate this way routinely.

 Hardest push from Greg ▶ 14:49 Greg questions the ethics of copying ads

Greg directly challenges the moral cleanliness of taking another company's expensive ad creative and remixing it in seconds with AI.

 Biggest teaching moment ▶ 6:14 Brand recognition training data schooling

Jacob explains how foundational models store concepts for high-spend brands versus low-visibility products, providing technical insight Greg admits he didn't know.

 Greg holds their own ▶ 22:58 Greg steers LCA's luxury brand positioning

Greg takes firm creative control, rejecting Jacob's newspaper layout proposal and articulating LCA's high-end executive positioning using a luxury watch comparison.
the scores for every segment, with the reasoning behind each
ChapterTopicGreg as informed peerGuest teachingGuest disagreementGreg pushing backWhy
Episode Roadmap and Framework Overview 5500 Jacob lays out the multimodal framework of ChatGPT-4o and how image references replace strict prompting. Greg adds value by connecting the tool to performance marketing principles like multivariate persona testing and ROAS optimization.
Text-Only Prompting and Internal Brand Knowledge 2700 Jacob schools Greg on why pretraining recognizes major brands like Nike or Ridge Wallet natively while smaller brands require reference images. Greg openly acknowledges he did not know this distinction.
Prompt Discovery and Chat Reset Workflow 4601 Jacob demonstrates prompt discovery via Sora's explore feed and explains the critical technique of opening fresh chats to prevent hallucination loops. Greg relates this workflow to narrowing time-to-value for non-technical creators.
Ad Template Remixing and Performance Ethics 4525 Greg challenges the ethics of cloning competitor ad layouts, asking if ripping high-spend designs feels dirty. Jacob pushes back gently, framing it as the standard industry workflow that agencies already practice via templates.
Seasonal Adaptations, Newsjacking, and Ad Research Tools 5500 The discussion covers seasonal variations, rapid newsjacking, and ad research tools like Creative OS and Foreplay. Greg contributes actionable marketing angles like micro-holidays and birthday triggers.
Live Demonstration: Conceptualizing the LCA Campaign 6414 Greg intervenes when Jacob suggests a newspaper layout for LCA, asserting firm brand ownership by redefining the agency as an elite luxury SWAT team fitting a vintage Rolex aesthetic. Jacob adapts the prompt accordingly.
Live Demonstration: Iterating in a Fresh Chat 4501 Jacob explains model temperature and iterative prompting in a new conversation window after the initial output generated an unwanted Rolex watch. Greg notices a hallucinated client logo (Disney) during the output review.

Statements from this episode (10)

Assertion Not checkable as stated
Pozel: ChatGPT Can Generate Ad-Account-Ready Creatives in Two Minutes
“We're going to teach people how to create ads ready for the ad account using ChatGPT. In about two minutes, it's that simple.”
Jacob Pozel Apr 2, 2025 ▶ 1:24
Insight
Reference images matter more than prompts in ChatGPT ad creation
“The prompting is way less important. And the prompting, you can actually be pretty straightforward, and it has a much better understanding of intent and creativity. So the important thing I found in creating good ads are actually the inspiration images and the…”
Jacob Pozel Apr 2, 2025 ▶ 2:31
Prediction Held up
API access will let marketers generate 100 ad variants in five minutes
“And when the API becomes generally available, this is going to be trivial. It's literally going to be like, This is your inspiration or like reference ad. These are all the different audiences, just like boom, boom, boom, smash them out and you'll have a hundr…”
Jacob Pozel Apr 2, 2025 ▶ 4:51
Insight
Pozel: Multimodal AI requires reference images for smaller brands
“A super important thing to understand about these models is that they have a really good understanding of well-known concepts, ideas, brands. So it knows Nike really well. It knows Adidas really well. Even like the classic online econ brands, like it knows Rid…”
Jacob Pozel Apr 2, 2025 ▶ 6:16
Opinion
Pozel: Sora has looser policy guidelines and better photorealism than ChatGPT
“Another really cool thing is I think Sora is better at it's, it has less strict policy guidelines and it's better at photo, photo realism. I don't actually understand why that's the case considering they probably use the same underlying models, but I think tha…”
Jacob Pozel Apr 2, 2025 ▶ 9:48
Insight
Pozel: Start a new chat when iterating AI image prompts
“If you struggle with the initial prompt, it's important not to carry on in the same conversation and try and adjust your image in the same conversation because you generally get the recycle of worse and worse rather than it actually improving because it's usin…”
Jacob Pozel Apr 2, 2025 ▶ 12:41
Opinion
AI ad remixing is ethical because agencies already copy competitor templates
“The way that most of these companies work is they are finding inspiration ads. There's like 50 companies that do this. They provide you inspiration ads, or they're getting templates, which are recreated ads in Figma files or Canva files. There's companies that…”
Jacob Pozel Apr 2, 2025 ▶ 15:25
Prediction Not checkable as stated
Isenberg: AI will make ad targeting more personalized and entertaining
“The targeting is going to get a lot more personal. It's going to be a lot more funny. It's going to be ads are going to be a lot more entertaining. And it's really up to, you know, the idea people to come up with these ideas.”
Greg Isenberg Apr 2, 2025 ▶ 17:59
Insight
Pozel: AI enables brands to rapidly newsjack short news cycles
“Companies now have the ability to take advantage of short, like new cycle moments. Like they never could have before because it was expensive, prohibitively expensive and slow.”
Jacob Pozel Apr 2, 2025 ▶ 18:29
Insight
Pozel: Iterating on AI-generated reference images degrades quality over time
“The logo is still a little bit worse because it can never be more perfect than the reference. Like that's kind of inherently impossible. And so it's always going to get a little bit worse over time.”
Jacob Pozel Apr 2, 2025 ▶ 34:56
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 300 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more