Speaker 1: It seems like marketers these days are either all in or completely clueless on how to use ChatGPT.
A split screen appears. On the left, a photo of a very muscular, shirtless man sitting on a couch using a laptop. On the right, a still from the video game Grand Theft Auto: San Andreas showing a character who resembles Mr. Bean looking confused. The word "or" is between them.
Text on screen: "How to use ChatGPT"
Speaker 1: And here at Motion, we work with some of the best performance marketers in the world who use our tools to analyze over five billion dollars in Meta ad spend annually.
A series of photos of people with their names underneath them slides across the screen. "Sarah Sanchez", "Dara Denney", "Lachezar Voynov".]
> [VISUAL: Animated graphics of marketing dashboards appear. One shows a comment box ("The lifestyle shot worked b..."), another shows a bar chart ("Monthly Reports - Top Perform..."), and a third shows a line graph ("Spend vs ROAS").]
> [VISUAL: A purple bar chart animates upwards, with the number "$5,000,000,000" appearing above it.]
> [VISUAL: A screenshot of the Meta Ad Library website.
Speaker 1: And that doesn't happen without being extremely creative, which is why we interviewed Stanford professor and AI researcher Jeremy Utley to put together five of the biggest mistakes to avoid when marketers are trying to brainstorm with AI tools to get as creative as possible.
A man stands on a stage, smiling and gesturing. A lower third identifies him: "Jeremy Utley, Director of Executive Education at Stanford, and Author".]
> [VISUAL: Text on screen: "5 biggest mistakes to avoid"
Speaker 1: PS, this guy Jeremy is like a freaking AI whisperer and everything I'm sharing is stuff that he taught us during our interview with him, so pay attention if you want to tap into virtually unlimited creative potential with AI.
A close-up of Jeremy Utley's smiling face. His hands are in front of him, framing the green ChatGPT logo which is superimposed in the center.
Speaker 1: Anyway, for this video, I'll pretend that I'm a performance marketer for a nutrition brand that I'm making up called EnergizeMax.
Text overlay points to the speaker with the words "Performance marketer".]
> [VISUAL: A black bottle with a red label that says "EnergizeMax" is shown against a black background with the words "ENERGIZE MAX" repeated in white text.
Speaker 1: And I'm going to show you how I would use ChatGPT to help me create amazing video ads based on Jeremy's tips. Okay.
The ChatGPT logo appears next to the speaker.]
> [VISUAL: A man's head (Jeremy Utley) peeks out from behind the speaker's laptop.
Speaker 1: Mistake number one, being too vague.
Text on screen: "Mistake 1 Being vague"
Speaker 1: If you ask ChatGPT, "please give me some advertising ideas," that's way too broad of an input.
Screenshot of the ChatGPT interface. The prompt typed in is: "please give me some advertising ideas"
Speaker 1: Just look at how generic the responses are. Not very helpful. To correct this, you need to provide ChatGPT with detailed and specific prompts.
Screenshot of a generic ChatGPT response. It lists "Digital Advertising" with bullet points for "Social Media Campaigns," "Influencer Partnerships," "Content Marketing," and "Email Marketing."
Speaker 1: Let me show you how. Let's type in, "ChatGPT, let's assume you are an expert advertiser working with an experienced team of performance marketers. I need three creative video ad concepts for a fake brand called EnergizeMax, a supplement that boosts energy and focus. The ad should appeal to young professionals who love technology and often work long hours. The tone should be humorous and engaging, suitable for TikTok."
A screenshot of the ChatGPT interface shows the detailed prompt being typed out.
Speaker 1: Now, look at these outputs compared to the previous ones. Way better, right?
A screenshot of the detailed ChatGPT response. It shows "Concept 1: 'Techie Transformation'" with scene-by-scene breakdowns, followed by "Concept 2: 'The EnergizeMax Hack'" and "Concept 3: 'EnergizeMax Time Warp'".
Speaker 1: Well, that's because ChatGPT uses large-scale machine learning models trained to generate responses based on patterns that it's learned.
The ChatGPT logo animates on screen.]
> [VISUAL: A screenshot of a webpage from data.iku.com. A sentence is highlighted: "large language models (LLMs), are capable of generating text on a seemingly endless range of topics."
Speaker 1: So when you use vague inputs, it defaults to more general patterns of outputs.
An animation shows the words "Vague inputs" with an arrow pointing to the ChatGPT logo. Another arrow points from the logo to the words "General output" repeated three times.
Speaker 1: Mistake number two, settling for the first idea.
Text on screen: "Mistake 2 Settling for the first idea"
Speaker 1: Look, it's tempting to run with the first good idea that you get from ChatGPT. And as humans, we love the feeling of completing a task. There's actually a psychological term for this called cognitive closure. But as much as you want to, keep digging deeper with ChatGPT and don't just mark it as complete. Here's how you can do that. Let's say, "ChatGPT, that's a good start. Now, give me three different twists on this concept, focusing on unexpected situations where EnergizeMax could be hilariously helpful. Be as silly as you want to be."
Screenshot of the ChatGPT interface. A follow-up prompt is being typed: "That's a good start, ChatGPT. Now, give me three different twists on this concept, focusing on unexpected situations where EnergizeMax could be hilariously helpful. Be as silly as you want to be"
Speaker 1: Okay, holy cow. These ideas are way more fleshed out and are sparking a whole lot for me to consider as a performance marketer.
Screenshot of the ChatGPT response. It shows three new concepts: "Concept 1: 'EnergizeMax at the Family Reunion'", "Concept 2: 'EnergizeMax in a Traffic Jam'", and "Concept 3: 'EnergizeMax at a Boring Meeting'".
Speaker 1: So when you request multiple iterations or variations on a theme, you're relying on the model's ability to generate diverse ideas based on its training to recognize a wide range of potential scenarios, which avoids cognitive biases that limit creativity. Makes sense.
Speaker 1: Mistake number three, not using an iterative approach.
Text on screen: "Mistake 3 Not using an iterative approach"
Speaker 1: You need to interact with ChatGPT like you are Slack messaging a coworker on your team. And at first, this can feel really weird, but you can very much define your ideas with a back-and-forth dialogue. Let me show you what I mean. I'll type in, "I like the idea of EnergizeMax helping in unexpected situations. Can you develop this into a few different scenarios where someone ends up in a funny situation due to their heightened focus and energy?"
Screenshot of the ChatGPT interface. A follow-up prompt is being typed: "I like the idea of EnergizeMax helping in unexpected situations. Can you develop this into a few different scenarios where someone ends up in funny situations due to their heightened focus and energy?"
Speaker 1: You see how I keep asking more questions and requesting more detailed information? Iterative dialogue mirrors more complex human thought processes, which prompts AI to recalibrate its responses.
Screenshot of the ChatGPT response. It shows new scenarios: "Scenario 1: 'EnergizeMax at the Gym'", "Scenario 2: 'EnergizeMax at the Office Party'".
Speaker 1: It can then progressively refine its outputs that align more to your creative goals.
Speaker 1: Mistake number four, skipping the details.
Text on screen: "Mistake 4 Skipping the details"
Speaker 1: You have to give ChatGPT tons and tons of details before you ask it any questions. More context allows ChatGPT to tailor its responses better. So make sure to provide information about your target audience, product features, and the desired emotional impact.
Text appears on screen: "Target audience", "Product features", "Desired emotional impact".
Speaker 1: Let me show you how that can give us a better ad concept with our fake brand here. So, let me type in, "Refine the previous ad concept for YouTube. Incorporate visual elements that highlight the confusion, and suggest dialogues that could resonate with tech-savvy young professionals who thrive on a busy schedule."
Screenshot of the ChatGPT interface. A follow-up prompt is being typed: "Refine the previous ad concept for YouTube. Incorporate visual elements that highlight the confusion, and suggest dialogues that could resonate with tech-savvy young professionals who thrive on a busy schedule"
Speaker 1: Bada boom, bada bing. Look at this. Detail-rich prompts are, they're a gold mine for ChatGPT, let's be honest.
Screenshot of the ChatGPT response. It shows a refined ad concept titled "EnergizeMax: Unlock Your Inner Superhero" with detailed scene breakdowns including visuals, dialogue, music, and on-screen text.
Speaker 1: They're details that you're adding and they're engaging specific pathways within the model's neural network, which enables greater nuance and more targeted text outputs.
An animation of a neural network with nodes and connecting lines.
Speaker 1: Mistake number five, ignoring the creative process.
Text on screen: "Mistake 5 Ignoring the creative process"
Speaker 1: Look, AI should be used to augment creativity, not replace it. Here's how you can use ChatGPT to spark more creative thinking rather than just dictating ideas back to you. I'm going to go ahead and type in, "This concept is shaping up well, but it needs a creative hook. ChatGPT, think outside the box and propose a unique angle that could make this ad go viral on Meta and TikTok. Consider incorporating trending memes or current events related to workplace culture."
Screenshot of the ChatGPT interface. A follow-up prompt is being typed: "This concept is shaping up well, but it needs a creative hook. ChatGPT, think outside the box and propose a unique angle that could make this ad go viral on Meta and TikTok. Consider incorporating trending memes or current events related to workplace culture."
Speaker 1: By encouraging outside-the-box thinking like this from ChatGPT, you are prompting it to cross-reference and combine disparate ideas from its training data and potentially leading, and hopefully leading, to unexpected creative ideas for you.
Screenshot of the ChatGPT response. It shows a new concept: "EnergizeMax: Unleash Your Main Character Energy" which incorporates the "quiet quitting" and "main character energy" memes.]
> [VISUAL: An animation shows lines connecting different prompts, which then lead to four lightbulb icons.
Speaker 1: Look, I know this might get nerdy and complex, but that's what this concept is for. That's what this video is for. I want you to interact with ChatGPT much more like what Jeremy Utley taught us in this video.
A still of a woman's face, blurred, with the ChatGPT logo over it. It cuts to a man with yellow sunglasses.]
> [VISUAL: The speaker is talking, and Jeremy Utley's head peeks out from behind the laptop again.
Speaker 1: Just like we did with our fake brand here in EnergizeMax, we hope this helps you, and I'm curious to know what prompts you're using to get the most out of AI.
