OpenAI's Glossy ChatGPT Ads Miss the Mark. Here's What Marketers Should Learn
OpenAI launched its biggest ad push yet for ChatGPT with cinematic spots intended to feel relatable. The campaign centers on two ads - "Dish," where a guy tries to impress a date with a prompt, and "Pull Up," where another tries to feel stronger by asking the bot for help.
Launched during NFL Primetime, the work will run across TV, streaming, paid social, outdoor, and influencer channels in the US and UK through the end of 2025. The scale is serious. The impact, according to independent testing, isn't.
What the research actually says
Adweek reported that System1 tested both ads with a US consumer panel using Paul Ekman's universal emotions framework: happiness, surprise, sadness, fear, anger, disgust, and contempt, plus neutral. They also captured second-by-second emotional response to diagnose moments that land or fall flat.
The result: both spots ranked in the lowest fifth for long-term growth potential and immediate sales impact. Most viewers couldn't identify the brand until the ChatGPT logo appeared at the end.
System1's approach is built to predict effectiveness, and Ekman's universal emotions are a proven lens for decoding response. If you care about market share and short-term lift, this is the kind of signal you don't ignore.
Even the bot wasn't impressed
In a twist, ChatGPT itself rated one of the ads a 5/10 when asked by a marketing columnist. Its critique hits the core issue: "Pull-Up is strategically on-brief and nicely made, but it underweights distinctive assets and mid-ad branding, so it risks becoming a likeable, generic 'AI-helped me' story rather than a memorable ChatGPT ad that builds future sales."
Translation: strong craft, weak brand linkage.
Why the creative underdelivered
- Branding shows up too late. If the logo or name only appears in the closing frames, you've paid to build category, not brand.
- Generic story beats. "AI helped me do X" could belong to any competitor. Without distinctive assets, memory won't stick.
- Weak product role. The ads show life moments more than product truth. Viewers remember the scene, not the solution.
- Emotional lift without attribution. Positive emotion that isn't linked to brand cues fails to convert into sales or future demand.
Actionable takeaways for marketers
- Brand early and often. Get name, audio, color, or UI on screen within the first 3 seconds. Repeat lightly throughout.
- Use distinctive brand assets. Ownable codes (sound, typography, UI, color, character) drive recognition under low attention.
- Make the product the hero. Show the interface doing the thing, not just the outcome. UI close-ups beat abstract montage.
- Tie emotion to brand triggers. If the peak moment isn't branded, you're gifting equity to the category.
- Anchor to category entry points. Map common "buy moments" (e.g., "quick dinner for a date," "first gym day") and link them clearly to your brand.
- Say the name in VO or captions. Don't be precious. Clarity beats clever.
- Pretest creatively, not just media. Use second-by-second diagnostics to fix muddled beats and add signature brand cues.
- Avoid interchangeable narratives. If you can swap your logo with a rival and the ad still works, it doesn't work.
How to fix spots like "Dish" and "Pull Up"
- Open with brand presence. Start on the ChatGPT interface or a branded audio sting; keep it on screen during key beats.
- Increase mid-ad branding. On-screen prompts, branded UI overlays, or device close-ups in the middle of the story.
- Make the solution visible. Show the exact prompts and the response driving the outcome. Annotate with a simple line: "Generated in ChatGPT."
- Add a distinctive sign-off. A repeatable sonic cue or mnemonic character that ties emotional payoff to the brand.
- End with a specific CTA. Example: "Plan your next first-date dinner in ChatGPT." Specificity lowers friction.
Measurement that matters
- Balance long and short. Track both immediate sales proxies and long-term brand-building potential. Optimize for both.
- Use emotion as a predictor, not a goal. Positive emotion without brand linkage won't move share.
- Run recognition tests. Can viewers name the brand at 2 seconds, 5 seconds, 10 seconds? Iterate until the answer is yes.
- Test category confusion. If viewers think "AI" rather than "ChatGPT," you have a branding problem, not a media problem.
The bigger picture
These spots look great. That's not enough. Craft without distinctive brand assets is charity work for competitors.
If your AI product is hard to show, simplify to a few high-frequency jobs-to-be-done and dramatize them clearly. Show the interface doing the work. Stamp every memory peak with a brand cue. Repeat. Then scale.
Next steps for your team
- Audit current creative for early branding, asset use, and product visibility.
- Map 5 category entry points and script to them.
- Build a distinctive asset library (sonic, visual, verbal) and mandate usage.
- Pretest with second-by-second diagnostics before you spend big on reach.
If you're upskilling your team on practical AI and marketing use cases, see the AI Certification for Marketing Specialists and our ChatGPT training resources.
Your membership also unlocks:
