On the NoRobots podcast, Jack Paxton, founder of Top Growth Marketing, shares what his agency learned after spending tens of thousands of dollars testing ChatGPT ads for DTC brands: which products convert, which don’t, where a small brand should put its budget first, and why GEO may be the bigger opportunity right now.
Watch the full conversation on YouTube.
Do ChatGPT ads actually work?
ChatGPT ads hit a $1 billion run rate in under 200 days. Jack’s agency tested them early, across a range of brands and products, and spent tens of thousands of dollars on the ad network.
His first takeaway: the platform is still basic and straightforward, so almost anyone can try it. And whenever a new ad channel appears, he thinks it’s worth testing. You never know. Tracking, though, is very limited for now.
Products under $100 convert
The first tests worked best for impulse and lower-priced items, roughly anything under $100. The customer journey is short enough that people don’t mind clicking an ad in a ChatGPT answer.
One reason, Jack explains, is how the ads appear. They’re shown to free ChatGPT users, and it’s very clear that the ad isn’t part of the AI’s summary answer. If your product shows up next to the organic options, is a fair comparison and isn’t a big investment, people are willing to click and buy. For these products, the agency got a positive ROI and a trackable click-to-conversion path. His advice for anything under $100: definitely run those ads.
Products over $200 don’t, for now
With more expensive products, usually $150 and up and closer to $200 on average, the results were poor. The ad felt too obviously like a paid plug-in to a bigger decision, one that probably involves more research.
The tests were run very early, so the team couldn’t track further down the funnel. Jack expects the platform to improve and to integrate with tools like Northbeam and Triple Whale, so brands can see delayed attribution and where the customer journey really started. That, he thinks, will be a big unlock for ChatGPT ads. But in the direct click-to-purchase model, higher-priced products didn’t pay back, and many of those brands weren’t comfortable continuing to spend there.
Organic vs paid for higher-priced products
Vlad suggested that for expensive products, investing in organic visibility might work better. Jack’s honest answer is the “super annoying” one: you need to do everything.
Still, for higher-priced items, ranking organically in ChatGPT gives much better results than paying for ads. Whether you can get cited is another question. It depends on the query, the industry and how competitive it is. But if you can, the conversion rate and ROI will be much higher.
If you’ve tried and can’t get cited, ads can still be worth it for awareness, so you at least appear in the comparison when people look at their options. Just don’t expect a direct click to checkout in that same session.
Will visual ad formats change things?
ChatGPT is expected to introduce more visual ad formats. Jack thinks taking up more space in the answer will help.
In the early tests, the team had to get very granular with keywords and strategy. If the ad doesn’t match the query, the impression is wasted. His example: someone searches for the best toothbrush for people with receding gums, and you show them toothpaste. It’s relevant, but it’s not what they want to buy. Early on, there was a lot of that.
With creative, brands can sell the benefit or the end result rather than just match the query. Bundles could work well here. A gum care kit with toothpaste, a toothbrush and floss is probably a better ad to show someone searching for a toothbrush than toothpaste alone.
ChatGPT ads are not the new early Facebook
Many brands hope ChatGPT ads will be a rerun of Facebook’s early days. Jack doesn’t think so. CPMs and cost per click are roughly on par with other platforms. When Facebook ads started, CPMs were around two or three dollars.
Facebook’s early success also came from things that won’t repeat: people getting apps on iPhones, mobile internet becoming part of everyday life. Now the competition for ad space is so high that the platform can’t really start that cheap.
Where a small brand should spend its budget
Vlad asked how a small e-commerce brand with a limited budget should split it between ChatGPT, Meta, Google and TikTok.
It depends on the product and the customer, Jack says:
- Google Search gives a very good ROI if you solve a very specific problem and know the exact keywords people search for.
- Meta ads work better for more general products, especially creator collaboration ads.
- ChatGPT ads are best kept as an experimental channel, about 20% of the budget.
Collaboration ads are, in his view, one of the best uses of ad spend across the brands his agency spends a few million dollars a month on. They now put roughly 10% to 40% of brand budgets into them. Get a few early users and boost their posts, and you’ll typically see lower CPMs, more engagement, a lower cost per click and far more authentic content than ads from your brand page.
He wouldn’t suggest small brands start with ChatGPT ads. The channel takes a bigger budget and a lot of experimentation: you start wide and narrow down, rather than starting small and expanding. For now, he sees it as an expansion channel that adds to Meta and Google, not a replacement.
Why GEO is the real early-mover opportunity
If you’re a small brand on a small budget, Jack recommends optimizing for AEO and SEO first, then moving into ads. As funny as it sounds, brands that do it now are still early movers. Not many are well optimized for AI search yet.
Getting cited early also builds authority faster. AI platforms are starting to look at the history of the sources they cite, not just the quality of the answer. Right now, a young brand can still give better answers than bigger brands and get its website cited over theirs. As bigger brands catch on, that window will close.
The traffic is worth it, too. Jack says GEO traffic converts at two to three times the rate of the agency’s old organic traffic. The volume is lower, but conversion rates for brands that get cited are very high. ChatGPT is currently the leader as a referral source for sales, leads and traffic, while Claude cites and links out less.
What happened when clients tried AI-only tools
Platforms like Meta are moving toward fully automated ad creation and targeting. So why hire an agency at all?
Jack has had one or two clients leave Top Growth Marketing to try an AI service. They came back within two or three months. The tools aren’t there yet if you care about detail: wrong sizing, off-brand visuals, slightly wrong messaging, occasional targeting issues or misrepresentation.
There’s a strategic problem too. Many AI platforms chase the most successful metric, which is often bottom-of-funnel retargeting like abandoned carts. They spend the whole budget there, without the insight to say: no, ads should drive new customer growth, while email and SMS handle the bottom of the funnel and lifetime value.
The best model, in his view, is an experienced human directing AI. The AI does the heavy lifting: the analysis and the time-consuming work. Then it lays out options, and the human picks the right one based on the product and the client. Clients’ priorities also shift. The best ROAS might come from a product shot with an 80% off sign, but many brands still want their ads to look good. That takes human judgment.
He adds one more point. People will always want someone to answer for it when something goes wrong. If an AI system accidentally spends $20,000, there’s no one to go back to.
When a client wants to try AI on their own, the agency lists the pros and cons and shares what it has seen with others. Then the client makes the call, since it’s their money, time and brand. Often, Jack says, brand guidelines go out the window once everything is handed to ChatGPT or Claude.
AI raised the bar, not the profit margin
Jack’s team didn’t shrink because of AI. Instead, the bar has been raised across the industry. There are more reports, more analysis and more strategy docs. Decisions that used to rest on experience now often come with five AI-generated reports to confirm them. The average person is working more hours because deliverables have grown in speed and volume.
He doesn’t think he’s squeezed more profit out of AI. Instead, the service is better value for brands. A media buyer can now run deep-dive analysis that once needed a data scientist, as long as they have the right data and guardrails.
The agency is also training its Claude organization on around ten years of chats, emails, reports and presentations. That’s why he believes individual brands will struggle to use AI as well as agencies can: agencies simply have more data to train on.
The junior talent gap
Vlad shared a similar view from development: on large projects, one mistake can cost $20,000 in ten minutes of sales, so even AI-made work needs heavy QA before production.
Jack agrees. His software teams have gotten smaller, but they still need experienced developers to architect and check the work. The part he finds most disappointing is the effect on junior roles. With fewer people coming into the industry, learning and growing, he expects a skills gap in about five years, when senior talent moves on or retires.
What agency Jack would start today
“The grass is always greener,” Jack says. Everyone with an agency wants a product, and everyone with a product wants a service. If he started today, he’d still start an agency, but a different one: an AEO, GEO and SEO agency, or a whitelisting and creator agency. In his view, ad and email agencies are saturated, and smaller ones will consolidate as they struggle to keep up with what the service now requires.
Any agency has to keep pivoting. Top Growth Marketing started as a Facebook-only ad agency about twelve years ago. Today it covers all the ad platforms, email and SMS, creative and whitelisting. An agency is great for building momentum, he says, but if he were starting fresh, the goal would be to turn it into a productized service.
The skill marketers need most
For Jack, it’s common sense and real-world experience. Many people give AI too little context, get generic answers and follow them without question.
AI can give you every possible solution. Experience tells you which one is right, because you’ve done it three times before. Every report and idea can sound amazing in ChatGPT or Claude. The skill is being able to say: no, that’s not how it usually works. Better data plus better experience makes better outputs.
Vlad added that using AI properly takes real work: setting up Claude skills, MCPs and connectors for different tasks. Jack agrees that many people use ChatGPT and Claude like Google, asking basic questions with no context. The setup is the difference between using it effectively and not.
About Jack Paxton and Top Growth Marketing
Top Growth Marketing has been running paid social and search ads for more than ten years across Google, Meta, Snapchat, TikTok and other platforms. The agency works with DTC e-commerce brands, usually on Shopify and Klaviyo, and also runs whitelisting ads, gifting and TikTok Shop campaigns, plus email and SMS. For many clients, the channels it manages drive 60% to 70% or more of revenue.
The agency usually sees the biggest impact with brands already doing at least $100,000 a month in revenue. Jack compares growth to pushing a boulder: once it’s moving, you can speed it up much faster than starting from a standstill. A good product, good marketing and strong retention all still have to go right.
FAQ
Do ChatGPT ads work?
In Top Growth Marketing’s tests, yes, for impulse and lower-priced products under about $100. Products around $200 and up did not get a good ROI on a direct click-to-purchase basis.
Are ChatGPT ads cheap like early Facebook ads?
No. Jack says CPMs and cost per click are roughly on par with other platforms, unlike the $2–3 CPMs of Facebook’s early days.
How much budget should a small brand put into ChatGPT ads?
Jack suggests treating ChatGPT ads as an experimental channel at about 20% of the budget, with Google Search or Meta collaboration ads as core channels.
Does GEO traffic convert better than SEO traffic?
In Jack’s experience, GEO traffic converts at two to three times the rate of the agency’s old organic traffic, though the volume is lower.
Can AI tools replace a marketing agency?
Not yet, according to Jack. Clients who left for AI-only services came back within two to three months, citing issues with detail, branding and strategy.
Who is Jack Paxton?
Jack Paxton is the founder of Top Growth Marketing, a growth marketing agency that runs ads, email and SMS for DTC e-commerce brands and manages a few million dollars a month in ad spend.
Where to find Jack Paxton
Learn more about the agency at topgrowthmarketing.com, or connect with Jack on LinkedIn. You can also find Top Growth Marketing on NoRobots.
Watch the full interview: ChatGPT Ads Actually Work (But Only Under $100)
Looking for a growth partner for your e-commerce brand? Browse the marketing and digital agencies listed on NoRobots.
