We onboarded 10 Shopify clients to ChatGPT Ads in the first quarter of the 2026 beta opening. The cohort spans apparel, home goods, beauty, fitness, kitchen tools, and dropshipping. Order volumes ranged from $20K to $400K per month. Total combined ad spend over the 90-day window was just under $180K. This is the consolidated writeup of what we learned.
The headline numbers
Across the 10 clients, the cohort produced the following 90-day aggregate results:
- Total spend: $178,400
- Total engagements: 312,000+
- Total conversions: 18,900
- Average engagement-to-conversion: 6.1 percent (range: 3.2 to 9.8 percent)
- Average engagement cost: $0.57
- Average ROAS: 4.8x (range: 1.9x to 8.4x)
For context, the cohort's Google Shopping benchmark during the same period was 3.4x ROAS. ChatGPT Ads outperformed Google Shopping on raw ROAS by roughly 40 percent on average, with the gap narrowing in categories where Google Shopping is highly optimized and widening in categories where it is not.
What worked consistently
Five operational patterns showed up across the highest-performing accounts:
1. Aggressive SKU concentration
Top performers concentrated 70 to 85 percent of spend on 15 to 25 SKUs. Average performers spread spend across 60 to 150 SKUs. The concentration strategy outperformed by roughly 35 percent on ROAS.
2. Rewritten merchant descriptions
Clients who wrote fresh merchant descriptions for ChatGPT Ads outperformed clients who used default Shopify descriptions by 40 to 60 percent on conversion rate. This was the single highest-leverage creative change.
3. Conservative initial bidding
Clients who started at platform-default bids and let the system learn outperformed clients who bid 30 percent above floor in week one. Aggressive early bidding consistently produced worse 30-day performance.
4. Patient optimization cadence
Clients who held the line on "no changes before day 14" outperformed clients who optimized daily. Premature optimization based on noise consistently destroyed value.
5. Combined GEO investment
Clients who had been doing meaningful GEO work for 6+ months prior to ChatGPT Ads launch had 25 to 40 percent higher click-through rates from placements. The compounding effect of organic AI visibility is real.
What did not work
Four patterns from underperforming accounts:
1. Treating ChatGPT Ads like Google Shopping
Some clients tried to apply Google Shopping operational playbooks to ChatGPT Ads: aggressive bid laddering, broad keyword targeting (which does not exist on ChatGPT Ads anyway), constant creative refresh. None of it worked. The channel rewards different operational habits.
2. Discount-led creative
Clients who opened merchant descriptions with discount offers ("Save 30 percent," "Free shipping over $50") underperformed clients who led with use case and product specifics. The conversational placement context punishes discount-led messaging.
3. Inadequate product page experience
Two clients had strong ChatGPT Ads engagement rates but poor click-to-conversion. The diagnosis: their product pages were not optimized for the AI-mediated visitor who arrived with specific use-case interest already established. Generic product pages that did not reinforce the use case lost the conversion.
4. Channel under-funding
One client started at $400 monthly budget and could not generate enough data to learn anything in 30 days. We restarted the test at $1,800 monthly and the channel began producing meaningful insights within 21 days.
Category-specific learnings
Some patterns specific to individual categories:
Beauty: Buyers responded extremely well to specific ingredient and concern-led merchant descriptions ("for combination skin," "with niacinamide and vitamin C"). Vague positioning ("clean beauty for everyone") underperformed dramatically.
Home office: Comparison panel placements drove most of the volume. Direct inline cards underperformed. Optimization meant ensuring all comparison-relevant attributes (weight capacity, height range, warranty) were filled in the feed.
Kitchen tools: Highest conversion rates of any category in our cohort. The "best X for Y" query pattern is structurally aligned with this category. Almost any decent creative converted.
Fitness: Recovery and rehabilitation positioning outperformed performance and elite-athlete positioning. The buyer base on ChatGPT skews toward problem-solving rather than aspiration.
Apparel: The weakest category in our cohort. Even our best-performing apparel client landed at 3.5 percent conversion vs the cohort average of 6.1 percent. Style-driven purchases do not fit the channel format well.
The biggest surprise
The single most surprising pattern was how much the ChatGPT Ads system rewards merchant description quality relative to other creative inputs. We went into the beta expecting image quality to matter most (as it does on Meta) and category mapping to matter second. In practice, merchant description quality drove larger performance differences than either of those.
This was operationally significant because it shifted where we invest creative budget. We now allocate 40 to 50 percent of ChatGPT Ads creative budget to description writing and rewriting, compared to 20 to 25 percent on image production. The ROI on better descriptions is consistently higher.
What we are doing differently in the next 90 days
Based on the first 90 days, our updated playbook:
- Lead client onboarding with merchant description workshops, not feed setup tutorials
- Default new accounts to 25-SKU concentration rather than 100-SKU breadth
- Require 14-day optimization holds with no exceptions
- Run GEO audit and improvements as a prerequisite, not a parallel workstream
- Build category-specific creative templates for the high-performing categories (kitchen, beauty, fitness, home office)
If you want a hands-on team that has already run 10 ChatGPT Ads launches and refined the playbook, our ChatGPT Ads service covers exactly this. Book a free 30-minute discovery call if you want to discuss your store specifically.