ChatGPT Product Feed Campaigns and Carousel Ads: A Practical Guide
Product-feed and carousel formats arrived in ChatGPT Ads Manager in August 2026. Here is how to prepare your feed and when these campaigns make sense.
For most of 2026, a ChatGPT ad was a single chat card: one headline, one short body, one image, one link. That works well for a hero product or a service. It works less well for a store with 300 products, where the right item depends on what someone is asking about.
In August 2026, OpenAI added product-feed and carousel formats to ChatGPT Ads Manager, along with oCPC (optimized cost per click) in beta for product-feed campaigns. Together they give ecommerce brands a way to advertise a catalog, not just a single product.
This guide explains what has been reported about these formats, how to prepare a product feed, how to choose between feed campaigns and classic chat cards, and how to measure them. Because the formats are new and still in beta in places, check Ads Manager for current specs and eligibility.
What product-feed campaigns are
A product-feed campaign connects your product catalog to ChatGPT Ads Manager. Instead of writing every ad by hand, you supply a structured list of products (titles, descriptions, prices, images, links, availability) and the system can show relevant products in response to conversations that match.
The carousel format lets a single ad show several products side by side. Picture someone asking ChatGPT for "a lightweight rain jacket for travel under $150." A carousel could show a few of your jackets that fit, each with its own image, price and link.
What stays the same:
- Ads appear below ChatGPT's answer, labeled as sponsored, and do not change the answer.
- Only logged-in adult Free and Go users see ads.
- OpenAI's ad policies apply to every product in your feed, not just the first one shown.
Who should use product feeds
| Business | Product feed fit | Why |
|---|---|---|
| Ecommerce store with many SKUs | Strong | Different products fit different questions |
| DTC brand with a small range (3–10 products) | Moderate | Can help, but hand-written chat cards may do just as well |
| Single hero product | Weak | A classic chat card gives you more control over the message |
| Consumer app or service | Not applicable | Nothing to put in a product feed |
| Course creator | Usually weak | A few offers are easier to advertise individually |
If your catalog is large enough that you cannot sensibly write an ad for every product, a feed campaign is worth testing.
Step 1: Audit your product data
Your feed is your ad copy. A product title like "SKU-4471 Blue M" will not earn clicks in a chat interface. Before connecting anything, audit your catalog.
- Titles. Clear, descriptive, readable by a person: "Women's Packable Rain Jacket, Navy" beats "PKJ-W-NV."
- Descriptions. Lead with what the product is and who it is for. Remove marketing fluff and anything you cannot prove.
- Images. One clean main image per product, product clearly visible. No text overlays with prices or claims, no before/after.
- Prices and availability. Must match your site exactly and stay in sync. Showing an out-of-stock product or a stale price breaks message match and can trigger policy issues.
- Links. Each product links to its own product page, with the correct variant selected.
- Categories and attributes. Size, color, material and gender fields help matching.
Most ecommerce platforms already produce a product feed for other channels. Start from that, then clean it up for a chat context.
Step 2: Write titles and descriptions for questions
People ask ChatGPT questions in natural language. Feed data that answers those questions matches better and earns more clicks.
Take a hypothetical $58 merino wool T-shirt:
- Weak title: Merino Tee
- Better title: Men's Merino Wool T-Shirt, Odor-Resistant, Lightweight
- Weak description: Our best-selling tee, loved by customers everywhere.
- Better description: Lightweight merino wool tee for travel and everyday wear. Resists odor between washes. Machine washable.
The better version answers likely questions: is it good for travel, does it smell after a day, can I wash it normally. It also avoids an unproven "best-selling, loved everywhere" claim. Tools like the product description generator can help rewrite at scale, but review every claim before it goes into the feed.
Step 3: Connect the feed and set up the campaign
The general setup, subject to how Ads Manager presents it in your account:
- Connect or upload your product feed in Ads Manager.
- Fix any items the system flags (missing images, invalid prices, broken links).
- Choose which products or product groups to include. Start with your best sellers or highest-margin items, not the entire catalog.
- Set context hints for the campaign that describe the situations your products suit.
- Choose a bid type, CPC or oCPC where available.
- Set a daily budget above the minimum (about $25 per campaign in the US).
Start with a focused product set
Including all 300 products on day one spreads budget thinly and makes results hard to read. A better starting point is a group of 10 to 30 products that share a buyer situation, such as "travel clothing" or "gifts under $50," with healthy margins and steady stock.
Step 4: Choose between CPC and oCPC
oCPC, introduced in beta for product-feed campaigns in August 2026, still charges per click but adjusts bids based on how likely each click is to convert, using your pixel and Conversions API data.
Use CPC if:
- Your conversion tracking is new or not yet verified
- You expect few purchases in the first weeks
- You want a clean baseline
Use oCPC if:
- Purchases are tracked reliably, including value
- The campaign already produces a steady flow of conversions
- You want the system to favor clicks more likely to buy
A sensible path is to start on CPC, verify tracking, and then test oCPC against the CPC campaign. Our guide to CPC vs CPM vs oCPC covers the trade-offs, and conversion tracking setup is a prerequisite for oCPC.
Feed campaigns vs classic chat cards
Feed campaigns do not replace hand-written ads. They do a different job.
| Classic chat card | Product feed / carousel | |
|---|---|---|
| Best for | A hero product, a service, a specific angle | Matching many products to many questions |
| Message control | High: you write every word | Lower: driven by feed data |
| Setup effort | Per ad | Up front on feed quality |
| Strength | Sharp, proven message | Breadth and relevance across a catalog |
| Risk | Limited reach per ad | Weak feed data produces weak ads |
Many stores will run both: chat cards built around proven messages for hero products and key buyer questions, plus a feed campaign covering the wider catalog.
Measuring feed campaigns
Measure product-level results, not just campaign totals. Within a feed campaign, a handful of products often produce most of the revenue.
Here is a hypothetical example. A feed campaign spends $2,100 over three weeks across 20 products and generates $5,040 in tracked revenue, a ROAS of 2.4. On closer look, four products generated $3,900 of that revenue. If you can see spend by product, you may find those four are profitable and the remaining sixteen are not. Moving budget to them, or giving them dedicated chat cards, typically improves results more than any bid change.
Check overall efficiency with the ROAS calculator, and compare against your break-even ROAS before scaling.
Common feed problems
- Price mismatches between feed and site after a sale ends
- Out-of-stock items still being shown
- Generic titles that do not answer any question
- Images with text overlays making claims or showing discounts
- Restricted products (supplements, alcohol and similar) included without checking category rules
- Too many products in one campaign, so none gets enough budget to learn
Where SecondWin fits
SecondWin focuses on writing original, policy-checked chat card ads built around the proven messages in your niche and the questions buyers ask ChatGPT. For catalog brands, those ads complement a feed campaign: they carry your sharpest messages for hero products while the feed covers breadth. Campaigns run in your own OpenAI ad account, and plans with more campaigns (Growth and Scale) suit stores that want both. See pricing for details.
To see which messages and products we would lead with, run a free analysis of your store's URL.
FAQ
What are ChatGPT carousel ads?
Carousel ads are a format added to ChatGPT Ads Manager in August 2026, alongside product-feed campaigns. A carousel can show several products in one ad unit, each with its own image, price and link, below ChatGPT's answer. They are suited to stores with many products, where different items fit different questions. Check Ads Manager for current specs.
Do I need a product feed to advertise on ChatGPT?
No. Classic chat card ads, with a hand-written headline, body, image and link, work without any feed. Product feeds are optional and most useful for ecommerce stores with many SKUs. Single-product brands, apps, services and course creators will usually do better with individually written ads.
Can I use oCPC without a product feed?
As of its August 2026 beta, oCPC was introduced for product-feed campaigns. Availability may change, so check Ads Manager. Whatever campaign type you use, oCPC depends on reliable conversion tracking through the OpenAI pixel and Conversions API, plus enough conversions for the system to learn which clicks are likely to buy.
How many products should I include in a feed campaign?
Start focused. A group of 10 to 30 products that share a buyer situation and have healthy margins and steady stock is easier to learn from than an entire catalog. Once you see which products convert, expand gradually or give top performers their own dedicated chat card ads.