ChatGPT Ads Ships a Product Carousel, and a Deadline: What to Do Before August 17

Two changes landed inside ChatGPT Ads in the same week, and only one of them made headlines.

The product feed carousel format went live on August 7, letting e-commerce catalogs show multiple products in a single ad unit instead of one at a time.

The quieter change sits ten days behind it: Automatic Advanced Matching becomes the default for every existing ChatGPT Ads pixel on August 17, unless you opt out first.

If you’re running or testing ChatGPT Ads right now, both changes affect you. Here’s what shipped, what the deadline actually means, and what to check before Monday.

The carousel turns ChatGPT into a shoppable shelf

Until this month, a ChatGPT Ads placement below a conversation showed one product from one advertiser. Digiday verified screenshots confirming the new format on August 6: multiple products from the same retailer now appear side by side in a single unit, pulled directly from the advertiser’s product feed rather than built as individual ads.

The mechanics work the way Google Shopping does. Retailers connect a structured catalog, OpenAI decides which products are eligible, and the system assembles the carousel automatically based on what the conversation signals. No manual carousel campaign to build, no per-product ad creation. If your product feed is already connected to ChatGPT Ads, this format applies to your account without any new setup.

The timing isn’t an accident. Digiday reported that OpenAI is targeting a $2.5 billion ad revenue year in 2026, and Q4 is when that target gets tested. A shoppable, multi-product unit gives holiday-season browsers more to compare without leaving the chat, and it gives OpenAI a placement that looks more like retail media than a single sponsored link.

For brands on Shopify, this connects to work you’ve likely already done for Google Shopping or Meta’s dynamic product ads: the same feed hygiene (accurate pricing, real-time inventory, clean titles) carries over. For WooCommerce and BigCommerce brands, native feed support is thinner. WooCommerce merchants currently need a third-party plugin to generate an OpenAI-compliant feed, since there’s still no first-party WooCommerce integration comparable to Shopify’s.

What actually changed on August 17

Automatic Advanced Matching, or AAM, is the piece with a hard date attached. OpenAI sent advertisers a product-update email during the week of August 3 covering nine changes, and this is the one that acts without you doing anything.

Here’s what AAM does: the ChatGPT Ads pixel reads customer information typed into forms on your site, hashes it in the browser using SHA-256, and sends the hash along with the conversion event so OpenAI can match more purchases back to the ads that likely drove them. OpenAI’s own documentation is explicit that raw customer data never leaves the browser unhashed, and no manual pixel configuration is required.

The mechanism exists to solve a real measurement gap. ChatGPT’s primary attribution signal is a first-party cookie tied to a click identifier, which works well within one browser session and less well across devices or after the cookie expires. Advanced matching is designed to catch conversions that identifier alone would miss, the same reason Meta built its own version of advanced matching years ago.

AAM is already the default for any pixel created from now on. What happens August 17 is that it turns on for every pixel that already exists, unless you go to Tools, then Conversions, then Data Source, then Edit pixel, and opt out first.

Why this should sound familiar

AdBeacon’s whole reason for existing is that ad platforms grade their own homework: they report the conversions their own systems attribute, using rules they set and don’t fully disclose. That’s exactly the dynamic AAM extends to ChatGPT, and it’s worth remembering that your attribution data already has a blind spot for ChatGPT-driven discovery even before AAM enters the picture.

Advanced matching isn’t unique to OpenAI. Meta has run its own version for years, and the pattern it produces is well documented: better hash-based matching means more conversions get attributed, and more attributed conversions push reported ROAS higher, whether or not the ad actually drove the extra sales. It’s the same mechanism behind the gap AdBeacon flagged on one account where Meta reported 3.23x ROAS against an AdBeacon-measured 0.93x, once view-through credit and modeled conversions came out of the number.

ChatGPT Ads is still a young channel, so the scale of any inflation from AAM is unproven. But the structure is identical to what’s already playing out on Meta and Google as CPMs climb: a platform improving its own ability to claim credit for conversions, reported inside the same dashboard the platform controls end to end. Nothing about AAM is dishonest on its own, better matching is a real improvement to a real measurement gap. The issue is what happens when a brand takes the resulting number at face value with no independent check against it.

What to actually do before Monday Aug 17th

Two things, both quick:

Check whether your product feed is connected.

If you’re running, testing, or planning to test ChatGPT Ads, confirm your catalog is live in Ads Manager and that pricing, availability, and images are current. A stale feed doesn’t just miss the new carousel format, it risks products getting excluded from eligibility entirely.

Decide on Automatic Advanced Matching before it decides for you.

This isn’t a call to universally opt out. Better match rates are a legitimate improvement if your consent and privacy documentation already cover sharing hashed contact information with an advertising platform.

If you’re in a regulated category, or you’re not sure what your forms currently collect, that’s a five-minute conversation with whoever owns your privacy policy, not a decision to leave on autopilot. Either way, treat the reported ROAS from a ChatGPT Ads pixel with AAM enabled as directionally useful, not as a verified number, the same way you’d treat Meta’s.

If you want to see what click-only, first-party attribution looks like sitting next to whatever ChatGPT Ads or any other platform reports, book a live AdBeacon demo.

—-

FAQ

What is the ChatGPT Ads product carousel?

It’s a feed-based ad format that shows multiple products from one retailer in a single unit below a ChatGPT conversation, similar to how Google Shopping displays multiple results from one merchant. It launched in testing in early August 2026.

What is Automatic Advanced Matching in ChatGPT Ads?

AAM is a pixel feature that hashes customer information from your website forms in the browser and sends it with conversion events, so OpenAI can match more purchases back to the ads that drove them. It’s already the default for new pixels and becomes the default for existing pixels on August 17, 2026.

How do I opt out of Automatic Advanced Matching before August 17?

In ChatGPT Ads Manager, go to Tools, then Conversions, then Data Source, then Edit pixel, and disable Automatic Advanced Matching there. If you take no action, it turns on automatically.

Does the ChatGPT Ads carousel work with WooCommerce or BigCommerce feeds?

Shopify has the most direct feed integration today. WooCommerce and BigCommerce merchants generally need a third-party feed plugin to format their catalog to OpenAI’s specification, since native support lags behind what Shopify offers.

Does Automatic Advanced Matching affect my ROAS reporting accuracy?

It can raise reported ROAS by attributing more conversions to ChatGPT Ads, but that number still comes from OpenAI’s own matching logic. Better matching is a real improvement, but it’s still the platform reporting on itself, so it’s worth checking against independent, click-based attribution rather than taking it at face value.