Your Product Feed Just Became Your Storefront

Two announcements landed within days of each other this month. OpenAI launched product feed campaigns in ChatGPT's Ads Manager. Google started testing Sponsored Shops — a format that bundles multiple products from one retailer into a branded unit inside Shopping results.

Neither is a response to the other. But both are saying the same thing to ecommerce advertisers: the product feed is no longer the boring spreadsheet that sits behind your advertising. It is the advertising.

That CSV you've been treating like back-office plumbing? It just got promoted to the shop window.

The Feed Is the Ad Now (No, Literally)

OpenAI's product feed campaigns, now live in beta, let retail advertisers upload a structured product catalogue via SFTP and run campaigns directly from that data. The feed needs between 1,000 and 2 million products. Once it's in, advertisers create campaigns by selecting the feed as the campaign type, filtering which products are eligible, and setting up a layout template.

Here's the thing: there is no separate step where you write ad copy or design a creative. Your product titles, descriptions, and images are pulled directly from the feed and served to users mid-conversation. That title you wrote in eleven seconds while juggling three other tabs? That's your ad now. The system matches eligible products to user conversations that indicate purchase intent — someone asking about running shoes, comparing kitchen appliances, researching skincare — and serves the most relevant item from your catalogue.

Each ad account is currently limited to one feed connection, and products only appear in paid placements during the beta. They won't surface in organic ChatGPT responses, though OpenAI has indicated that capability may follow. (Because why stop at paid when you can own the whole funnel, right?)

Early performance data suggests the format works. OpenAI has reported that a global DTC brand doubled its click-through rate and halved its cost per click after switching from standard ChatGPT ads to feed-based ads. Feed campaigns have been among the strongest-performing formats in the programme to date.

The speed of this rollout is worth a moment of quiet appreciation. OpenAI launched advertising in ChatGPT in early 2026. In roughly five months, the platform has moved from impression-based pricing to CPC campaigns, introduced a conversion tracking pixel, built out an ads manager, and now added product feed campaigns. Each of those steps took Google and Meta years. OpenAI has compressed the timeline by adopting proven infrastructure rather than inventing new mechanics — making onboarding deliberately frictionless for advertisers who already manage feeds elsewhere. It's the advertising equivalent of moving into a fully furnished flat. Why build the kitchen when someone else already installed the cabinets?

Google's Building Mini Storefronts Inside Search (And It Changes the Game)

Google's Sponsored Shops test, spotted in Shopping results this month, takes a completely different approach to the same underlying idea. Instead of individual product listings scrapping it out against each other — the standard Shopping format we all know and tolerate — Sponsored Shops groups several products from a single retailer into one branded ad unit. The unit displays the store name, seller ratings, and a curated product selection. It's basically a pop-up shop sitting inside search results.

The shift from product-level to store-level competition is significant. In standard Shopping, you win a placement by having the better feed data and the stronger bid for a specific product. In Sponsored Shops, the competitive unit is no longer the individual listing — it's the retailer. A competitor with a deeper catalogue, stronger seller ratings, and better brand presentation holds a structural advantage that bid optimisation alone cannot offset. You can't out-bid your way past a better shop.

Google hasn't confirmed a broad rollout, and the format remains a test. But it's consistent with the direction they've been building toward. In 2025, Google introduced Merchant Brand Profiles, which let retailers build brand-presence pages in search with lifestyle images, videos, and business descriptions. Sponsored Shops looks like the next logical step — bringing brand identity directly into the Shopping ad unit. Google's own communications about its 2026 priorities describe the goal as making search "a more powerful tool for discovery, where ads can inspire and answer all at once." (Which sounds like something a platform says right before it starts charging more for everything.)

The measurement implications are also worth flagging. A single ad unit with multiple clickable elements — store name, individual products, ratings — creates multiple potential click paths. How traffic splits across those paths, and how that maps to existing attribution models, is an open question that Shopping advertisers should be thinking through before the format scales. Attribution modelling was already giving us headaches. This promises migraines.

Here's Where It Gets Genuinely Fascinating

Step back from the specifics of either announcement and the structural pattern becomes clear. Google, Meta, Amazon, TikTok, and now OpenAI have all converged on the same fundamental input: a structured product feed. The file formats differ — field names, required attributes, validation rules vary by platform — but the underlying logic is identical. You maintain a machine-readable catalogue with standardised fields for title, description, price, image, availability, and category. The platform ingests it and uses it to generate, target, and serve ads.

What's changing is where the feed sits in that chain. Historically, the feed was the data layer. The platform read it, processed it, and built something on top of it — a Shopping ad, a carousel, a sponsored listing. Your job was to keep the data accurate and let the system handle the rest.

Both of this month's announcements move the feed closer to the surface. In ChatGPT, the feed data is the ad creative — titles and descriptions served directly to users mid-conversation, with no intermediate creative layer. In Google's Sponsored Shops, the feed data is the storefront — product imagery, titles, and catalogue depth presented as a branded unit representing the retailer as a whole. In both cases, the quality of your structured product data isn't just an input to the advertising system. It's visible to the consumer. It's the thing being judged.

That makes feed management a fundamentally different activity than it was even twelve months ago. It's no longer about keeping a data file accurate enough to pass validation. (The "does it clear the minimum bar so the platform stops yelling at me" approach.) It's about ensuring that every product title reads clearly, every description differentiates the product, every image holds up when displayed alongside an AI-generated response or inside a branded storefront. The feed is the copy. The feed is the creative. The feed is the brand presentation.

What This Means If You Have a Big Catalogue

For sellers with broad product ranges, both developments play to existing strengths. OpenAI's feed campaigns are explicitly designed for catalogues of 1,000 to 2 million products — the kind of breadth where automated matching at scale delivers the most value. Google's Sponsored Shops format rewards assortment depth, giving retailers with comprehensive catalogues a more compelling storefront than competitors showing three products where a rival shows ten.

The compound effect matters. A large-catalogue seller who has invested in comprehensive structured data, high-quality imagery, and consistent product information across platforms now holds a strategic advantage that multiplies with every new surface reading from the same feed. The same data discipline that improves Google Shopping performance also powers stronger ChatGPT feed ads, better Meta catalogue campaigns, and more competitive Amazon listings. Feed quality isn't a platform-specific concern anymore — it's a cross-platform asset, and it compounds like interest on a savings account nobody told you about.

The practical priority: consistency. Ensuring product titles, descriptions, and imagery are optimised not just for one platform's requirements but across the full set of surfaces where feed data is now the presentation layer. A title that works for Google Shopping but reads like gibberish in a conversational ChatGPT ad, or imagery that performs well in a standard listing but looks thin inside a Sponsored Shops storefront, creates gaps that competitors will exploit.

What This Means If You Don't

The honest picture for smaller sellers is mixed, but it's not a funeral.

OpenAI's 1,000-product minimum excludes them from ChatGPT feed campaigns entirely. Google's Sponsored Shops, if it scales, favours the assortment depth that smaller catalogues cannot match by definition. These are real constraints, and they're part of a broader pattern in which AI-mediated advertising consolidates advantages for operators with scale. (We've seen this film before. It usually ends with "enterprise pricing available upon request.")

But standard Google Shopping isn't going away. Product-level competition — where the best feed data and the strongest bid win the placement — remains the primary battleground, and smaller sellers can still compete effectively there. The same is true for Meta catalogue ads, which impose no minimum product threshold, and for Amazon's advertising ecosystem, where individual listing quality still drives results.

Here's the thing: feed quality per product arguably matters more, not less, when your catalogue is small. A 200-product feed with precise titles, strong imagery, and complete attributes outperforms a sprawling 5,000-product feed with mediocre data on every surface that doesn't impose a minimum threshold. Smaller catalogues are easier to audit, easier to optimise, and easier to keep consistent — advantages that are worth more now that feed data is increasingly the layer the consumer actually sees.

Seller ratings and brand signals are also store-level assets that don't require a massive catalogue to build. If Sponsored Shops scales broadly, a smaller seller with strong reviews, a clear brand identity, and a well-built Merchant Brand Profile is better positioned than a large seller with weak ratings and no coherent brand presentation. The format rewards brand clarity, not just breadth. Which, if you've been doing the hard work of building a brand rather than just listing products, should feel like vindication.

The Bottom Line

The product feed has graduated from an operational data file to the primary brand presentation layer in digital advertising. Google is building store-level formats around it. OpenAI is using it as the entire ad creative. Meta, Amazon, and TikTok already treat it as the foundation of their retail advertising ecosystems.

For ecommerce sellers — whether you manage fifty products or fifty thousand — the implication is the same. The feed is no longer something you maintain. It's something you compete with. And the sellers who treat their product data as a strategic asset rather than an administrative chore are the ones building for where advertising is going — not where it's been.

That product title you wrote in a rush between meetings? It's not hiding behind the scenes anymore. It's the first thing a customer reads. Might be worth a second look.

• Canva clears the calendar for 5,000+ staff to explore AI for a week and finds the bottleneck is behavior, not tools: people stuck to familiar uses until they learned to write and test specs like developers.

• A bug in Meta's AI customer-service chatbot let hackers reset passwords on 34,000 Instagram accounts just by asking, breaching 20,000 including Obama's old White House account; Meta blames back-end checks, not the agent itself.

• Across 6 experiments people feel moral disgust and are 24.6% less likely to recommend a brand once they learn an emotional message was AI-written, with the backlash sharpest when AI text is signed by a human.

• Google tests a Chrome flag routing all omnibox queries straight into AI Mode and bypassing standard Search; it's Canary-only, off by default, and Google says there are no plans to ship it, calling the appearance an error.

• AI clones split the creator economy between those licensing digital twins for brand deals and those finding unauthorized copies of themselves;

• Amazon lets shoppers design print-on-demand merch from AI prompts via Alexa in its Shopping app, undercutting platforms like Redbubble by turning AI-generated products into just another shopping option.

🖥️ Cursor AI - native code editor and pair programmer for shipping faster.

🖼️ ImagineArt - All-in-one image, video and audio generation workspace.

🌐 Clay - Enrich and scrape web data to build hyper-personalized prospecting lists.

🤖 Raycast - Mac launcher with AI, clipboard history, snippets and extensions.

😍It was great, all I needed

🙂Good, not great

🙄It sucked