Why your ChatGPT product feed is the most important feed you're not running in 2026

There is a search engine processing 50 million shopping queries every single day that most UK merchants aren't listed on. That search engine is ChatGPT.

While marketers debate Performance Max bid strategies and TikTok Shop commission rates, OpenAI has quietly built one of the fastest-growing product discovery surfaces in ecommerce history. And unlike every other channel that's launched in the last decade, ChatGPT's shopping feed has a specification that most feed management tools don't support yet.

The scale is already enormous

OpenAI confirmed in 2026 that ChatGPT processes over 50 million shopping queries daily. That's not search queries in general — that's queries with explicit commercial intent, where a user is asking ChatGPT to recommend a product, compare options, or find the best deal on something specific.

These aren't browsers. These are buyers.

The nature of ChatGPT shopping queries is fundamentally different from Google Search. Users don't type "waterproof jacket" — they type "I'm going hiking in the Scottish Highlands in October, I need a waterproof jacket that's genuinely waterproof, not just shower-resistant, under £200". ChatGPT then evaluates your product data to determine whether your jacket is relevant to that specific, highly-informed query.

The merchants who build complete, accurate, factually-rich product feeds for ChatGPT now will own the top positions when the platform opens fully to UK traffic.

Why it's different from every other feed

Most merchants assume they can submit their Google Shopping XML to ChatGPT. They can't.

The ChatGPT product feed is a completely separate specification using JSONL format (not XML), delivered via SFTP push (not uploaded to a merchant centre), with unique fields like is_eligible_search, is_eligible_checkout, variant_dict, and listing_has_variations that don't exist in any other feed format.

The most common mistakes merchants make when they first build a ChatGPT feed:

• Including HTML in descriptions. ChatGPT rejects any HTML markup in the description field. Every <p>, <br>, <strong> tag will cause the product to fail validation.
• Incorrect variant_dict. If you include a colour dimension in variant_dict for a single-colour product, OpenAI's engine generates all possible colour combinations from your catalogue. A 7,000-product catalogue can become 22,000+ phantom products overnight.
• Non-ASCII characters. Curly quotes, em dashes, copyright symbols, and non-breaking spaces all cause JSONL parsing failures.
• Static prices. ChatGPT surfaces real-time pricing. If your feed price doesn't match your website price, your listing quality drops.

The opportunity window

ChatGPT Shopping launched initially for US merchants. UK expansion is confirmed throughout 2026. This creates a specific opportunity: merchants who build and validate their ChatGPT feed now will be positioned to activate the moment UK access opens, rather than scrambling to build it when everyone else is.

The parallel in history is Google Shopping circa 2012. Merchants who built clean, complete product feeds for Google Shopping in 2012 had a significant first-mover advantage as the channel scaled. The merchants who waited until 2015 when Google Shopping was "proven" paid dramatically higher CPCs to compete against established listings.

ChatGPT Shopping is that moment, right now.

What a good ChatGPT feed looks like

A well-built ChatGPT JSONL feed has several characteristics that distinguish it from a hastily converted Google Shopping XML:

• Clean, factual descriptions. No marketing language ("Amazing quality", "Best in class"). ChatGPT's AI evaluates factual richness — specifications, materials, use cases, compatibility.
• Correct variant grouping. Every variant of the same product shares a group_id. The variant_dict only contains dimensions that genuinely vary within the group.
• Complete required fields. All 14 required fields must be present on every product row. Missing fields cause silent rejection — your product simply doesn't appear, with no error message.
• Consistent seller context. seller_name, seller_url, and return_policy must be accurate and consistent across every product.

The bottom line

50 million shopping queries per day. A specification most feed tools don't support. A UK rollout confirmed for 2026. And an opportunity to build your position before the market catches up.

The ChatGPT product feed is not a nice-to-have. For UK merchants serious about multi-channel commerce, it is the most important feed you should be building right now.