Otto Just Put Its Entire 19-Million-Product Catalogue Inside ChatGPT
Updated: 1 day ago
Otto launched a ChatGPT plugin exposing around 19 million products to AI agents and began testing ads inside ChatGPT, reported 20-21 October 2026.
The catalogue goes where the shopper asks
Otto has done something quietly radical. On 20-21 October 2026 the Hamburg retailer confirmed it had launched a ChatGPT plugin that exposes roughly 19 million products to AI agents, and that it is separately testing advertising inside the assistant.
The logic is simple and uncomfortable in equal measure. A growing share of product research no longer starts on a retailer's own site or even on Google. It starts inside a conversation with an assistant. If Otto's inventory is not legible to that assistant, Otto is invisible at the exact moment a purchase intention forms.
By plugging nineteen million items directly into ChatGPT, Otto is choosing to be found wherever the shopper is researching, rather than fighting to drag that shopper back to otto.de first. This is the practical meaning of agentic commerce: the store now comes to the query.
The scale is not incidental to the strategy either. Nineteen million products is the kind of long tail that a human shopper could never navigate but an AI agent can filter in an instant. Otto is effectively betting that breadth, not curation, is its advantage inside an assistant, because the agent, not the shopper, does the sorting.
Owning the shelf versus renting the assistant
There is a strategic tension hiding in this move. For two decades the retail orthodoxy has been to own the customer relationship end to end, from discovery to checkout to data. Feeding your catalogue into a third-party assistant hands part of that discovery layer to a platform you do not control.
Otto appears to have decided that absence is the bigger risk. A shelf nobody can see does not protect a relationship; it simply loses the sale to whichever competitor did make its range machine-readable. Presence inside the assistant, even on someone else's terms, beats purity.
The parallel is retail media. Brands once resisted paying a retailer to advertise on its own platform; now it is a multi-billion-euro category nobody can ignore. Agentic commerce may follow the same arc, and Otto is positioning to be early rather than defensive.
What global retailers should copy
The transferable action is a question every retail and e-commerce leader should put on the agenda this quarter: is our product data structured, complete and machine-readable enough for an AI agent to recommend us accurately? Nineteen million items is a scale story, but the underlying discipline is data hygiene, not size.
The brands that win agentic commerce will be the ones whose feeds are clean, whose attributes are rich and whose availability is live. Ragged, incomplete catalogue data is now a visibility problem, not just an operations one.
The second action is to treat assistant presence as a channel to measure, not a science experiment to admire. Otto is testing ads inside the same surface, which means it is already thinking about how discovery and paid placement interact. Retailers who wait for the category to mature will arrive to find the best placements already taken.
There is a defensive dimension too. If the biggest domestic retailers make their catalogues legible to assistants and a smaller competitor does not, the assistant will simply route the recommendation to whoever it can read. Invisibility inside the agent compounds quietly, with no dramatic moment of lost traffic, only a slow leak of consideration that never shows up in a site-analytics dashboard because it never reached the site at all.
Quick FAQ
What did Otto launch inside ChatGPT?
Otto launched a ChatGPT plugin that exposes roughly 19 million products to AI agents, and it is separately testing advertising inside the assistant. The news was reported on 20-21 October 2026.
What is agentic commerce?
Agentic commerce is shopping mediated by AI agents that research and recommend products on a user's behalf. Otto's plugin makes its catalogue legible to those agents wherever the shopper asks.
Why does this matter for other retailers?
It signals that clean, complete, machine-readable product data is now a visibility requirement. If an assistant cannot read your catalogue accurately, it cannot recommend you.
Otto has grasped something a lot of retailers are still resisting: a shelf no AI agent can read is a shelf nobody sees. Handing your catalogue to an assistant feels like losing control, but absence is the greater loss. The real work now is unglamorous, clean, complete, machine-readable product data. — Daniel Nikolla, Founder of Merx Marketing
In agentic commerce, visibility belongs to the retailer whose product data an AI agent can actually read, not the one with the prettiest site.
Is your product catalogue clean and complete enough for an AI agent to recommend you correctly?
If you would like to know more about this topic, please contact us on danieln@merxmarketing.co.uk
Sources: Horizont, W&V, Meedia
Related reading
Written by Daniel Nikolla, Founder of Merx Marketing Ltd and Marketing Minute
