Is Your Intershop Catalog Ready for AI Buyers?

For years, B2B commerce has been about the storefront. We made search better, pages faster, the customer portal smarter. All of it assumed a person would land, browse, and click.

That assumption is starting to fail. More product discovery, and sometimes the purchase itself, now happens through AI agents working on a buyer’s behalf: ChatGPT’s shopping features, procurement copilots, and assistants like Intershop’s Copilot for Buyers. Intershop now positions itself as an agentic B2B commerce provider, and its Spring 2026 release pushed firmly in that direction.

The issue is straightforward. An agent never sees your storefront. It doesn’t use your navigation or notice your design. It reads your data. So everything your interface normally does for a human, the catalog now has to do on its own, in a format a machine can parse and trust. A catalog built for people browsing a website is not ready for buyers who never visit it.

Three things “AI buyer” can mean

The term gets used loosely, so it’s worth separating what’s real for B2B today.

The first is discovery through tools like ChatGPT. Using the Agentic Commerce Protocol (ACP), you can expose structured catalog data so your products appear when someone asks ChatGPT for them. For B2B this is mostly about visibility and pulling high-intent buyers to your product pages, not in-chat checkout, which is still mainly a consumer behaviour.

The second is GEO, or generative engine optimization. SEO targets a search index. GEO targets how language models read and cite content, which rewards clear, structured, question-shaped writing. If an LLM can’t confidently understand your product, it won’t recommend it, and there’s no second page to fall back on.

The third is the in-platform Copilot for Buyers, and this is where agentic commerce is most useful for B2B right now. Sitting inside a customer portal and connected to your data, it can send maintenance reminders, book service appointments, flag parts that need replacing, and identify a component from a photo even when it’s worn or dirty, then start an order. For manufacturers and wholesalers with large spare-parts catalogs, the payoff is clear.

All three depend on the same foundation: clean, complete, machine-readable product data.

What an AI buyer needs from your catalog

When a machine does the reading, the requirements get specific. Product data has to be structured, not locked inside PDFs or images. Technical attributes like dimensions, materials, certifications, and part numbers need to be discrete fields, formatted consistently across the whole catalog, because agents match on attributes. A product with thirty clean attributes wins the recommendation; a similar one with three does not.

Content matters too. Titles and descriptions should carry the brand, the key specs, and the words a buyer actually uses. Formats like FAQs and how-to sections get picked up and cited far more readily than marketing copy. Every sellable product needs several clean, high-resolution images, which is what makes photo-based part identification work. Price and stock have to be accurate and refreshed often, because the agent states them as fact and the buyer can’t cross-check against your site. You also need reliable identifiers and variant modelling so the agent picks the exact SKU rather than a close miss.

Turning it on is easy. Winning with it isn’t.

On Intershop, switching on an ACP feed is mostly configuration. You set it up as a product data feed on a channel, choose locale and currency, decide how incomplete products are handled, map your images, schedule it, and register with ChatGPT. A capable team can do that in an afternoon.

So when someone asks whether they could just enable it themselves, the answer is yes. But enabling the feed and getting results from it are two different projects, and the work sits on either side of that switch. Upstream: is your data complete and rich enough that products actually surface and convert, or will part of the catalog be rejected as incomplete while the rest loses to competitors with better attributes? Downstream: are validation rules stopping bad data from shipping, are URLs clean, are images mapped correctly, and is the feed tuned to run reliably at your catalog’s size?

That space between having the feature and getting the result is where the real effort goes.

Where the work actually is

For a real B2B catalog, getting agent-ready is a data and integration project. It usually comes down to a handful of things. Auditing and enriching the catalog in your PIM so attributes are complete and normalized. Configuring the feed with custom validation and mapping. Building a GEO content pipeline, where Intershop’s Product Content Agent gives you a starting point to extend. Setting up proper retrieval for the Copilot, which needs a dedicated vector layer rather than bundled defaults. Embedding Copilot for Buyers and connecting it to your installed-base and service data. And owning the security work, since identity, secrets, secure coding, and testing for these AI features are the customer’s and partner’s responsibility, not the platform’s.

The catalog audit is almost always the largest piece, and for technical catalogs it’s where most of the value comes from.

A quick gut check

Is your technical data structured, or stuck in PDFs? Do titles and descriptions read like answers, or like internal labels? Does every product have good images? Are price and stock accurate enough to trust without checking? Would an incomplete-product filter reject a big share of your catalog today? Is there anything beyond a storefront, like retrieval and a Copilot layer? Has anyone owned security for these features?

If a few of those gave you pause, your catalog isn’t ready yet, and most aren’t. The companies that fix it first will be the ones AI buyers find while everyone else is still deciding whether this matters.

We’ve delivered more than 40 Intershop projects, including deep platform development, so we don’t stop at flipping switches. We make the catalog underneath good enough to win: PIM enrichment, GEO content, feed configuration with custom validation, the retrieval layer, Copilot enablement, and the security that goes with it. We start by auditing where your catalog stands against what AI buyers need, then build what closes the gap.

If you want to know how agent-ready your Intershop catalog is, we’ll run a short review and show you the gaps and the fastest wins.