“These flyers need to be in Antwerp by Thursday, what will that cost?” Since late August, a ChatGPT plugin has been answering questions like this with an order-ready configuration. The Dutch online print platform Print.com has developed its own plugin for ChatGPT. The connection reportedly runs as MCP, meaning it works for any AI assistant, not just ChatGPT.
Its own plugin in ChatGPT, free for all Print.com customers, running in both the free and paid versions of the AI assistant. The company announced the move rather casually: an entry in the product notes at the end of August, followed by a post on its own LinkedIn channel in early September. Print.com is now available in every conversation with ChatGPT, the post says. Anyone submitting a request doesn’t get a product list back, according to the post, but exactly the configurations, including paper and production method, that will actually meet the requested deadline, plus prices and links to the matching product specifications. The order is then placed in the web app, via a shareable link that the plugin is said to provide.
This goes beyond a price calculator. Print.com described the scope of the feature set to the British trade magazine PrintWeek as follows: in the chat, the complete product catalogue can be searched, including thematically, whether that’s weatherproof products for an open-air concert or recyclable products for an event that has made sustainability its mission. On top of that, delivery times are checked, calculated backwards from the delivery date, and order status is queried. A connection to the company’s own CRM or ERP is also planned.
Not Just for ChatGPT
Print.com writes in its LinkedIn post that, over the past few months, it has made billions of product combinations on the platform readable for AI — “not just for ChatGPT, but as MCP: the open standard that lets any AI assistant [connect to Print.com, ed.],” as Daan de Jong, Lead AI at Print.com, is quoted as saying.
While a plugin behaves like a sales channel, an MCP server can be seen as an interface that any agent can dock onto — today ChatGPT, tomorrow Claude, Gemini, Copilot, or the purchasing bot of an agency sourcing trade-show materials for three clients at once. A comparable plugin for Claude is reportedly already under review, according to the PrintWeek report.
That’s exactly what the billions of product combinations are about — not a chat interface, but a body of data from which binding statements can be derived. Co-founder and Chief of Strategy Lucas Baarspul frames the expectation behind it this way, according to PrintWeek: a growing share of all print enquiries will in future begin with an AI assistant. “We want that question to land on a real specification, not an estimate. With our own production sites and more than 200 print partners, we know what can be produced where, tomorrow.”
That, incidentally, is exactly what Initiative Online Print, together with international partners, is also trying to achieve by developing a print vertical for UCP (Universal Commerce Protocol).
Why This Is More Than a Product Announcement
Anyone who has followed the past few months closely will recognise a pattern. At the end of August, Adobe made around 70 tools directly available inside ChatGPT. Cimpress has been supplying the print production for Canvas Print Shop since the summer. And now an online print platform is moving not just design into the AI assistant, but calculation, scheduling, and ordering too.
A similar development is emerging from another direction too. In mid-September, Printess unveiled an AI assistant that builds a sales-ready web-to-print product from a single prompt, complete with form fields, logic, and a buyer interface. So on one side, the customer enquiry becomes the prompt; on the other, it’s the template work.
For print shops, this shifts the question of what their actual edge consists of. A shop with a clean configurator was a competitive advantage for ten years. But if the enquiry no longer starts in the shop at all, something else decides it — whether product data, pricing logic, and delivery times are machine-readable. Being retrievable alone isn’t enough, though. If an agent promises a delivery date, the knowledge of which machine is free until when has to be behind it — otherwise the answer comes quickly, but it’s wrong.
Four things decide whether a print shop shows up in an agent’s answer at all — and whether what’s stated there can actually be delivered on:
- Product data that clearly describes variants, substrates, and finishes, instead of hiding them in body text
- Pricing and scheduling logic that can be queried as an interface, not just calculated in the frontend
- Capacity knowledge that knows what’s feasible by when — because an agent doesn’t forgive embellished delivery promises
- An open connection, so that the business’s own offering shows up wherever the customer happens to be asking
MCP Connects, But It Doesn’t Translate
While individual providers build their own access points, a second development is unfolding that mostly gets lost in the coverage of plugins. MCP governs how an assistant talks to a system: connection, tools, data access — all of it bilateral. Every provider describes its own world in its own terms, and an AI agent has to learn its way around exactly that particular world.
But what actually gets exchanged is governed by other protocols. In January, Google, together with Shopify, introduced the Universal Commerce Protocol (UCP); catalogue, shopping cart, and identity linking have been running live on Google surfaces in the US market since spring. OpenAI and Stripe counter with the Agentic Commerce Protocol (ACP), which ChatGPT also builds on. Both have their origins in retail. Both understand size, colour, variant, stock level. Paper weight, finishing, 4/4, colour accuracy, preflight requirements, or a delivery date that depends on machine scheduling? Nowhere to be found.
To close this gap, an initiative came together in May 2026 under the leadership of Initiative Online Print (IOP), together with the Bundesverband Druck und Medien (“German Printing and Media Industries Federation,” bvdm) and Intergraf (“European Federation for Print and Digital Communication”). Since then, dpsuisse, VIGC (“Flemish Innovation Centre for Graphic Communication”), and CIP4 (International Cooperation for the Integration of Processes in Prepress, Press, and Postpress) have joined the project, along with the PRINTING United Alliance and the Ghent Workgroup in July, and APTech (Association for PRINT Technologies, also from the US) in August. The goal is a print-specific vertical for UCP, with submission announced for the fourth quarter. CIP4 and the Ghent Workgroup are contributing what the commerce protocols are missing. Because a print order doesn’t end at checkout, it has to be translated into print data and a job record — the Ghent Workgroup maintains the specifications for print-ready PDF data, CIP4 the formats for job data.
Print.com and the effort to develop an industry standard in the form of a UCP print vertical aren’t standing in each other’s way, then, they sit on different levels. A technical connection can later be switched over to a different protocol; the terms in which a product is described remain.
Only the shared vocabulary isn’t finished yet, while the first platforms are already delivering. If agents spend two years learning how a single provider names papers, deadlines, and surcharges, a later standard will have to accommodate those terms, rather than setting them.
A vertical always extends a specific protocol, in this case UCP. The terms the vertical defines, however, aren’t tied to UCP — whether they can later also be mapped onto ACP remains to be seen.
No Numbers, No Ad Space
How many Print.com customers are actually already starting their print orders in the ChatGPT chat remains an open question — no solid usage figures for the plugin are available. Print.com itself sees the plugin as the start of something bigger; the company says early customers are currently being brought into the further development.
More interesting than user numbers is the power question anyway. Whoever runs their own MCP access supplies the specification that an agent orients itself by. Whoever doesn’t run one doesn’t show up in the AI agent’s answer.
Visibility can certainly be bought: Google has been showing ads inside the AI answer itself since June, and OpenAI has been inserting advertising underneath it since August. But an ad slot doesn’t replace a connection — anyone can be advertised, but only those who are connected can calculate and deliver. Smaller print shops, though, can neither build nor maintain their own interface. For them, the UCP print vertical will decide whether their offering becomes describable in the future without an individual connection.
Clean Up Your Data Instead of Building a Plugin
Few can afford their own plugin. The groundwork, on the other hand, they can: putting product data, prices, and delivery times in enough order that a machine can derive a binding answer from them. Whoever only calculates prices in the configurator, maintains delivery times as a rule of thumb, and stuffs product characteristics into marketing copy has nothing an agent can work with. This clean-up is worth it, regardless of which assistant and which protocol ends up winning out. Whoever has described their products cleanly can adapt them to any standard.
My Take: Print.com is probably right: a growing share of print enquiries will in future begin with an AI assistant, and access to it has to be open. That can be useful for the initiative behind the UCP print vertical, because every connection shows what information an AI agent actually needs. For the individual print shop, in the meantime, this task remains: anyone thinking about their next shop relaunch should also ask what the AI knows about their own products — and whether it’s correct.
