Google’s UCP is making AI shopping real. Is your product data ready?

Google’s Universal Commerce Protocol (UCP) is an open framework designed to enable AI-driven commerce. It allows AI systems to discover, compare, and purchase products on behalf of users.

Instead of traditional browsing journeys where customers search, click links, and navigate websites, UCP enables AI assistants to interpret user intent and surface products that best match that intent.

For brands and retailers, this means product discovery increasingly happens inside AI interfaces rather than on traditional websites.

Agentic commerce refers to a model where AI agents assist or act on behalf of shoppers throughout the buying journey.

Instead of manually searching for products, customers describe what they need and AI agents:

• interpret the request
• evaluate available products
• recommend options
• potentially complete the purchase

This shifts commerce from keyword-driven discovery to intent-driven recommendations.

In AI-driven commerce, product data is the foundation that allows machines to understand and recommend products.

Traditional product pages were written for human browsing. AI systems need structured information that explains:

• what the product is
• who it is for
• how it should be used
• how it compares to alternatives
• what problems it solves

If this information is missing or unclear, AI systems may not recommend the product at all.

AI-ready product data includes more contextual information than traditional product listings.

Important data elements include:

• product attributes and specifications
• usage scenarios
• customer questions and answers
• care and maintenance information
• certifications and materials
• compatibility and related products

The more context AI systems have, the better they can match products to customer needs.

AI commerce affects multiple functions across an organization.

Successful initiatives typically involve collaboration between:

• marketing
• e-commerce teams
• product data teams
• customer experience teams
• technology and architecture teams

Because AI shopping affects discovery, merchandising, customer relationships, and technology infrastructure, companies benefit from approaching it as a cross-functional initiative rather than a single department project.

Niclas Mollin:
Okay, so welcome to this live webinar.

Today we’re going to talk about agentic commerce in general, and specifically Google’s UCP.

I’m co-hosting this today with Erik Wikander, CEO and Co-Founder from Wilgot. Welcome, Erik.

Erik Wikander:
Thanks. Great to be here. How are you doing today?

Speaker 1:
I’m good. It’s snowing outside in Stockholm. I don’t know about Malmö.

Erik Wikander:
Sunny, as always.

Speaker 1:
Yes. Before we kick things off, there are a few practical things. You have the opportunity to use the chat and ask questions, and we’ll try to pick them up during the session. If not, we’ll have a Q&A at the end. So please interact and give us different angles we can build on.

Maybe, Erik, we should start with who you are and why you’re interested in this space.

Erik Wikander:
Sure. I’m Erik, one of the co-founders of Wilgot.

My background is that I used to be a CMO in fintech. We spent a lot of money on paid marketing — probably too much — and we started asking ourselves what a different route could look like. That’s how I got into the whole SEO rabbit hole.

We managed quite well. After working on it for about a year, we got around 90% of our customers through SEO. That company was later acquired, and I wanted to build my own company. Around that time, ChatGPT came along, and I started thinking: what if we could solve the problems I had experienced with SEO and content generation using AI?

So the idea was to combine content generation with SEO research to allow people to actually write toward what people are searching for. One of the things I experienced firsthand was that there was a big divide between search behavior and content generation.

At that time it was still GPT-3.5, so the content that came out wasn’t very good. It was content, but not very good content. A lot has happened since then.

About a year and a half ago, we started seeing that people were beginning to treat ChatGPT like a search engine. That made us think: what happens the day AI takes over discovery from humans? Because that requires something completely different. It requires a new piece of infrastructure that allows products to be understood and recommended by machines rather than humans.

All of the content on the internet has been written by humans for humans. But now we need to rethink that, because we’re entering a new era.

The last few months have been particularly interesting. We were quite early in spotting that the market would go in this direction, and now that view has really been reinforced by all the updates coming out and by the data showing that this transformation is happening in real time.

That’s also why, when I met you maybe seven or eight months ago, it felt like we shared the same vision of where the market was heading. We also felt that Occtoo plus Wilgot was a fantastic combination because we come from different worlds, but they fit together really well.

So when you suggested doing this webinar, we were super excited. I think this shift — and especially the UCP announcement — kind of flew under the radar. Most people didn’t really stop to reflect on what it actually means for the market. And that surprises us, because this is a fundamental shift. It’s already happening in real time in the US, UK, and Australia.

Niclas Mollin:
I think the sound was breaking up a little for me there, but it’s great that we could do this together.

I’ve been working at the intersection of web, product data, and commerce since 2005 — so for over 20 years. I’m super excited about this. When e-commerce first took off, a big part of it was still mail order. And I see this shift as even bigger.

But the audience is here to learn more about UCP. Erik, I know you’ve prepared a few slides to help us understand the opportunity here. Would you like to take it away?

Niclas Mollin:
Or it seems like we lost Erik.

Erik Wikander:
I’m back.

Speaker 1:
There we go. You scared me a little there.

I was just saying: if you want to take it away and share with the audience what UCP is and how they can benefit.

Erik Wikander:
Yes. Can you see my screen?

Speaker 1:
Absolutely.

Erik Wikander:
Nice. Let’s kick off.

The first thing we usually say is that the times we live in are interesting because everything we know, to some extent, no longer applies. Very few companies actually realize the extent of this change.

That’s why, when we meet brands, most of them aren’t really prepared for this shift. But once we explain our view of the world, the shift from not understanding to understanding happens very quickly. They realize they need to act.

In our view, this is a fundamental transformation, and it needs to spread through the entire organization. Everyone inside your company — if you’re in e-commerce — needs to be aware of this shift.

To understand what’s happening, though, we need to take a step back.

In the last 12 months, 60% of users searching never reach a website. That means 60% of clicks on the internet have disappeared. And they’re not coming back, because of AI.

AI fundamentally changes the user journey. We used to search for something, get results, click the result, and go to the website. Now AI takes over that part. It gives you the information you’re looking for in a summarized way, so there’s no need to go to the website.

This shift started in other verticals first — informational queries like “How do I fix this drain?”, “How do I repair my car?”, or health-related symptom searches. Companies in those sectors have already been hit quite hard. As an acquisition channel, SEO has in many ways died there.

But the interesting thing is that commerce has been safe until now, because product discovery is one of the most complex use cases for search.

What’s driving this overall change is the shift from keyword-based queries to conversational and contextual search.

In the old world, you might type “blue shirt” into a search box. You entered what you expected the search engine could return. That keyword-based way of searching defined the early days of search and really lasted up until recently.

But as people got used to AI like ChatGPT, they realized they could just describe what they actually want. They could explain their need, and the AI would give them answers.

That conversational way of searching is where we’re headed, but it’s only the start. The next phase is what we call complex, multi-intent, contextual search.

If you search for something and I search for something, we may get completely different results. That’s the contextual and personalized part of search, and it’s fantastic for users because it makes search much more relevant.

But it also creates a completely new challenge for brands trying to optimize for how people search.

In the old world, the e-commerce game was about ranking keywords. You wanted your category page to show up at the top. If you ranked high, you got customers almost for free.

Now, this opens up a kind of blue ocean opportunity. If you’re selling niche products, there are customers with niche needs — and this new world allows you to reach people you would never have reached before.

That applies both to organic and paid. All of this is changing at once.

If discovery was the prelude to what’s next, then I think we got the real signal at NRF in January, when Google and Shopify announced the Universal Commerce Protocol, UCP.

This also went almost unnoticed by many people, which surprised us. We’re deep in this space, so maybe we read more than most, but what’s important is this:

This is not something coming in two or three years. It’s already live in the US, Australia, and the UK, and it’s planned for 170 markets in the first half of this year.

The difference between ChatGPT and this protocol is that Google is the ground that e-commerce companies and brands are already standing on. You are already reliant on that ecosystem for acquisition.

So this is not just something sitting on top of commerce. It’s changing the ground underneath it.

That’s why, when we looked at UCP, we felt it was more extensive and more deeply thought through than other protocols.

Niclas Mollin:
I wanted to ask you about that. Why do you think UCP will be more successful, and what’s the difference? Can you help the audience understand?

Erik Wikander:
I think the protocols are similar in terms of what they want to achieve, and we’ll talk more about that. But the main difference is that Google has been doing commerce for 28 years, while OpenAI has been doing it for maybe three.

In our view, OpenAI was the spark that lit the fire. But if you look at Google’s position over the last few years, it’s clear they still have an enormous advantage. They invented much of this technology. They have the ecosystem, they have the data, and they have the world’s biggest shopping graph.

That’s important, because even ChatGPT currently draws from Google’s shopping graph when it shows products. There isn’t yet a mature alternative feed.

We often get the sense that many brands think ChatGPT is what’s going to change commerce, without realizing that Google is already deeply embedded in how commerce works today.

I think one of Google’s major challenges was monetization. They needed to figure out how to make money in a world where AI finds and recommends things for you. But when their Q4 report came out and search revenue was up significantly, it was clear that they had figured something out.

So UCP is an open-source, vendor-agnostic protocol, developed by Google and Shopify. Shopify is especially interesting because they’re involved in both ecosystems.

The way we look at it is this: HTTP standardized how browsers request information. UCP standardizes the layers of trade.

To us, this marks the end of the click era and the beginning of what we call zero-surface commerce and agentic commerce.

The big shifts are:

• moving from keyword-based search to intent-led discovery
  
  
• moving from traditional funnels into AI-driven journeys
  
  
• moving from optimizing for human attention to optimizing for machine understanding
  
  

The AI agent interprets needs and then serves products that fulfill those needs.

In the old world, someone would search, browse results, go to a website, add to cart, and purchase. Now everything can happen inside one interface — the AI assistant.

That assistant could be Google, or it could be something else.

This changes optimization completely. We used to optimize for human attention with visual content and category pages. Now we need to optimize for machines that are evaluating products.

That means your product data becomes the key to this new world. The AI has to understand your product well enough to compare and recommend it in response to extremely complex user intent.

So product data has always been important — but now it’s imperative. It becomes the technical gatekeeper for whether you even exist when AI is shopping on behalf of humans.

If you don’t have the right attributes, you’re not just limited in SEO or paid shopping. You could become invisible to these complex queries.

That means brands need to rethink the importance of product data entirely.

In our view, the most important thing you can do in this new world is optimize product data based on how people search and why people buy, so that AI fully understands your products.

In a way, AI turns your product data into a salesperson. Every product needs to be able to sell itself.

Erik Wikander:
The UCP framework has four layers. We won’t go through every detail of all of them because they’re quite extensive, but it helps to understand what UCP actually is and why it matters.

We focus primarily on the discovery layer, but the other three layers are essentially the new infrastructure that companies will also need to consider — unless they are on Shopify, which has a major advantage since Shopify co-developed this.

If you’re on another commerce platform, you may already be calling your vendors and asking them to move faster, because these capabilities are becoming essential.

The other layers are technical capabilities that allow AI to transact. But before you think about transactions, merchandising, or pricing, you need to start with discovery. If your products are not visible and understandable, there’s no point in tackling the other layers first.

So discovery is the foundational layer.

A lot of what you do for UCP also benefits other protocols. The core issue is making products understandable to machines.

The knowledge layer feeds the shopping graph. It helps the AI answer questions with certainty even when a shopper never visits your site.

And that’s one of the biggest implicit changes in all of this: you may not have customers on your website anymore.

To feed the AI properly, there are already new types of fields becoming important — things like:

• Q&As
  
  
• usage scenarios
  
  
• care and maintenance
  
  
• material composition
  
  
• certifications
  
  
• other contextual information
  
  

The more you can feed the AI with real customer questions and the right answers, the better it can represent your products.

The same goes for usage scenarios — who the product is for, how it’s used, and in what context. Historically, product pages have been written for humans and designed to convert quickly. In this new world, you need to optimize for context.

The more context you provide, the better the AI can match your product to a user’s actual needs.

Niclas Mollin
That’s interesting, because some of this information already exists in different parts of the company. If you produce garments, for example, some of it might already be in the PLM.

But I can also see that there’s a need to generate new attributes — the sort of AI-ready dataset you mentioned.

Would you say these are recommendations, or requirements?

Erik Wikander:
To be competitive, I would say they are requirements.

Some of this information already exists in different systems, absolutely. But a lot of it will still need to be generated or enriched. The good news is that it can often be generated from data you already have.

Today, many companies are already answering these questions — they’re just doing it in fragmented ways. The answers sit in customer reviews, customer service conversations, search data, internal documents, and other places.

One of the things we do is ingest all that data so we can generate content with confidence and help the AI answer on your behalf. But first, you need to get the data into shape.

And if you’re a larger company with a lot of SKUs, that can be a real challenge.

That’s also why we’re excited about our partnership. One of our biggest pain points is that customers often have data spread across many different places. Having that foundation prepared speeds everything up dramatically.

Erik Wikander:
The next step is the actual buying part.

Zero-surface commerce is the term we use for the moment when a transaction happens inside the AI itself.

This is one of the biggest things with agentic commerce and UCP: the fact that buying can happen inside the AI. Google demonstrated a fully agent-led purchase flow quite a while ago, but it took time to operationalize.

This completely changes the funnel. There is no website sitting on the right side of the process anymore. That means the role of websites will fundamentally change. Websites become containers for your products and content — and, in a way, servants of the AI.

Your job becomes making the AI the best possible salesperson for your products.

That has major implications.

Think about loyalty. What happens when the AI buys and the customer never visits your website? How do you create loyalty in that model?

The good news is that Google has confirmed that the merchant remains the merchant of record, which means you still get the customer data needed to serve the customer. But the way you serve them will be very different.

Another example is the cart. Today, a customer visits your site, sees upsell options, adds multiple products, and increases average order value. But in an AI-driven flow, that behavior changes.

Google obviously also has an interest in solving this. If average order values are too low, the economics break. So it’s in their interest to make bundling and upsell work inside the AI environment.

I think one of the most interesting emerging areas will be optimizing product relationships and bundles in a machine-readable way, so the AI can build higher-value baskets.

Niclas Mollin:
I’m looking at the time here, but let me ask something important.

Some people in the audience might think this is something that will happen at the end of the year, or next year. Should they wait to act, or should they act now?

Erik Wikander:
You can wait — but I don’t think it’s a good idea.

Based on the rollout plans that have been announced, this is already live in several markets and will reach 170 markets in the first half of 2026, including the EU.

We also have customers already seeing these shifts in the US. So this is not theoretical.

And even if you don’t start with full agentic commerce, there is already a huge opportunity in the discovery layer. You can improve how AI finds and compares your products in tools like Gemini and ChatGPT.

That alone is reason enough to act now.

The first step is to make sure AI can actually find your products.

This new Google AI mode combines generative search with the shopping graph and personalized product grids. It allows users to ask follow-up questions, which makes search feel almost like a conversation with your product catalog.

And yes — it’s coming, and it’s coming quickly.

Erik Wikander:
The benefits are clear:

• higher conversion potential
  
  
• better discovery for niche and long-tail products
  
  
• stronger ability to compete on relevance, not just budget
  
  
• direct customer relationships remain intact
  
  

But there are also serious risks:

• website traffic will decline
  
  
• some traditional digital tactics will lose value
  
  
• brands with weak product data could become invisible
  
  

Erik Wikander:
To explain what we do in this context:

We upgrade product data so that AI agents can find, recommend, and sell products. One of our core theses is that product content has to move from being static to dynamic.

We call this self-optimizing catalogs.

We’ve built models that continuously analyze search intent, buying intent, and product visibility, and then optimize product data at scale.

Every product page effectively becomes its own salesperson.

We ingest product data, customer review data, customer service data, search data — and soon also AI conversation data — and use all of that to detect patterns, generate optimized content, and improve merchant center feeds so products remain visible and recommendable over time.

Niclas Mollin:
I think that’s really interesting, because right now in the market you have PIMs and other software categories that claim to do this.

What’s your point of view on why they don’t really reach the same results?

Erik Wikander:
That’s a very relevant question.

We do see a lot of PIM systems adding AI content generation features. But our core thesis is that content needs to go from static to dynamic.

It’s one thing to generate content. It’s another thing entirely to make that content perform.

To do that, you need to integrate all the relevant signals and continuously improve the content based on performance. That’s the big difference.

I sometimes use the analogy of entering a Formula One race in a Fiat Panda. You can enter, but you probably won’t win.

We’re extremely focused on content performance, and the only way to make content perform is to continuously ingest and respond to data.

That makes this feel less like traditional product content management and more like a performance marketing discipline.

Speaker 1:
That’s really good.

So before we go into questions, maybe we should show the audience a little more about why Wilgot and Occtoo are such a strong combination.

We’ve been using the analogy of heart and brain. We really are made for each other in a good way.

The Occtoo part is the heart — gathering and orchestrating data from many different places. Wilgot is the brain — optimizing the content so that it can be reused and repurposed across all the different sales channels a company has.

I think it’s a very interesting partnership.

As something we can offer the audience, we also decided to provide a health check. If you don’t really know where to start, you can reach out to me and Erik and we can offer a health check around agentic commerce and UCP to see how ready you are to embrace this opportunity.

Erik Wikander:
And what makes this so interesting is that it spans so many domains.

It’s not only about getting your product data in order. It’s also about rethinking your marketing, your customer experience, and your internal collaboration.

The best conversations we have are with cross-functional groups — marketing, tech, customer service, loyalty, merchandising. Everyone has their own angle on this shift.

That’s also why companies often feel confused. It’s a broad and complex topic.

Niclas Mollin:
Perfect. I can see the questions now.

The first question is:

“If UCP shifts the goal from capturing attention to satisfying agent fulfillment, does media spend move from a CPC model to an agent commission or CPA model?”

Erik Wikander:
That’s a very interesting question.

I think in theory that’s where it should go. When OpenAI announced shopping, they mentioned commission-based models, which makes a lot of sense in this kind of environment.

The challenge is that they also layered in impression-based charging, which becomes the worst of both worlds because you pay for both exposure and outcome.

But over time, I do think it’s possible that commerce shifts toward more commission-based or performance-based models. It would be a much better value proposition for brands — you only pay for what you get.

Whether or when that happens is still open, but I think it’s a very logical direction.

Niclas Mollin:
Another question:

“With the transaction layer moving offsite, how do premium brands maintain their identity beyond just price and specs?”

Erik Wikander:
That’s a great question, and I think it points to the need for an entirely new discipline — something like agentic brand positioning.

The first step is for brands to understand themselves very clearly. How do you want to be perceived? What do you want to be associated with?

The only things you can really influence are the images, the copy, and the product data that the AI has access to. So it becomes critical to ensure that your brand values come through in every product listing and every representation of the product across every market.

Luxury brands are especially interesting here. We’ve had conversations with a few of them, and many luxury brands don’t like overly verbose product pages. Their premium feel often comes from restraint and minimalism.

So the question becomes: what if the AI knew enough to subtly recommend a product without needing huge amounts of visible copy? That may be where this goes — richer machine understanding behind the scenes, without sacrificing brand expression on the surface.

But it requires a very clear internal understanding of your brand.

Niclas Mollin:
There’s a follow-up question:

“Do you see this model working for all of e-commerce, or is it better suited for some industries than others?”

Erik Wikander:
I think it will affect all of e-commerce, but in stages.

Commodity products will probably be the first to shift fully into AI-driven transactions. If I’m buying aspirin or a carton of milk, I may not care much who fulfills the order or what the branding looks like, as long as the need is met.

But for more complex categories, where discovery is more research-driven, the transition may take longer. People may still ask more questions, seek more reassurance, or spend more time validating the right choice.

So yes, all sectors will be affected — but not all at the same pace.

Niclas Mollin:
When we talk about this partnership, we’ve also discussed who should actually own this inside an organization.

Right now, it seems to come partly from tech, but there’s also strong interest from marketing.

Who do you think should own this?

Erik Wikander:
I think the success factor is involvement from at least both marketing and tech.

But the most productive conversations we’ve had are often driven from the marketing side, because this is fundamentally also about protecting and improving performance in existing acquisition channels.

If you think of this as an investment, you’re already spending money in Google Shopping and related channels. This is both about protecting that investment and improving the performance of your product content.

And beyond that, creating product content has never been easy. It’s always been a major task, especially across multiple markets, with many stakeholders involved.

So I think companies should use this moment to think more holistically — not just patching something together, but rethinking how they produce and manage product content altogether.

Niclas Mollin:
There’s another question I’ve been thinking about:

Will AI-generated content be good enough?

Erik Wikander:
That question has changed a lot in just the last 12 months.

I’ve worked with AI-generated content since GPT-3.5, and what we have today is something completely different.

In the latest studies we’ve seen, AI-generated content can outperform human-written content in many contexts — partly because AI is very good at structuring information in ways that machines and search systems can interpret well.

We’re already at a point where AI-generated content is on par with, and sometimes better than, what humans can produce — especially if it is guided by the right data and intent.

So I think if companies dismiss AI-generated content outright, they risk putting themselves in a weaker position than those who adopt it properly.

Niclas Mollin:
That’s reassuring. Many companies are still afraid of being penalized for using AI.

And one final question:

“What happens to social retargeting if no one enters the website?”

Erik Wikander:
Yes — that’s going to be tricky.

There are a lot of things that stop working in the same way if people no longer visit the website. Retargeting is one of them. So there will definitely be a need to rethink a lot of website-based mechanics.

But what’s interesting is that UCP is not really limited to Google.

If you think about Instagram, for example, the algorithm is already very good at surfacing products you’ve never seen before. So what happens if you can also buy directly inside that interface?

That’s not hard to imagine. And if that happens, it may well run on the same kinds of underlying protocols.

So in many ways, these ecosystems may start to merge. Social, discovery, and transaction become more tightly connected inside AI-powered interfaces.

Niclas Mollin:
Sounds good.

Erik, it’s been a pleasure. You really know your stuff, and I’m very happy about this partnership.

And to everyone in the audience: if you have more questions, don’t hesitate to reach out to either me or Erik.

Hopefully we’ll also see many of you from the Nordic audience at D-Congress, where we’ll be attending.

Thank you very much, and have a great day.

Erik Wikander
Thanks.