AI agents will soon buy products on their own: structured data, feeds and consistent information will become essential to get selected
AI agents that shop on your behalf, ACP and UCP protocols, agentic commerce… I bet you hear about it everywhere. Trouble is, between acronyms and technical jargon it’s a mess to navigate and actually make sense of it all.
But today we’re setting things straight: we’ll unpack this topic in a simple, clear way, even if you’re not a nerd about this stuff.
For this episode of SEO Confidential we’ll dig into these topics with an exceptional guest: Alex Moss. And when I say exceptional, I’m not just saying it!
We’re talking about someone who has been doing technical SEO for over 25 years, co-founded FireCask, the Manchester agency that has worked with names like Expedia, the Financial Times and Beauty Bay, and is now Principal SEO at Yoast.
Oh, and as if that weren’t enough, he has also built WordPress plugins downloaded more than 2.4 million times. In short, someone who truly knows the web inside out.
For a while now Alex has specialised in a topic that is becoming more and more important: structured data, product feeds, data integrity and everything you need to be found (and chosen) by AI agents.
He writes about it for Search Engine Journal, talks about it on stages all over the world, and in this interview he explained it in a way that, cross my heart, anyone can understand.
Why does this conversation matter?
Well, because the agentic era is not some far-off science fiction: it’s already here, and anyone running an e-commerce (or simply wanting to sell products or services on the web) has to reckon with it.
Get ready, because what Alex told us is truly illuminating. Let’s go!
Agentic Commerce Optimization is aimed at a new consumer: the AI agent
More and more often we hear about AI agents that don’t just recommend a product but can go as far as completing a purchase directly. How much do you think this will change the role of SEO in e-commerce over the next few years?
The role of SEO, in itself, will stay the same: optimising wherever possible to ensure the accuracy and consistency of information. When it comes to agentic commerce, though, the perimeter of SEO expands as new protocols emerge.
Crawlability, information architecture, entity accuracy and structured data: agents rely on all of these elements, probably even more than human users do.
Large-scale adoption will take time, but the main obstacle will be humans trusting AI agents with decisions and purchases. That shift will happen over the next few years and, when it does, I predict that transactions carried out by agents will surpass those carried out directly by humans.
At that point, data integrity will be the main element an agent relies on.
I’ve started calling this discipline “Agentic Commerce Optimization“, but, just as good GEO is also good SEO, it’s the same discipline applied to a new consumer of your site: the agent.
If AI agents make decisions based on structured data, product feeds and brand reputation, what will the priorities be for an e-commerce that wants to be chosen by these systems?
Truth and consistency above all. If a product page shows one price, the feed reports another and the structured data shows a third, a human shopper might not notice. An agent will, and it will have no reason to give you the benefit of the doubt when a competitor’s data is consistent and easy to verify.
The priority has to be data integrity across every surface: unambiguous entities, correct relationships between entities, feeds consistent with the information on your pages, and availability that actually matches the product’s real availability.
Then there’s entity perception, a layer some SEO professionals tend to overlook. Agents don’t just read your site: they also ingest and weigh the information and perspectives coming from third-party sources.
You can have perfect structured data and still lose the final selection because of an inconsistency, or any other piece of information that leads the agent to a wrong interpretation and, as a result, to excluding you from consideration.
Many describe Agentic Commerce as the next big revolution in online commerce. Do you think it’s genuinely a change destined to transform the industry, or are we still in a very experimental phase?
Both statements are true. I believe it will be a major revolution, but it will take longer than we’d like. The main barrier isn’t the technology, it’s consumer adoption.
The protocols themselves are still extremely young: both ACP and UCP launched within the last 12 months. The big tech companies are showing they want to accelerate the process, but large-scale adoption will take time.
With the arrival of protocols like UCP, AI agents will no longer just suggest products but will be able to manage the entire purchase journey. Do you think this will change how brands need to design their e-commerce sites?
Your site will become the source of truth for your data, but interaction with the site will increasingly be mediated by agents. Brand loyalty will matter even more than in the past, and UCP can play a significant role across the entire customer lifecycle.
The site’s capabilities will need to be made accessible to agents: clean, consistent feeds, Schema.org Actions, and endpoints that let a machine check a product’s availability and complete a purchase.
Humans build a more emotional bond with brands and make decisions shaped by trust, by stories and by the points of view a company communicates. Agents, however, will also need to be able to factor in these elements, because they will keep influencing loyalty towards a brand.
For years SEO has focused on earning clicks. If in the future it’s increasingly AI agents choosing which products to show or buy, what will the new decisive factors be for getting selected?
Verifiability. An agent making a purchase on someone’s behalf takes on a kind of responsibility that a plain search result never had.
Ambiguity is the real weak spot. A human can tolerate an unclear product page and browse through the site’s different sections to find the information they need. An agent simply moves on to the next candidate.
The job of SEO professionals is to maximise data integrity and accuracy while minimising every form of ambiguity. Essentially, that sentence now describes the entire job.
As Agentic Commerce evolves, do you think schema.org, Merchant Center and product data quality will become an even bigger competitive advantage than in the past?
Yes. Schema.org is enjoying a second life. Both UCP and ACP rely on Schema.org types to give agents a reliable information base.
The same goes for Merchant Center, where feed quality used to be seen mostly as a Shopping-campaign concern, while today it represents a layer of connection with agentic surfaces.
The competitive advantage will go to the companies whose data is consistent and easy to verify across pages, feeds and markup.
Many e-commerce businesses are still investing mainly in content and traditional optimisation. Do you think they should also start preparing now for a web where the main “customers” will increasingly be AI agents?
Yes, but without abandoning what they’re already doing. It’s not about adopting a new SEO method, it’s about improving what we’ve come to take for granted over time.
Content still matters, but editorial standards need to be higher. Content must have structure and format, bring genuinely new and distinctive information, and be clear and solid enough to be quoted verbatim by an LLM.
Agents have a much lower tolerance for ambiguity. That’s why you need to improve your grounding system for agents and minimise every possible source of hallucination.
How much does a page’s structure matter today in AI systems’ ability to interpret its content correctly?
It hasn’t become more important than it was before these systems arrived, but it remains very important.
Systems like NLWeb directly ingest headings, lists and tables, along with Schema.org markup, sitemaps and feeds. A table with a product’s specifications, if structured as a real HTML table, is interpreted more accurately than the same information presented without proper semantic structure.
The same information dropped into a bunch of styled divs with no clear structure, or presented as an image, can be practically invisible to an agent.
In recent months there’s been more and more talk about the llms.txt file. What’s your take on this standard and what role do you think it can play in the future?
I think there’s been some confusion around llms.txt. Some have described it as indispensable for gaining visibility in LLM systems. It’s not the miracle solution it’s often made out to be, but the proposal is that the file exists and agents can consult it, should they choose to.
I see it as a form of insurance. In case a system does actually check for the file, I want to make sure it’s available. The implementation cost is practically zero and the underlying concept is sound, even if its impact is nowhere near that of good SEO.
Model Context Protocol servers are starting to enter the daily work of many professionals. How do you think they will change the way we analyse data and make decisions in SEO?
I think MCPs are great, but at least for now I can’t rely on their accuracy.
The MCP concept will keep evolving over time, but today the main obstacle is the complexity and sheer volume of the data we analyse. LLMs, moreover, keep delivering answers with great confidence even when those answers aren’t always accurate.
So let the MCP explore the data with you, not instead of you.
Conversational interfaces will play an increasingly important role in the future, don’t you think? How do you imagine the relationship between websites, search engines and AI systems will evolve over the next few years?
The web is splitting in two: one built for humans and another for agents. Search engines will be used differently by the two audiences and, in the long run, I believe websites themselves will change to serve both effectively, just as responsive design emerged as mobile usage spread.
The website will need to become the official source of a company’s data. The same information can then be used in different ways: shown to people through the site’s pages, or read and interpreted by AI agents through structured data and machine-readable content.
This might sound like a downsizing of the website’s role as we know it today, but I believe the opposite is true. When multiple systems compete to acquire and use our data, the responsibility of keeping it accurate, consistent and reliable falls increasingly on us.
Your next sale will be made by an AI agent: your e-commerce has to be ready!
This conversation with Alex Moss was exactly what we needed, don’t you think?
I think we needed to clear some of the fog around agentic commerce, a concept that will be hugely important in the coming months.
The message is clear and worth repeating: your site has to be ready.
Ready in the technical sense: clean structured data, consistent feeds, information that doesn’t contradict itself across product page, Merchant Center and markup.
Because yes, an AI agent doesn’t forgive ambiguity the way a human would. If it finds an inconsistency, it simply moves on to the next competitor.
And it doesn’t end there, because what third-party sources say about you matters too. An agent doesn’t just read your site, it also weighs how the rest of the web talks about you. So reputation and information consistency everywhere become an integral part of SEO (and GEO, of course!).
The interesting thing is that agents don’t have an emotional side like we do.
They don’t grow attached to a brand because of a well-told story or a bright, empathetic tone of voice. But in their own way they will still take into account the trust and reliability a brand demonstrates over time.
So the advice I’ll leave you with is simple: get ready now, because an ever-growing share of traffic and sales will come precisely through AI agents.
Thanks so much for reading this far. See you next Monday with another great figure of international standing, right here as always, on SEO Confidential.
