Agentic AI is moving quickly from theory to reality. McKinsey predicts that agentic commerce could account for up to $5 trillion in global e‑commerce purchases by 2031, while Morgan Stanley suggests that 20% of US e‑commerce revenue could be realised through agent‑led transactions over the same period. That’s the equivalent of $385bn in purchases completed not by people, but by AI agents acting on their behalf.
This shift is already underway. In January, Google announced the launch of its Universal Commerce Protocol (UCP), designed to allow AI agents to discover products and complete checkouts directly on websites. In recent months, Google has focused on making this experience as seamless as possible for users, while working with payment partners such as Mastercard to improve security and functionality.
Importantly, this isn’t just about commerce. While shopping is one of the fastest‑growing use cases – and likely to become the most commercially significant – agentic experiences are expanding more broadly. In April, Google announced that agent‑led restaurant bookings are now available in AI Mode in the UK. Search, browsing and decision‑making are all beginning to change shape.
What this means for brands
Whether we like it or not, agentic browsing and commerce will become part of the search ecosystem. The scale and speed of adoption remain uncertain and come with clear caveats, but the technology is live and already being used. That makes it something brands need to understand and plan for now, not later.
Which brings us to a key question: what role does a website play in an increasingly agentic world?
From a practical perspective, there are two distinct challenges brands need to address:
- Making websites agent‑ready (relevant to all brands)
- Optimising for agentic commerce (relevant to e‑commerce brands)
Making websites agent‑ready
For years, website optimisation has focused on delivering the best possible experience for people. The next step is understanding how AI agents experience the web. The encouraging news is that what’s good for users is often good for agents too. When an agent interacts with a website, it evaluates several key elements:
- Visual layout: Agents take screenshots of pages to understand layout and infer how users are expected to interact with them.
- HTML structure: Beyond visuals, agents assess how pages are structured and labelled, using HTML hierarchy to understand meaning and importance.
- Accessibility: Agents also reference the accessibility tree to interpret how elements such as sliders, toggles, and input fields are intended to function.
In many cases, brands that have invested in usability, clarity and accessibility are already some of the way there.
Optimising for agentic commerce
While concepts like the UCP can sound daunting, implementation can be relatively straightforward. For example, Shopify stores can enable agentic commerce through a simple setting in the admin panel. Other platforms may require more development effort, but the barrier to entry is lowering.
However, technical readiness is only the starting point. The bigger challenge is ensuring that agents choose a brand’s site when completing a task. Factors that influence this include:
- Clear, well‑structured product titles
- Detailed and accurate product descriptions
- Reliable pricing and availability information
- Effective use of structured data
- Scenario‑based content and FAQs that answer real questions
In an agentic environment, clarity, accuracy and completeness become competitive advantages.
The reality check
There’s no doubt that agentic browsing and commerce could represent a significant shift in how some people search, choose and buy online. But adoption won’t be universal, at least not in the short term.
The scepticism is understandable. While there are many positive use cases, there are also well‑documented instances of agents ignoring guardrails or misinterpreting instructions, sometimes with damaging consequences. Trust remains one of the biggest barriers to wider adoption, particularly for higher‑risk or more considered purchases where reassurance and control matter most.
Research into how trust in AI agents is built is ongoing, and it’s an issue technology companies will need to address if agentic experiences are to scale meaningfully.
So, are websites still the destination?
Agentic commerce and browsing are still in a growth phase. Companies like Google and OpenAI, alongside partners such as Mastercard, continue to invest heavily in advancing these capabilities.
If forecasts of up to $5 trillion in global agent‑led purchases by 2030 are even partly realised, brands can’t afford to ignore the need to be agent‑ready. At the same time, user trust remains a significant hurdle, and even in an agent‑first future, there will be a substantial audience that prefers to browse, compare and complete tasks directly on websites.
Which brings us back to the original question: what role does a website play in an increasingly agentic world?
The answer is that websites are very much here to stay – but their role is evolving. They need to serve both people and agents, supporting discovery, decision‑making and transactions in parallel. The opportunity isn’t to choose one over the other, but to design for both.
That’s where the right guidance and tools matter. Helping brands become agent‑friendly – across e‑commerce and non‑commerce experiences – is now part of building future‑ready digital strategies. And for those who get it right early, agentic AI doesn’t just change the destination. It opens a new route to growth.
If you'd like to hear more, don't hesitate to get in touch. We'd love to hear from you and get a conversation started.
