Your next website visitor might be an AI agent. Is your brand ready?
The customer journey has acquired a new middleman
For years, the digital journey followed a familiar pattern. Someone searched, clicked, landed on a website and decided what to do next.
That sequence is starting to change.
Salesforce’s latest State of Commerce research reports that the use of agentic search as the first step in a shopping journey grew 200% year on year. Instead of opening ten tabs and working through them, customers can ask an AI assistant to find the best option that fits a set of needs, compare the evidence and produce a shortlist.
The agent may eventually make the purchase too.
This remains an emerging behaviour rather than a universal one. Trust is uneven, the technology still makes mistakes and plenty of people will want to remain firmly in control of the final decision. But brands should not wait for fully autonomous shopping before paying attention.
Research and comparison are already valuable parts of the journey. If an agent handles those stages, it can influence which brands a person ever gets to see.
The fight is about more than convenience
Amazon recently blocked Meta’s Muse agent from browsing and buying on its platform. Amazon raised questions about authorisation, account access, security and customer experience.
Those concerns are legitimate. They also reveal the commercial stakes.
The platform that begins the journey gets to shape the recommendations, observe what the customer wants and decide which options appear. The business completing the transaction risks becoming little more than a supplier behind somebody else’s interface.
For brands and retailers, this creates a difficult choice. Third-party agents could bring valuable customers, but they may also weaken direct relationships and reduce control over how products are presented.
Mastercard’s recently announced agentic-commerce tools frame the same challenge from the merchant’s side: how can businesses make products available to AI assistants while retaining control over pricing, fulfilment, business rules and the brand experience?
There will not be one answer for every organisation. But ignoring the question leaves the answer in somebody else’s hands.
Being found is no longer enough
Traditional search optimisation helps a page appear. An AI agent needs to understand what is there and whether it satisfies a particular request.
That requires more than keywords.
Product details need consistent names and attributes. Prices, stock levels, delivery information and returns policies need to be current. Important content cannot be hidden inside images, vague headlines or awkward interactions. Claims need evidence an agent can verify.
A controlled study published earlier this year compared two versions of the same ecommerce website. The products, prices and journeys were identical, but one version included clearer structure, labels, action cues and machine-readable data. Browser agents successfully completed 89.3% of tasks on the agent-ready version, compared with 49.3% on the conventional version.
It is preliminary research rather than a universal benchmark, but the principle is hard to ignore: clearer digital foundations help agents make fewer mistakes.
They usually help people too.
Generic brand language has nowhere to hide
This shift could be uncomfortable for brands built on broad, unsupported claims.
“Premium quality” means very little to an agent. Premium in what way? Compared with what? Is there independent evidence? Does the warranty support the claim? Do customers consistently mention the same strength in their reviews?
The agent becomes a ruthless comparison layer. It turns fuzzy positioning into a practical question: why should this product be recommended for this particular need?
That does not reduce the importance of brand strategy. It exposes where the strategy has never been made concrete.
A clear proposition should influence product information, proof points, service policies, customer reviews and the language used throughout the experience. If those signals tell conflicting stories, an AI assistant may misrepresent the brand or exclude it from consideration entirely.
Machine-readable must not mean forgettable
There is an obvious danger here. Brands could respond by filling their websites with uniform product feeds, comparison tables and clinically efficient interfaces.
That might make the agent happy. It will not necessarily make anyone want the product.
Human beings still notice design, tone, story, status, identity and belonging. We still use brands to reduce risk and make decisions feel right. An AI recommendation might bring someone to the door, but the experience that follows has to earn trust.
The job is to serve both audiences.
For machines, that means structure, accuracy, consistency and evidence.
For people, it means clarity with character: a distinctive point of view, thoughtful design, recognisable language and an experience that feels coherent from recommendation to delivery.
Good digital design has always balanced usefulness with feeling. Agentic browsing simply makes that balance more visible.
A sensible place to start
Brands do not need to rebuild everything around a technology that is still developing. They do need to examine the foundations.
Start by asking an AI assistant to compare your offer with several competitors. Look at what it gets right, what it misses and which sources it relies upon.
Then inspect the information behind the experience:
Are your products and services described consistently across your website, retailers and third-party profiles?
Can an agent understand important differences without guessing?
Are your strongest claims backed by accessible evidence?
Are price, availability, fulfilment and returns details clear and current?
Does the experience still feel recognisably yours when the information is stripped back to its essentials?
This is partly a technical audit, but it is also a brand and content exercise. Structured data cannot rescue a muddled proposition. Equally, a brilliant identity cannot help an agent interpret information that is incomplete or contradictory.
Design for the handover
The most useful way to think about agentic commerce is not as machines replacing customers. It is as a new handover within the customer journey.
An agent may identify the need, conduct the research and narrow the options. The person still needs confidence in the recommendation and reassurance that the brand will deliver.
Brands that prepare well will make that handover feel seamless. Their information will survive the journey through an AI interface, and their personality will still be present when the customer arrives.
The front door may be moving. The need to give people a good reason to come in has not.
