AI agents are reshaping how consumers discover and buy products — and brands are realising they may no longer own the consumer journey in a way they historically have.
According to Salesforce’s Connected Shoppers Report, 43 percent of retailers are actively piloting autonomous AI agents. As brands prepare for a world where consumers use AI agents to discover and shop for products, a race to understand what that means for the consumer relationship is taking place.
For decades, brands owned the consumer experience. The information they would get from ad clicks, browsing behaviour, and transaction data all historically informed retail marketing.
Now, because of conversational AI tools like ChatGPT, Claude, and Gemini, product discovery has been compressed into an instantaneous search function, where consumers can find out the best product for their specific needs instantly.
Instead of only searching for products in a conversational chatbot, consumers can give their personal agents a goal, such as ‘Find me a Korean facial cleanser for acne-prone oily skin that’s under £20’ and the agent can browse, compare cleansers, and even purchase one autonomously.
With this newfound use of personal shopping agents, the consumer journey, once owned by brands – including the data collected from browsing and transactions – is being taken by LLMs. The agent does the searching, the shopping and the purchasing, meaning brands have little understanding of the journey taken to transaction.
“The distinction that matters most is who owns the agent,” comments Aidan van Vuuren, Head of Digital at performance marketing agency Peak Digital. “A brand’s on-site AI agent gives them the conversation data, the recommendation logic, and the loyalty relationship. A third-party agent – ChatGPT, Perplexity, Google’s AI Overviews – gives them none of that. The brand gets the transaction but loses the relationship. The agent earns the loyalty, not the brand.”
The response from some has been blunt. Amazon sued AI search engine Perplexity to block its shopping agents from accessing the platform altogether. But for most retailers, shutting agents out isn’t viable — consumers are already using them, and that number is only growing. The question is no longer whether AI agents will reshape retail. It’s whether brands will be ready when they do.
A Brand’s Digital Shop Assistant
An AI agent that sits within an LLM is a general-purpose agent that can use external tools to complete tasks set out by a user. In comparison, a brand agent sits on a commerce website and acts as a personal shopping assistant to a consumer – typically tapping into the retailer’s inventory and aligning with brand guidelines.
In June 2025, American supermarket giant Walmart introduced its own AI shopping agent Sparky. The agent helps customers discover Walmart products, build custom grocery baskets and assess reviews without leaving the retailer’s website.
For a retailer like Walmart, a brand agent works well for consumers who can shop numerous brands under one trusted name. If comparing products across the entire web, however, shoppers might opt for an LLM agent over a brand agent that can access and compare products across multiple retailers.
Brands are now faced with the decision of whether they should grant access to third-party LLM agents but risk missing out on important customer data, or block these agents from entering their website but forfeit potential customers finding their products through this new purchasing channel.
The Gatekeeper to Your Customer
Already, personal agents from third-party tools aren’t only researching the web for products most aligned with user requests, but are talking to brand agents. This agent-to-agent communication is widely seen as the next step in how consumers shop.
For brands, agent-to-agent conversations will shed some light on user intent, as brands will find out what consumers want – not from themselves, but from their personal shopping agent.
“The big difference comes down to who the agent works for,” says Ed Freed, Global Chief Transformation Officer at marketing agency RAPP. “An agent on my phone, baked into Siri or Android, is loyal to me and those platforms, not to your brand. It builds the shortlist, and it has an unrivalled understanding of the person it works for.”
In this sense, who an agent is loyal to plays a major role in the types of brands it suggests. He adds: “It gatekeeps two or three options based on that person’s needs and intent, and the customer only ever sees the curated set. Brands building their own agents are kind of missing the point. Consumers will have their own agents, with unique utility to them, and increasingly those will be built into the devices we already own.”
“So the job for brands is to “AI encode” themselves, so a consumer’s agent can tell a Burberry shirt from a Marks and Spencer one and pick the right answer for that person. Any agent a brand does build won’t be talking to consumers. It’ll be talking to their agents.”
The Resurgence of First-Party Data
Despite third-party agents being able to surface products from multiple retailers, brand agents hold a significant advantage: depth of data.
“An on-site agent can access everything from past purchasing history and browsing behaviours to personal preferences, loyalty data and engagement metrics,” says Matt Sherwen, MD of digital consultancy Sherwen Studios. “This nuanced understanding of each customer makes it far easier to deliver personalised, relevant recommendations that align with their existing brand relationship. At the moment, third-party agents simply cannot replicate this.”
For brands, the priority is ensuring that data is both high-quality and accurate enough for these systems to learn from. While third-party agents will continue to shape how consumers discover products, it’s the brand-owned environment where conversion is most likely to happen. The longer-term challenge is making sure that both owned and third-party agents have access to sufficient data, so the experience remains strong and consistent regardless of where the customer journey begins.
As AI intermediaries become more common, first-party data may become one of the few remaining assets brands have full control over. “We’re already seeing brands invest more heavily in first-party data strategies, loyalty programmes and owned experiences,” notes Amanda Walls, Head of Organic Media at digital marketing agency Cedarwood Digital. “The brands seeing the greatest success are those using AI to remove friction from the buying process — not simply adding another chatbot to their website. The businesses that win in the next phase of digital commerce will be those that make it easy for both consumers and AI agents to understand, trust and recommend their products.”
Going Beyond GEO
Blocking third-party agents entirely isn’t a realistic option for most brands. Do so, and you forfeit visibility at the very moment consumers are increasingly delegating their product discovery to these tools. But allow them in unchecked, and you surrender purchase intent signals, clickstream data, and any meaningful insight into how your brand is being found.
The answer, for most, lies in becoming easier for agents to read. That means ensuring products are reviewed positively on authoritative platforms — major publications, Reddit, specialist forums — and that product feeds and content are written in the kind of conversational language agents are trained to understand. It also means getting the technical basics right, including consistent product details across every channel, and structured metadata that agents can interpret without friction.
Most brands, Freed says, are currently focused on Generative Engine Optimisation. (GEO) — the practice of optimising online content to surface better in AI tools – treating it largely as an extension of existing SEO practice. But optimising to be found is only half the challenge. Few are doing the harder work of encoding what their brand actually means into a form agents can act on.
“Some are starting the foundational work: making product, pricing, inventory and policy data machine-readable, implementing the semantic web, and beginning to think about ‘share of model’,” he says. “But in truth that’s just the table stakes for existing in an agent-driven world.”
The brands that pull ahead, he argues, will be those that go further — embedding not just product information but brand identity itself into the data layer. “For anything above the lowest-consideration products, purchase decisions are a mix of logic and emotion. The brands that win will be the ones that encode the essence of their brand as data for the agent and as experience for the customer, and know how to weave the two together at the right moments.”
In a world where an AI agent may never visit your website, third-party LLMs might be the only introduction a consumer gets to a brand.
