Something significant is shifting in how people shop online. Increasingly, the person choosing the product is not really a person at all - it is an AI agent acting on their behalf. Users are asking ChatGPT which air fryer to buy, letting Perplexity compare broadband deals, and getting product recommendations surfaced directly inside Google AI Overviews without ever visiting a retailer's website. If your brand is not part of what the agent recommends, you are not in the consideration set. Full stop.
This is not a distant hypothetical. AI-powered shopping agents are already influencing purchase decisions for a meaningful share of online consumers, and that share is growing. The challenge for marketers is that the mechanisms driving these recommendations are poorly understood, and most loyalty strategies were built for a world where a human being was doing the browsing.
Why Traditional Loyalty Thinking Does Not Transfer
Brand loyalty, in the conventional sense, rests on accumulated experience. A customer tries your product, has a good outcome, remembers your brand name, and comes back. That cycle depends on direct interaction between a person and a brand. AI agents disrupt every link in that chain.
When an agent handles the search, evaluation, and shortlisting, the customer may never form a direct association with your brand at all. They asked for the best value running shoe under £80. The agent picked one. The customer bought it. The agent made the choice, not the customer. If the experience is positive, loyalty may form - eventually - but the initial selection was entirely agent-driven. And next time, the agent may pick something different based on updated pricing, availability, or review signals.
The implication is straightforward but uncomfortable: brands cannot rely on the emotional equity they have built with human shoppers to carry over into agent-mediated environments. Agents do not have nostalgia. They process signals.
What Signals AI Shopping Agents Actually Use
Understanding which signals influence agent recommendations is the foundation of any sensible strategy here. Based on what we know about how AI models like ChatGPT and Perplexity generate product recommendations, and how Google AI Overviews surfaces shopping results, a few factors consistently matter: structured product data, review volume and sentiment, third-party editorial coverage, and the clarity and specificity of claims made on your own website and product pages.
Agents are, fundamentally, synthesising publicly available information. They are reading your product descriptions, pulling review data, and weighing up what credible sources say about you. A brand with sparse product copy, thin review profiles, and no meaningful third-party mentions is an invisible brand to an agent - regardless of how much brand equity it has built through traditional advertising.
This is where the discipline of Generative Engine Optimisation (GEO) becomes commercially relevant, not just for organic search traffic, but for agent-mediated commerce. Making your product attributes explicit, ensuring your claims are specific and verifiable, and building a presence on the sources agents tend to draw from - review platforms, specialist publications, comparison sites - all become part of the loyalty equation.
The Google AI Overviews Clarification Marketers Should Sit With
Alongside the conversation about shopping agents, Search Engine Watch also reported something worth examining carefully: Google has stated that brands do not need to specifically pursue AEO or GEO tactics to appear in AI Overviews. The implication from Google is that good content naturally gets surfaced.
This is technically defensible but practically misleading for most brands. Google is correct that there is no separate submission process or special flag to set. But the characteristics of content that performs well in AI Overviews - clear structure, authoritative claims, specific answers to specific questions, good E-E-A-T signals - are exactly what GEO practitioners focus on. Calling it something other than GEO does not change what the work involves.
For UK brands, the practical takeaway is this: do not wait for Google to validate a particular methodology before investing in the underlying content quality improvements. If your product pages are vague, your FAQ content is thin, and you have no third-party editorial coverage, no amount of semantic reassurance from Google will fix your AI Overviews presence.
Building Loyalty When You Cannot See the Agent
The measurement problem here is genuine. If an AI agent recommends your product and a customer buys it, that conversion may appear in your analytics as direct traffic, or as a generic organic visit, or possibly not at all if the purchase happens within an agentic interface. Attribution is difficult, and loyalty metrics built around repeat visit behaviour become unreliable.
What this pushes brands towards is focusing on the post-purchase experience with renewed intensity. If you cannot reliably control whether an agent recommends you in the first place, you can control what happens after the sale. A strong post-purchase experience generates the reviews and word-of-mouth content that feeds back into agent recommendation signals. It is a longer loop, but it is one brands can influence.
There is also a case for thinking about your brand's presence in the specific communities and publications that AI models treat as authoritative sources. Consumer review platforms, category-specific editorial sites, Reddit threads relevant to your product area - these are not new channels, but their importance is amplified when agents are pulling from them to make decisions. A proactive review generation programme and a consistent presence in relevant online communities becomes infrastructure, not optional.
What Paid Search Teams Should Be Watching
For brands running Performance Max or AI Max campaigns, the rise of shopping agents adds a layer of complexity to how you think about audience signals and creative. Performance Max already operates across Google's full inventory, including Shopping surfaces where agent-driven recommendations increasingly appear. The campaign's reliance on first-party data and audience signals means brands with richer customer data will be better positioned to stay competitive as agent-driven commerce scales.
Product feed quality also becomes more important, not less, as agents rely on structured data to make comparisons. If your titles, descriptions, and attributes are inconsistent or incomplete, you are handicapping both your paid Shopping performance and your organic agent visibility at the same time. These are the same data problems, with compounding consequences.
The Practical Starting Point
Brands that will adapt well to agent-mediated commerce share a common trait: they treat content quality, data quality, and customer experience as connected systems rather than separate workstreams. That is easier said than done in organisations where SEO, paid media, and CX sit in different teams with different budgets.
But the brands that get this right - that produce specific, credible, well-structured product content, maintain clean and complete data feeds, actively cultivate review volume, and deliver a post-purchase experience worth talking about - are the ones that will appear consistently in agent recommendations over time. That is the new version of loyalty. Not an emotional attachment, but a signal profile strong enough that agents keep picking you.