\n<\/p>\n
\n\t\tOpinions expressed by Entrepreneur contributors are their own.\t<\/p>\n
When most people hear \u201cAI marketing,\u201d they think of using AI to do marketing. This may mean writing copy faster, generating ad creative, automating outbound sequences, etc.\u00a0That\u2019s useful and has a huge impact on how marketers approach their jobs, but it\u2019s not what I\u2019m talking about.\u00a0<\/p>\n
I\u2019m talking about marketing to AI. Specifically, to the agents that are starting to shop on behalf of people.\u00a0<\/p>\n
The first wave of AI search optimization (AEO, GEO, AIO, pick your acronym) was built around one shape of AI: the chatbot or answer engine. A shopper asks a question, an LLM writes an answer, and somewhere in that answer your brand shows up, or it doesn\u2019t.\u00a0<\/p>\n
That framing got a lot right, but it is still largely rooted in how we think about traditional search, the only real difference being the shift from keywords to questions. That said, it has sparked a new discipline within marketing, one where many brands are tracking and investing energy in understanding how AI models read their site and content, which sources AI cites when answering brand-relevant questions, and how they stack up against competitors across LLM responses.\u00a0<\/p>\n
All of this is important, but just the tip of the iceberg. What marketers need to do next is evolve their GEO strategies to not only ensure their brand appears in the first answer, but also to ensure that as a shopper travels through the full buyer\u2019s journey within the same AI search conversation, their products and services remain relevant and ultimately win the customer decision. The later portion of the journey used to happen after a shopper clicked from their initial Google search to your website. Now it lives entirely between the shopper and their agent.\u00a0<\/p>\n
This is where we\u2019ve seen brands quietly lose. When my team ran 320 simulated shopper conversations<\/a>, nearly three in ten brand mentions never turned into a recommendation. The brand showed up in the answer, and then somewhere further down the conversation the agent picked someone else. That lines up with independent research<\/a> showing AI shopping assistants get worse as conversations get longer. Getting mentioned and getting picked are two different things, and right now most brands are only measuring the first one.<\/p>\n It is important to note that not all agents look like chatbots, and as AI Search evolves into new formats, so must our marketing strategies.\u00a0<\/p>\n Look at Meta\u2019s Muse. It launched on Sept. 8 and quickly climbed to the top of the App Store. It\u2019s not really a chatbot; it\u2019s a personal agent that runs on its own cloud machine, opens its own browser, scans retail sites, compares prices, and checks out for you. It keeps a shopping profile on each person, tracking their actions and tastes.\u00a0<\/p>\n And Muse is one example, but there are agents that live in text messages, inside browsers, in glasses, and in voice. Many brands and retailers are building their own agentic shopping experience too. Each iteration and version finds products differently, reads them differently, and ultimately guides their users down buyer journeys in different ways. These are all new potential blind spots that we, as marketers, must learn to address and influence to reach our shoppers via their agent.\u00a0<\/p>\n A few things already stand out:\u00a0<\/p>\n They pull from everywhere. <\/strong>In one early test,<\/a> most of Muse\u2019s product picks came from live websites rather than Meta\u2019s own catalog, even though the catalog offered far more options. Your site, your feeds and every catalog you\u2019re listed in all count.\u00a0<\/p>\n They act, not just answer. <\/strong>An agent that can check out has to get through your site first. Amazon has already blocked Muse. Plenty of brands are blocking agents without knowing it, through bot protection built for a world where bots were the threat.\u00a0<\/p>\n They remember. <\/strong>A profile that says \u201cWide feet, and need extra support for bad knees\u201d shapes every future recommendation. That\u2019s potential brand loyalty, written down by a machine.\u00a0<\/p>\n They report to a human. <\/strong>Muse asks for approval before it makes a purchase. The agent has to justify its pick and give its reasons.\u00a0<\/p>\n That last point matters more than most brands realize, and it\u2019s a big part of where that three-in-ten number comes from. Agents don\u2019t just find you, they do a lot of the \u201cchoosing\u201d on behalf of their shopper. In our simulations, in which shoppers were looking for headphones under $150, the shift from mentioning a brand to recommending it occurred once the shopper provided a budget. One premium brand was mentioned in 82 answers and recommended in none. Another got named as a premium option and then dropped as soon as the price cap came up, losing the chance to sell premium value to justify a higher cost.\u00a0 <\/p>\n Meanwhile, a value brand went from never being recommended when budget came up to almost always being recommended afterward, which is a price-fit story, not a trust story. The brands that dropped off weren\u2019t unknown to the agent, they just didn\u2019t fit the shopper\u2019s constraints, which meant the agent had nothing to justify the pick with when it came time to choose.\u00a0<\/p>\n Here\u2019s the issue with AEO and GEO as labels: they name the interface.\u00a0<\/p>\nChapter two will encompass more than one shape of agent.\u00a0<\/h2>\n
Chapter 3 is about naming the buyer, not the interface\u00a0<\/h2>\n