AI Agents Are the New Shelf Space

Brands have spent years optimizing for two shelf spaces: the physical aisle and the search results page. A third one is forming quickly—AI assistants that research, filter and sometimes buy on a customer’s behalf.

New data from several trackers shows how uneven that shelf already is. The assistant a shopper happens to open can deliver a completely different shortlist, and marketers are only beginning to map the rules.

One prompt, very different aisles

An analysis by Jellyfish’s Share of Model tool tracked product recommendations from Alexa, ChatGPT and Google’s AI Mode across eight categories in the US, UK, Australia and Singapore. When asked for toy recommendations, Alexa for Shopping showed toys from 177 brands—and nearly every one came from Amazon’s own store. The same request to ChatGPT produced recommendations involving 24 retailers. Google’s AI Mode cited 37.

In practice, that means AI visibility is partly a function of being stocked in the stores each assistant is willing to look at. The sale, in turn, depends on which assistant the shopper opened first.

Retailers are not equally visible

Microsoft’s Copilot cites retailers far more than other assistants, according to Tinuiti. In July, the top 100 e-commerce sites accounted for 34% of Copilot’s citations—more than triple their share on Google’s AI Mode and more than eight times their share on ChatGPT. Tinuiti partly attributes this to Reddit blocking Copilot’s crawlers, which leaves more room for stores.

Across the six assistants Tinuiti tracked, retailers’ share of citations slipped to 14% in July. The direction is clear: getting recommended is now a distribution problem, not just a content problem.

Marketers are already chasing the bots

A Dentsu Creative survey of 1,950 senior marketers across 14 markets found 59% are optimizing content and SEO for AI-enabled search. Half are investing in agentic commerce to support the customer experience. AI search optimization ties for the top innovation strategy CMOs say they already use, level with social commerce and AI-driven personalization at 59% apiece.

The practical playbook looks like this:

• Map each assistant’s retailer graph. If your product isn’t stocked where an assistant looks, no prompt will surface it.
• Treat structured product data as media. Availability, price, specifications and reviews feed the agent’s decision rules.
• Measure share of model, not just share of search. Track which assistants cite your brand, your category and your retail partners.
• Weight high-intent AI traffic separately. Adobe found AI-referred shoppers generated 53% more revenue per visit than other traffic in July.

That last stat matters more when you put it next to history. A year earlier, non-AI visits were worth 128% more. Adobe credits assistants getting better at surfacing relevant products to higher-intent shoppers, and the people using chatbots may already be further along in their decision.

The next mass-market agent is already at checkout

Meta’s Muse could force the pace. HUMAN Security began tracking Muse on Sept. 22 and logged 5.6 million requests in under two days—a target ChatGPT took 11 months to reach. Muse now accounts for an average of 72% of the daily AI-agent requests HUMAN sees, and 84% of those requests target product and search pages. In a recent sample, Muse was 25% more likely than other agents to reach the checkout.

Muse has a generous free tier and can lean on Facebook, Instagram, Messenger and WhatsApp. That gives it a distribution advantage few standalone assistants can match.

What to do before agents finish their training run

The brands that win won’t be the ones with the most ad impressions; they’ll be the ones with the cleanest data and the right retail partnerships. If AI agents are becoming the shelf, then shelf placement is no longer about aisle position—it’s about being cited when the agent answers.

Source: Digiday