Meta's AI Shopping Agent Favors Known Brands

Your product page is live, your price is competitive, your reviews are strong. Then a shopper opens Meta's new Muse assistant, types "best linen overshirt under $150," and your brand never appears. Not because your product is worse. Because the agent was told to prefer someone else.

That is the practical reading of an instruction file that the AI-tracking firm Profound says it pulled from Meta's Muse shopping agent on September 25, 2026, as reported by PPC Land on October 11. For US D2C brands that are not household names, it is the clearest look yet at how a consumer shopping agent decides who gets shown.

What the Muse instruction file reportedly says

Muse launched in the US on September 8, 2026. According to the reporting, which relies on Profound's analysis and a State of Brand write-up dated October 8, the file instructs the agent to rank well-known merchants first. It calls it "imperative" that the top five products come from well-known merchants or websites and match the shopper's request. The rule appears in three places: the ranking step, the display instructions, and the formatting rules.

Other rules in the same file matter just as much for smaller brands:

• Listings priced outside a category's normal range are discarded, whether far above or far below. Challenger brands that price low to get noticed are exposed to this.
• For each request, the browser agent checks no more than three merchant sites and returns no more than ten products.
• In-chat checkout is built only for Meta catalog products carrying an "agentic checkout" flag. Open-web products fall back to browser checkout on the merchant's site.
• When a shopper names a brand, that brand's own store ranks first, and direct-to-consumer brands are favored over resellers by default.

Read the caveats before you act on it

This is one analysis, not a published Meta policy. Profound's sample was 50 prompts from a single retail category, run through one Muse account, and Profound itself calls it small and exploratory. Profound sells software that tracks how AI agents treat brands, a commercial interest the report discloses. The file may have changed since late September, Meta has not commented, and the document does not define "well-known" or explain how the three sites are chosen.

So treat it as a signal about direction, not a ranking manual. The direction is consistent with what we see across agent-driven shopping: agents compress the choice set, and a compressed choice set favors recognizable names.

Answer: how does an unknown D2C brand compete for AI shopping agent placement?

Meta's Muse shopping agent is reportedly instructed to fill its top five picks with well-known merchants, discard prices outside a category's normal range, and check at most three merchant sites per request. An unknown D2C brand's realistic levers are catalog presence with the agentic checkout flag, category-normal pricing, and building a name that shoppers and the web already recognize.

Where the numbers say the opening is

The Profound sample shows how narrow the funnel is. A typical catalog search returned about 60 products from 30 merchants. The browser then visited three sites and found four products at two merchants. The final six picks usually split four from the browser and two from the catalog. The catalog supplied about 95% of candidates at the median, yet web results carried more weight in final selection. In the words of Profound economist Jennifer Zou, "Much more weight is placed on candidates returned by the web search."

Two implications follow. First, catalog entry gets you into the candidate pool but does not win the slot. Second, whatever the browser agent finds when it reaches your site, and whatever the open web says about you, decides the outcome. That is an engineering problem as much as a marketing one.

Five things a lesser-known brand can control

These are our engineering judgments from building agent-ready commerce, not guarantees of placement.

1. Get into the catalog properly

If Meta's agentic checkout flag is what unlocks in-chat checkout, a catalog feed that is incomplete or unflagged forfeits that path entirely. Treat feed quality, product identifiers, availability and price accuracy as production systems, not marketing tasks.

2. Price inside the category's normal range

An agent that discards outliers cannot be argued with. If your positioning depends on being dramatically cheaper or dramatically premium, the product page has to carry that story for the shopper, because the agent may never show it. Check where your price sits against comparable listings before assuming an aggressive price helps.

3. Make your own site fast and legible to a browser agent

With only three sites visited per request, a slow page, a blocking cookie wall, or a variant selector that needs a human to click can cost you the visit. Clean markup, server-rendered product data, and stable URLs are table stakes.

4. Build the "known" signal outside your site

Because the file does not define "well-known," nobody can say what Muse counts. Independent reviews, editorial mentions and consistent brand data across the web are the sensible hedge. Be honest about the limit: we cannot tell you which signals Meta weights.

5. Protect branded search

The reported rule that a named brand's own store ranks first is the one reliably good-news line. Every shopper who already knows your name and asks for it can be routed to you. That makes brand-building spend and agent-readiness work compound rather than compete.

What this means for your roadmap

Agent visibility is now split across several surfaces, each with its own ranking logic: Meta's catalog and browser agent, Google's Universal Commerce Protocol, OpenAI's ChatGPT shopping, and others. Optimizing for one reported rule file is a poor strategy. Building a clean, structured, fast, truthful commerce layer that every agent can read is the durable one.

At MnT Future, that is the work we do: AI-native, agent-ready commerce platforms for US D2C and marketplace brands, with the product data, checkout and compliance layers built to hold up when an agent, not a human, is the first visitor. See our MnT Commerce build and AI agents services. We build for ACP, Google UCP and Retail MCP, and we keep the advice honest about what is measured and what is inference.

Where to start this week

Pull your top ten products and run them through the questions an agent would ask. Is the price inside the category range? Is the feed complete and current? Does the page render its core data without JavaScript gymnastics? Does a branded search land on you? The gaps you find are your agent-readiness backlog.

If you want a second pair of senior eyes on that list, MnT Future offers a free agent-readiness audit. We will show you where your store stands for ACP, Google UCP and Retail MCP and what we would fix first. Book a free strategy session.

Sources: PPC Land, "Meta's shopping agent must fill its top 5 picks with well-known merchants" (Oct 11, 2026), reporting Profound's study (Oct 2, 2026) and State of Brand analysis (Oct 8, 2026).