We’ve spent years understanding how people discover products, evaluate brands and move towards a purchase. But AI agents are starting to take on more of that research and comparison, and Meta’s new personal AI agent, Muse, is one of the latest examples of what this could look like in everyday life.
Muse can research products, compare options and prepare purchases on consumers’ behalf. It reportedly reached five million downloads in just 22 days, but what interests me more is what happens when agents start influencing which brands consumers consider.
It’s early days, and there’s little reason for marketers to rethink their budgets just yet. But Muse raises some interesting questions about how product discovery could evolve and what that might mean for paid search, paid social, retail media and SEO.
How Muse decides which products to recommend
Muse connects to a user’s tools and services, allowing it to carry out tasks on their behalf. When shopping, it can find products through Meta’s catalog, retailer integrations and the open web, before presenting recommendations and helping users complete a purchase.
What’s interesting is how those products make it into the recommendations.
In a small study of 50 shopping prompts, researchers found that Meta’s catalog supplied approximately 95% of the initial product candidates, while Muse’s browser typically visited just three websites. However, the final recommendations leaned towards products found through the open web, suggesting that catalog inclusion alone doesn’t guarantee visibility.
The same reporting points to instructions favouring well-known merchants, although Meta hasn’t publicly explained exactly how it determines which brands qualify or how products are ultimately ranked.
It’s a small sample, but the implications are worth considering. Brands with strong recognition, established retail relationships and accessible product information may have an advantage. Smaller brands could face an additional challenge if they’re competing for visibility within a relatively narrow set of recommendations.
There’s also the question of monetisation. Meta has said Muse doesn’t currently carry advertising, but operating AI agents is expensive, and it’s reasonable to expect the commercial model to evolve. Sponsored alternatives alongside organic recommendations are one possibility, although there’s no confirmation that Meta intends to introduce them.
For now, the more immediate question is how these agents could influence the channels marketers already invest in.
What could this mean for search, social and SEO?
I expect Muse’s impact on advertising to be fairly limited for the foreseeable future. We don’t yet have enough evidence of consumers changing their shopping behaviour at scale to draw firm conclusions about channel performance. There are, however, a few hypotheses worth exploring.
Paid search: discovery could decline, but verification may remain
One possibility is that some non-brand discovery and comparison activity moves away from traditional search engines and into conversations with AI agents. Someone looking for a new skincare product, for example, might ask Muse to compare options based on their preferences rather than conducting several searches themselves.
That could reduce certain types of non-brand search volume over time. But it doesn’t necessarily mean search becomes less important throughout the buying journey.
I can also see branded search becoming a way for consumers to verify an agent’s recommendations before committing to a purchase. This would fit with Google’s Messy Middle framework, where consumers move between exploration and evaluation rather than following a linear path to conversion. An agent might do more of the initial exploration, while Google remains a place people go to check reviews, compare prices or reassure themselves about a brand.
Paid social: what happens to intent signals?
For paid social, I’m particularly interested in what this could mean for the signals platforms use to understand purchase intent.
One of the richest intent signals available to social platforms comes from how people engage with content. What they watch, click, save and interact with helps platforms understand their interests and predict what they’re likely to buy.
If consumers increasingly delegate product research to agents, some of those interactions could change. A person who asks Muse to find a product may no longer need to engage with the same volume of shopping-related content to reach a decision.
That could eventually affect the signals platforms use to understand purchase intent. Equally, agents could introduce new signals through the preferences and shopping requests people share with them. How those signals would be used within Meta’s advertising ecosystem remains unclear, particularly given the privacy boundaries Meta has described between Muse and its ad systems.
At this stage, I’d treat this as something to monitor rather than a reason to change paid social strategy.
SEO: product data and brand authority become increasingly important
For SEO, the implications are perhaps more immediately practical.
Agents need reliable information to identify products, compare options and check whether something is available to buy. Accurate pricing, stock information, structured product data and accessible product pages could therefore become increasingly important.
Brand authority matters here too. If agents favour established merchants, brands with stronger recognition and a credible presence across the web may be better positioned to appear in recommendations.
A lot of this is already good SEO and ecommerce practice. The difference is that the information may increasingly need to serve an automated buyer as well as a human one.
Retail media: could the point of influence shift?
For retail media, the question is whether more product discovery happening through agents could change how shoppers interact with retailer platforms. If consumers arrive at Amazon or other retail sites having already narrowed down their options, there may be fewer opportunities for on-site sponsored placements to influence their decisions.
That could make visibility earlier in the purchase journey increasingly valuable, whether through AI recommendations, social or CTV. But the impact will likely vary by retailer. Amazon has already restricted Muse’s access, highlighting the importance retailers place on controlling their shopping experiences and customer relationships, while others are taking a more open approach.
Retail media still plays an important role in reaching consumers close to purchase, and there’s little evidence that this is changing today. For now, I’d be watching how retailers develop their own agentic capabilities and how those experiences might create new opportunities for brands to reach shoppers.
Why I won’t be changing media budgets yet
It’s worth separating the interest in Muse from evidence that it’s materially changing how people shop.
Downloads and agent registrations tell us there’s appetite to experiment with the technology. They don’t necessarily tell us how frequently people are completing purchases through agents, which categories they’re using them for or whether they’re replacing existing shopping behaviours.
There are also some fairly significant barriers to adoption.
Consumer trust is one. Using an agent to compare products is a relatively small commitment, but giving it access to personal information, connected accounts and payment methods is another matter. Meta’s history with privacy concerns may make some consumers particularly cautious about how much access they’re willing to grant.
Retailer participation is another consideration. Amazon has reportedly restricted Muse’s access, which illustrates the difficulties agents may face when trying to operate across commerce environments they don’t control. If major retailers limit access, that could affect how useful these tools are for certain purchases.
Until we understand how these factors play out, I don’t see a strong case for making significant adjustments to paid search or paid social investment. The existing channels are still doing their jobs, and we have limited evidence that Muse is changing their effectiveness.
What I do think makes sense is looking at how prepared brands are for a shopping journey that involves an AI agent.
What brands can start doing now
The good news is that much of the work involved in preparing for agentic shopping overlaps with improvements brands should already be making to their ecommerce experience.
I’d start with three areas.
Make sure agents can find your products. That means reviewing product catalog coverage and relevant retailer integrations, including how products are shared with Meta. Given the role Meta’s catalog played in the early testing, it’s worth checking that eligible products are correctly represented and that supported checkout integrations are configured.
Make product information easy to understand and verify. Agents need accurate titles, prices, availability, product identifiers and clear purchasing policies. Product pages should make it straightforward to establish what an item is, whether it’s in stock and how to buy it. It’s also worth testing the checkout process to understand whether automated purchasing introduces any friction. Improving these fundamentals benefits human customers too.
Continue building brand recognition and trust. This is the longer-term consideration, particularly if agents continue to favour established merchants. The work brands do across search, social, content and other channels to build familiarity and credibility could influence whether they’re considered by agents in the future.
None of these requires a major change in investment. They’re practical ways to improve product visibility and the customer experience while we learn more about how agentic shopping develops.
What I’m watching next
For me, the next thing to watch is whether early interest in Muse translates into meaningful changes in purchasing behaviour. How frequently will consumers use agents to shop? Will they trust the recommendations enough to complete purchases? And will any of that begin to show up in search volumes, social engagement or channel performance?
There’s also the question of how Meta eventually chooses to monetise Muse, and what that could mean for advertisers.
Until we have more evidence, I’d focus on getting the fundamentals right and continue investing in the channels delivering results today. Muse is worth paying attention to, but I don’t think it’s a reason to change course just yet.
