What influencer marketing can and cannot teach brands about personal agents

It is easy to describe personal agents as the new influencers. Both stand between a brand and a shopper as a voice the shopper trusts, and both now sit close to the moment of purchase. I think that comparison hides the difference that matters most for brands: an influencer is valuable because of their point of view, while an agent is valuable because it is impartial and focused entirely on the shopper's problem.

Many merchants I talk with expect personal agents to play a big role. What they do not know yet is whether shoppers will trust an agent to buy for them, and shopping through an agent can feel like shopping through a pinhole. As a result, many are waiting to see. For now, agents feel like a more natural fit for simpler purchases, like reservations, food orders and everyday essentials. Influencer marketing offers some useful answers, as long as we are clear about where the comparison stops working.

Influencers are tastemakers. Their perspectives vary widely, and those points of view shape what they choose, what they recommend, and the community that gathers around them. Followers come back because that particular person's opinion, style, and community matter to them.

Agents are not human, so they have no taste or lived experience to draw on. Instead, they can catalog, compare, and analyze far more options than any person could, which lets them tailor a recommendation to the person they are helping rather than to a personal preference.

A request for a gift for a teenager leads to very different products depending on whether he is a gamer or plays sports. An agent can ask which, find out what he already owns, and narrow the options to a few, even when it starts from an influencer's pick. Its range is limited only by the product information it can find and understand. For a brand, that means the agent's recommendation is earned through information rather than through a relationship, and that changes most of what brands learned from working with influencers.

The two do have real things in common. Both save the shopper from evaluating every option, because someone they trust has already done part of that work. Both turn a vague request, like "a gift for my mother-in-law who has everything," into a short list of specific items. And both add to the purchase: a creator mentions the cord-management kit behind their TV, and an agent suggests the same kit along with the wall mount that fits that specific model.

We saw the same wait-and-see attitude in the mid-2010s, when social shopping was just emerging. Many of the most sophisticated brands in the world held back on influencers because there was no reliable way to measure what an influencer post was worth, and the whole category was worth only about $1.4 billion worldwide in 2014. Younger brands, possibly because they had less to protect and more to gain, treated influencers as an experiment and learned from the results.

Revolve started sending free clothes to fashion bloggers in 2009, before Instagram existed, and by 2018 influencers drove nearly 70% of its sales. Glossier grew out of a beauty blog its founder started in 2010 and was valued at $1.2 billion by 2019. Meanwhile, Revlon, a 90-year-old beauty brand, filed for bankruptcy in 2022 after missing the cosmetics boom that influencers drove. In 2025, more than a decade after the early movers, Unilever announced it would move half of its media spend to social and work with 20 times more influencers.

Brands had roughly ten years to make up their minds about influencers. They will not have that long with agents: agentic purchases through Stripe's Link increased 38x in a single month, admittedly from a small base.

An agent is committed to its user. A brand cannot pay to change where its product appears in the agent's recommendations, and the AI platforms that sell ads keep them apart from the answer: OpenAI says ads do not influence ChatGPT's answers. Shoppers may trust an agent for exactly that reason, because they see it as having no stake in what they buy. That could change. My worry, and we already see signs of it with Meta, is that agent providers will start with walled gardens of approved merchants, much as Google leans on Merchant Center and Amazon favors its own brands.

The agent's work also does not stop at the recommendation. After an influencer says "buy this," the shopper usually buys from the one store the influencer linked to, or forgets. An agent has memory, tracks the price, and compares the same product across retailers on price, delivery date, and return policy before it buys. For retailers, this means the agent is judging the store's brand promise as well as the product.

Influencer marketing gave brands a feedback loop. Influencers hear from their audience through likes, comments, and affiliate sales, and brands can see much of that reaction too. Many brands built their influencer programs around it, adjusting products, briefs, and partners based on what the audience said.

Agents take that loop away. An agent hears from one shopper in real time and changes its recommendation within the same conversation, but that conversation happens somewhere the brand cannot see, and OpenAI keeps conversations private from advertisers. A brand may never learn that an agent passed over its product because a size chart was missing or a material was not listed. The only part of that conversation a brand can shape ahead of time is the information the agent finds about its products.

Brands learned to give creators what they needed: a product, a brief, and a link to buy. Agents need the same three things in a different form.

Creators still matter in this picture, because agents read what they publish. In AI answers to US shopping questions, Reddit and YouTube are cited more than three times as often as every major retailer and marketplace combined. The agent then checks the details against product data. A creator can explain why a rain jacket is worth buying, and the catalog has to confirm the fabric, the fit, and whether the shopper's size is in stock.

We do not yet have hard data on what agents weigh most, but a pattern is emerging. Complete attributes get a product considered. Context helps it get recommended: answers to shoppers' questions, FAQs, the occasions it suits and what it's best for. Price and availability come once the agent has narrowed the options.

Some early Instinct users complained about product suggestions they had not asked for. An agent is useful only while the shopper believes it is working for them, and recommendations that feel like a pitch put that belief at risk, the same way undisclosed sponsored posts did for influencers. So even as agents add some taste, we expect the problem-solving role to stay at the center of why shoppers use them.

With influencers, a brand chose who spoke for it. With agents, the agent chooses the brand, based on what it can find and understand in the brand's catalog, and the brand cannot pay for that recommendation or watch the conversation that leads to it. That makes product information the main way a brand takes part.

Fabric enriches product data for AI shopping, shows how your products appear across AI channels, and will soon support the emerging Personal Agent Protocol. Start with a catalog assessment to see where your data is complete and where an agent would have to guess. Influencer marketing took years to develop measurement that brands trusted, and AI shopping is at the beginning of that process; our next post looks at how to measure performance on AI channels.