What happens to brands when agents do the choosing?
Brands shortcut decisions when our cognitive bandwidth is limited. What changes when an agent can compare unfamiliar alternatives against what we actually value?
Event / source date: .
At lunch on Monday, during Trusted Conference in Phuket, Freddy, Miro and I returned to a question that has been playing on my mind for some time: what does the agentic web mean for brands?
I have spent the past two years developing my hypothesis and research into web iv, the intelligent and agentic web. By 2030, I expect intelligent agents to handle a large part of the functional web on our behalf. They will help us find, compare, select and buy, while the human web remains a place for culture, identity, relationships and enjoyment. I am tracking the evidence for that transition through the web iv clock.
The branding question keeps coming back. I discussed it over lunch earlier this year with Josh Faulks, CEO of the Marketing Association of Australia, and in a recent meeting with Tom Goodwin. Both are people whose thinking I respect. Conversations like these help me develop my own thinking; the argument here is mine.
That raises a question for anyone working in marketing: what happens to a brand when the customer delegates the comparison?
Brands shortcut decisions when cognitive bandwidth is limited
The American Marketing Association defines a brand through the names, symbols and other features that distinguish one seller’s goods or services from another’s. The more interesting question here is what that distinction does for the buyer.
A familiar brand carries expectations. We remember an experience, recognise a reputation or associate it with people and values we identify with. It gives us somewhere to start and a reason to feel more comfortable choosing.
There is a substantial research foundation for that. In Brand Equity as a Signaling Phenomenon, Tülin Erdem and Joffre Swait explain how credible brands can reduce uncertainty, perceived risk and the cost of gathering information.
That makes sense. Brand is one way of shortcutting a decision because our cognitive bandwidth is limited. We have only so much time, attention and capacity to hold alternatives and trade-offs in our heads. We cannot investigate every product, supplier and alternative each time we need something. A familiar name lets us carry knowledge, experience and expectations from one decision into the next.
I do not want to become an expert in every category I buy from. There are only so many evenings in a week.
But brand also gives us meaning. Escalas and Bettman’s research on brand associations and identity examines how brands help people construct and express a sense of themselves. Advertising, culture, community and personal experience all contribute to that meaning.
The shortcut can carry all of that: an expectation of quality, a sense of trust, a personal connection. It helps us narrow the field before we have investigated every alternative. That is the part of branding I think agents will put under pressure.
Agents change the scale of comparison
Now put an agent into that decision.
A capable system, connected to suitable information, could compare product specifications, prices, availability, delivery times, service records, reviews and independent coverage at a scale no person could practically manage. It could assess those details against my needs, rather than requiring me to assemble and interpret them myself.
Electric cars make an interesting example. They combine a detailed technical purchase with questions of identity, trust and how we want to be seen. There is also a growing field of alternatives: the IEA counted nearly 1,000 electric car models globally in 2025, including battery electric cars and plug-in hybrids. That is a global total; an individual buyer’s available choices will be much narrower.
Ordinary software can do that arithmetic. The harder work is obtaining trustworthy, comparable evidence and correctly representing what matters to the buyer. Today’s agents can only work with information they can access, and finding a claim is different from verifying it. Greater processing capacity does not give them complete knowledge of the market.
Even so, the comparison could expand far beyond the brands and criteria a person can comfortably hold in their head. An unfamiliar car could make the shortlist because it fits someone’s life. They might still want to see it, drive it and decide whether it feels like them. But the brand has crossed the first barrier: it is being considered.
A brand could become relevant to me before it becomes familiar to me.
That is a meaningful change to how discovery works. Depending on the authority I give the agent, I might see a shortlist, receive one recommendation or approve a purchase without ever examining the alternatives.
Emotion can still shape the instruction
The tempting conclusion is that people choose emotionally and machines choose logically. I think that is too simple.
An agent does not need to feel my attachment to a brand to account for it. If I care about a particular design, a company’s purpose, buying locally or continuing a relationship with a business I trust, those preferences can become part of the instruction.
An agent can also learn from what I choose, reject and buy again. With access to that history and my feedback, it can infer preferences I have never explicitly stated and refine its understanding over time.
An emotional preference can therefore be represented in data, even if the representation is incomplete. Delegating the research does not require me to abandon the things I value.
There is also a difference between asking an agent to find a product from my favourite brand and asking it to find the best product for my needs. The first preserves my brand choice. The second gives the system room to introduce someone else.
That leaves brand with a powerful role before delegation: shaping what people want the agent to look for. It also creates an opportunity for discovery to work in the other direction. A good recommendation and experience could be the beginning of a new brand relationship.
A machine is not automatically impartial
We should also be careful about assuming that agents will reward the best match simply because they can process more information.
In research published at EMNLP in 2024, models showed brand biases across the categories tested, including more positive associations with global brands. A later experimental study of shopping agents found that choices varied with the model, product position and presentation. Those are findings from particular tests, rather than a verdict on every agent or purchase.
They are enough to challenge the idea of a perfectly rational machine comparing the whole market on our behalf.
An agent may inherit familiar brand associations. Its access to suppliers may be limited. The service operating it may have commercial incentives that differ from mine. A recommendation described as personalised still needs scrutiny.
The opportunity for smaller or unfamiliar brands is real as a possibility. It is not an automatic consequence of adding AI to the journey.
Brand has another audience to satisfy
My view is that the agentic web will fundamentally change the job of brand.
People will continue to care about identity, reputation, experience and belonging. Alongside that, brands will increasingly need their claims to stand up when a system examines them against a particular person’s requirements.
The evidence of what a business does becomes more consequential: whether its products perform, whether its service matches its promises and whether the values it communicates can be substantiated. Product information, customer experience and reputation all affect what an agent can establish.
If agents can represent what we value and find credible evidence of who delivers it, they could introduce us to brands we would never have discovered ourselves. That could change who gets considered, how trust begins and where loyalty develops.
The question I keep returning to is how much of today’s brand advantage comes from shortcutting a decision, and how much would survive a much more thorough comparison.
By 2030, I expect considerably more of that comparison to happen on our behalf. Brands will have to earn their place in it.
Reviewed 27 October 2026