Your Next Customer Might Not Be Human

A marketing strategy built for humans is only half the job. The other half is convincing the machines.

I had a conversation with a product team last month that stopped me mid-sentence. They’d run an A/B test on their pricing page. Version A converted humans at 14%. Version B converted at 11%. They shipped Version A. Obvious decision.

Except it wasn’t. Because when they dug into the data, they found that Version B had been disproportionately recommended by AI-powered shopping assistants, browser-integrated recommendation engines, and automated deal-finders. The humans who arrived via those channels converted at 23%. The humans who arrived organically converted at 9% on Version A and 4% on Version B.

The AI agents preferred Version B. And the humans those agents sent were more valuable than the ones who found the page themselves.

Nobody had optimised for the agent. They’d got lucky.

You won’t always be lucky.

The Customer You’re Not Seeing

Here’s the uncomfortable truth that most marketing teams haven’t fully absorbed: the entity evaluating your product is increasingly not a human being. It’s an optimisation function.

Google’s AI Overviews don’t just summarise your content. They select it, rank it, and present a curated answer that often eliminates the need to visit your site at all. Perplexity doesn’t just search; it synthesises, recommends, and attributes (or doesn’t). Apple Intelligence is building a layer of personal context that will mediate between your brand and the person holding the phone. Meta AI sits inside WhatsApp, Instagram, and Facebook, increasingly the first point of contact between a consumer’s intent and your brand’s response.

And that’s before we get to the shopping agents. Amazon’s recommendation engine already drives an estimated 35% of the company’s revenue. But that’s a primitive ancestor of what’s coming. We’re now building autonomous purchasing bots that can compare, evaluate, negotiate, and buy without a human ever seeing your landing page. Your beautiful, carefully crafted, conversion-optimised landing page.

The AI is your first customer. And in many cases, it’s the only customer you actually need to persuade.

The Invisible First Click

I’ve written about this elsewhere as the Invisible First Click, and it’s worth explaining what I mean by it, because the implications are wider than most people think.

In the traditional model, marketing is a funnel. Awareness, interest, desire, action. You capture attention, build a narrative, create emotional resonance, and eventually, if you’ve done your job well, the human clicks. That click is visible. You can track it, attribute it, optimise for it.

In the AI-mediated model, there’s a click that happens before the click. It’s the moment an AI agent evaluates your brand against alternatives and decides whether to include you in its recommendation. That click is invisible. You can’t track it. You can’t see it in your analytics dashboard. You can’t A/B test it (at least not in the traditional sense). But it determines whether you ever get the chance to make your pitch to a human.

If Perplexity’s answer engine doesn’t mention your brand when someone asks “what’s the best student accommodation in Leeds?”, you don’t exist. If Google’s AI Overview surfaces your competitor instead of you, that’s not a ranking problem; it’s an existence problem. If the shopping agent filtering options for a young professional looking at build-to-rent apartments in Manchester doesn’t include your development in its shortlist, you’ve lost the customer before they knew you were an option.

The first click is the one that matters most. And increasingly, it’s not a human making it.

A Strategy for Two Audiences

This creates a problem that marketing, as a discipline, is not well-equipped to solve. Because marketing has always been built on a foundational assumption: you are persuading a human.

Every framework, every model, every creative brief starts from the premise that the decision-maker has a psychology. They have emotions, biases, motivations, fears, aspirations. They can be moved by a story, persuaded by social proof, triggered by scarcity, or delighted by surprise. The entire apparatus of modern marketing, from brand storytelling to behavioural nudge theory, depends on the decision-maker being a person.

What happens when the decision-maker has no psychology?

An AI agent doesn’t care about your brand story. It doesn’t feel the warmth of your community-focused messaging. It’s not moved by your beautifully shot campus video or your development’s architectural ambition. It’s running a multi-criteria optimisation: price, location, reviews, specifications, availability, and a dozen other quantifiable factors. If you’re optimising for emotional resonance and the gatekeeper is a spreadsheet with a personality disorder, you’ve got a mismatch.

But here’s where it gets genuinely complicated: the human is still in the loop. They’re just not the only one in the loop anymore.

You’re now marketing to two audiences simultaneously. The human, who needs to feel something. And the agent, who needs to find something. The brand that wins is the one that figures out how to satisfy both without compromising either.

The Psychology of Delegation

Not everyone relates to AI agents the same way. This is where consumer psychology gets genuinely interesting, because the willingness to delegate purchasing decisions to an algorithm maps directly onto deeper motivational patterns.

Some people delegate naturally and enthusiastically. These are the individuals driven by efficiency, by optimisation, by the desire to get the best possible outcome with the least possible effort. They’re the ones who set up automated purchasing rules, who trust recommendation engines without interrogation, who view an AI shopping assistant as a force multiplier for their own decision-making. For them, the agent isn’t replacing their judgement; it’s extending it. They’d happily let the algorithm buy their groceries, book their holidays, and choose their next apartment, because what they care about is the outcome, not the process.

Others resist, sometimes fiercely. These are the people for whom autonomy is non-negotiable. The act of choosing is as important as the choice itself. They want to browse, to compare, to feel the texture of the decision. Handing that over to an algorithm feels like giving up control, and control is the thing they value most. They’ll use search engines, read reviews, visit in person, and make up their own mind. The AI agent is a threat to their identity as an independent decision-maker, not a convenience.

And then there’s the majority: the people who don’t even know it’s happening. They’re not consciously delegating to an algorithm; they’re just living inside an algorithmically mediated environment and assuming they’re making free choices. The personalised feed, the curated search results, the “recommended for you” section: these are AI agents doing their work invisibly. The person scrolling through options on Rightmove or SpareRoom doesn’t think of the ranking algorithm as an AI agent making decisions on their behalf. But it is. Every ordering decision, every filter default, every “you might also like” is the agent shaping the choice set before the human even begins to evaluate.

For marketers, this means your audience isn’t one audience. It’s three, and they need fundamentally different approaches. The delegators need to see that your brand is agent-optimised: structured data, clear specifications, verified reviews, machine-readable trust signals. The resisters need human-first proof: community, authenticity, the feeling of discovering something rather than being served it. The unaware middle needs the illusion of agency while you quietly optimise for the algorithmic layer they don’t know exists.

The Loyalty Problem

Here’s the question that should keep brand strategists awake at night: what happens to brand loyalty when the customer has no emotional attachment to you?

Brand loyalty, in the traditional sense, is an emotional phenomenon. It’s built on repeated positive experiences, identity alignment, social belonging, and the cognitive biases that make switching feel costly even when it isn’t. People are loyal to Apple because it’s part of who they are. They’re loyal to Nike because the brand says something about them. They’re loyal to their student accommodation provider because it was their first home away from home.

An AI agent has none of these attachments. It doesn’t identify with your brand. It doesn’t feel nostalgia about living in your building during freshers’ week. It doesn’t care about your sustainability commitments or your community values unless those commitments translate into a measurable metric it’s been told to optimise for.

When the agent is the one making the recommendation, switching costs are effectively zero. If a competitor offers a 3% better price-to-value ratio, the agent switches. No emotional friction, no identity crisis, no “but I’ve always used them” inertia. Just a recalculation and a new recommendation.

This doesn’t mean brand is dead. It means brand has to work differently. In an agent-mediated world, brand equity isn’t just what the human feels about you; it’s what the agent “knows” about you. Your brand’s presence in structured data, review aggregators, authoritative sources, and machine-readable content becomes as important as your brand’s presence in human consciousness. Possibly more important, because the agent’s recommendation is the gateway to the human’s consciousness in the first place.

The brands that thrive will be the ones that build equity on two tracks: emotional resonance for humans, and informational authority for agents. The ones that only do one will find themselves either loved but invisible (great brand, no agent presence) or visible but forgettable (high agent ranking, no human connection).

Neither is a viable business.

The Double Trust Problem

Trust, in the AI-mediated landscape, has become a two-layer equation, and both layers are fragile.

The first layer is human-to-agent trust. The person has to trust the AI making recommendations on their behalf. This is where the current wave of AI products is most vulnerable. When Google’s AI Overviews hallucinate a confidently wrong answer, it erodes trust not just in that specific feature, but in the entire category. When a recommendation engine surfaces a product that’s clearly wrong for you, you don’t just distrust that recommendation; you start questioning whether the engine understands you at all.

This trust is earned slowly and lost quickly. A single bad recommendation can undo weeks of good ones. And unlike a human salesperson, who can read your reaction and adjust in real time, an AI agent doesn’t get the chance to recover gracefully. It either got it right, or it didn’t.

The second layer is brand-to-agent trust. The AI agent needs to “trust” your brand enough to recommend it. This isn’t trust in the human sense; it’s more like confidence. Does your brand have consistent, accurate, structured information across the web? Do your reviews aggregate favourably? Is your data clean, your specifications complete, your availability up to date? Are you present in the authoritative sources that the agent’s training data and retrieval systems draw from?

If the answer to any of these is no, the agent’s confidence in your brand drops, and so does its likelihood of recommending you. This is a trust relationship that most marketing teams aren’t even aware they need to manage.

The fragility is compounded by the fact that these two trust layers interact. If a human trusts the agent, and the agent doesn’t trust your brand, you’re invisible. If the human doesn’t trust the agent, and the agent does trust your brand, the recommendation is ignored. You need both relationships to be healthy for the system to work in your favour.

And both can break for reasons entirely outside your control.

The Paradox of Personalised Service

There’s a deeper philosophical tension at the heart of all this, and it’s worth naming, because it affects how you think about strategy.

We’re building AI agents to serve human preferences. The entire value proposition is personalisation: the agent learns what you like, what you need, what you value, and it finds the options that best match. It’s supposed to be the ultimate expression of consumer sovereignty, a world where every recommendation is perfectly tailored to you.

But the agents are simultaneously reshaping the very preferences they’re designed to serve.

When an AI agent consistently recommends certain types of products, it doesn’t just reflect your preferences; it trains them. The person who always sees mid-range options starts to think of mid-range as “their” price bracket. The student who’s only ever shown accommodation in certain areas starts to believe those are the only viable areas. The recommendations become a self-fulfilling prophecy, and the human’s preferences gradually converge with the agent’s model of those preferences.

This is the filter bubble problem, but applied to commerce. And it has real consequences for brands. If the agent’s model is accurate, you benefit. If it’s slightly off, you get recommended to people who aren’t actually your ideal customer, and your conversion rate drops. If the agent’s model is significantly off, you’re invisible to the people who would love you and visible to the people who won’t.

The brands that understand this will invest in feeding the agent accurate, rich, multidimensional information about who they are and who they serve. Not because the agent needs a brand story, but because the agent needs to build an accurate model of the brand’s ideal customer, and it can only do that if the brand gives it enough signal.

What Marketing Leaders Should Do Now

I’m not going to pretend this is easy. The landscape is shifting in real time, and the playbook is being written as we go. But there are things you can do today that will put you ahead of the brands that are still optimising exclusively for human eyeballs.

Audit your agent-readiness. Go to Perplexity, Google AI Overviews, and the AI shopping tools that are relevant to your sector. Search for what your customers would search for. See if you appear. See how you’re described. See what information the agent surfaces about you. If it’s wrong, fix it. If you’re absent, that’s a bigger problem.

Structure your data. AI agents need machine-readable information. Schema markup, structured product data, complete and accurate specifications, verified reviews in aggregatable formats. This isn’t glamorous work, but it’s the foundation of agent-mediated discovery.

Build two brand strategies. One for humans (emotional, narrative, identity-driven). One for agents (factual, structured, authoritative). The intersection of these two strategies is where the magic happens, but you need both in place before you can find the intersection.

Test for the agent layer. Start running experiments that account for the agent intermediary. Track not just conversion, but the path the customer took to reach you. Did an AI agent recommend you? Did the customer find you through a summarised answer rather than a direct click? The attribution models you’re using probably don’t capture this yet, and that means you’re flying blind on the most important part of the journey.

Invest in the trust relationship. Both of them. For the human-to-agent trust, make sure your brand is the kind of thing an agent can recommend confidently: accurate information, strong reviews, consistent presence. For the brand-to-agent trust, make sure the agent’s model of you is correct, complete, and up to date.

The Uncomfortable Conclusion

Here’s where I land on this, and I’ll admit it’s not entirely comfortable.

We are building a world where the most important customer interaction might be one where no human is present. Where your brand’s first impression is made on an algorithm that doesn’t feel, doesn’t aspire, and doesn’t care about your mission statement. Where loyalty is a function of data quality, not emotional connection. Where the entity choosing between you and your competitor is an optimisation engine running on criteria you may not even know about.

This isn’t a dystopia. It’s just the next version of the market. And like every previous version, it rewards the people who understand the rules before everyone else does.

The brands that win in this environment won’t be the ones that abandon human-centred marketing. They’ll be the ones that extend their thinking to include the non-human customers who are already shaping their fortunes. The ones that build for two audiences without losing coherence. The ones that understand the psychology of delegation, the fragility of double-layered trust, and the paradox of agents that reshape the very preferences they claim to serve.

Your next customer might not be human. But the one after that will be. And the trick is making sure both of them choose you.

David Chadderton spent his twenties and thirties teaching people how to make life-or-death decisions at forty thousand feet. He now applies the same principles to consumer psychology, which, depending on the brief, can feel equally high-stakes. He’s the creator of the STAR Framework and the author of The STAR Framework: Rewriting the Rules of Consumer Engagement (NYC Big Book Award 2025), The STAR Operating System: Decode Mindset, Understand Motivation, Transform Human Behaviour, and Dear Algorithm, It’s Not Me, It’s You. By day, a Chief Marketing Officer. By night, a behavioural science obsessive who writes The Unoptimised Human because he can’t stop thinking about why people do what they do.