Your customer has company: marketing for humans, algorithms and agents

Lately, I’ve been thinking about who we’re actually making our marketing for, because the answer is getting a bit more complicated than the audience personas in the brand strategy, or the targeting slides in a marketing plan. 

There’s the human, obviously. The person with a problem, a budget, and probably seventeen other things on their mind. But between that person and your business, there are algorithms deciding what gets seen, and increasingly, AI tools helping them work out what to buy.

Which means the linear media model is kaput and we’ve now got three layers to think about: the human we want to reach, the algorithms distributing our content, and the AI systems interpreting our offer. They overlap technically, but they do different jobs. And accounting for all three is becoming part of the marketing brief.

As if the brief wasn’t already tangly enough.

The person, the feed, and the shortlist

Say you’re looking for under-eye patches for the dark circles brought on by thinking about all of this marketing stuff. Hypothetically, of course.

You might see a video while scrolling, search for a product, or ask an AI assistant to compare options for sensitive skin within your budget. The need is much the same, but the way a brand gets considered changes.

For the human, the product needs to make sense. What does it do? Will it suit me? Is there a reason to believe the claims? The brand, the explanation, and the evidence all have work to do.

If you’re reaching that person through a feed, there’s another set of considerations. Meta describes using signals and predictions to rank content, including recommendations from accounts someone doesn’t follow. Publishing something gives the system something to assess; it doesn’t reserve a place in your customer’s day.

That’s why we spend so much time thinking about the creative — the hooks, openings, formats, and what makes someone keep watching or share something. But a post can be very good at attracting attention and give people absolutely no reason to buy your eye patches. The algorithm’s enthusiasm is lovely. And vanity metrics are a massive dopamine hit. You still need a customer.

Then there’s the AI-assisted route. Someone might ask for a comparison and get a shortlist without browsing all the websites themselves. An agent could go further and carry out parts of the task on their behalf.

Now your product information has to support that process. Ingredients, price, availability, delivery, and suitability need to be available and accurate. A lovely photograph and a line about ‘glass skin’ leave rather a lot of homework for everyone involved.

Getting onto the AI shelf

Deana Burke coined the idea of an ‘AI shelf’ and even did an experiment building a fake brand and optimising the content to get recommended by AI engines. I really like the concept of the ‘AI shelf’ as a way to picture this. Your product could be considered somewhere the customer never encounters your carefully designed product page.

There’s actual infrastructure being built for it. Shopify’s agentic storefronts use its product catalogue to make products available through participating AI channels, with structured product information supporting discovery. 

But there isn’t one universal AI shelf, or one clever trick that gets you onto it. Different systems find and use information differently. Google explicitly says its AI search features require no special schema or additional optimisation beyond the existing search requirements.

So I’d be wary of anyone selling certainty here. What I would take seriously is whether your business can be understood by agents and robots from the information you’ve made available.

Can someone establish what you sell, who it’s suitable for, what it costs, and why they should trust it? Does your website agree with your product listings and the information elsewhere about your brand? Can the relevant systems access it?

These sound like basic questions. They are also questions that some very expensive websites answer remarkably badly.

Give all three a job in the brief

The temptation is to turn this into three more workstreams for an already stretched marketing team. Another optimisation exercise, another report, another person asking if we’ve done the AI thing yet.

I think it’s more useful to bring these considerations into the work we’re already planning.

Start with the person and what would help them choose. Then consider how they could encounter the offer, and what the systems involved need in order to surface or represent it properly.

For our hypothetical eye patches, that might mean a useful demonstration for someone scrolling, a clear product page for someone comparing, and accurate catalogue information for an AI-assisted recommendation. The same offer needs to hold together across those encounters, even when the format changes.

That’s the strategic job I think we need to get better at. Making something appealing to a person, earning distribution, and supplying usable information are connected parts of reaching a customer.

And the customer is still the reason for doing any of it. We need to understand their circumstances well enough to make a useful offer, then make sure that offer can travel through the systems they use.

Your customer has company. So, we’d better account for it.