Your Next Customer Is an AI Agent

My first online purchase was in 2010, almost 16 years ago. I had saved for months to buy a smartphone. I was a student in Pakistan, and buying online was not the norm back then. It seemed risky, but liberating too.

I remember spending weeks gathering all the information I could on the handful of online stores, reading reviews and looking at forums. And I still paid cash on delivery, because I didn't trust any store with my savings until the phone was in my hands.

Today, an AI assistant could do those weeks of research, verify a website's authenticity and buy for you in minutes. As more of us, humans, rely on agents to buy, shop and help us run the world. Marketing, how we do it, is also gonna change.

I have been thinking about marketing in the agent's era for sometime. And here is how I think about it.

Marketing has always been about humans and their emotions. You move a human with your marketing, you get a customer. Now there is another player in the picture – the agent.

This is how I used my AI assistant over the past few days: "reorder the toilet roll," "find me a vacuum cleaner under $50," "get bedsheets like my last ones." My agent was responsible for browsing, comparing, and buying. If an agent is making decisions on my behalf, the marketing will need to evolve to please agents. And to get there, we need to understand how they shop.

How agents shop

Humans are emotional, they buy into a story, get FOMO'd, like a packaging, they love celebrity endorsements but Agents are not emotional, they are logical, and diligent. What worked for humans won't work for agents any more.

• Specs over slogans. "Luxuriously soft" is nothing to an agent; "100% cotton, 400 thread count" does.

• The impulse is gone. An agent sent for eggs doesn't come back with a chocolate bar.

• Reviews are important. Agents can analyse thousands of reviews in seconds. Heck they can smell fake reviews too.

The important thing becomes the product, its spec and how it is being talked about on the internet.

Brand still matters, but it is nuanced

Two things will likely happen:

1. If you tell your agent, "get me a Patagonia jacket," the brand has already won and the agent will follow that command.

2. If you didn't tell the brand name, the agent will find the brand for you and for that it will analyse all data points diligently to reach a conclusion. "Get me a jacket" might mean scanning 100 jackets, 10,000 reviews, discounts, offers, delivery availability etc.

So branding shifts from impressing a human to writing authentic information an agent can read. So, what people write about you counts for more than how you look in an ad. This is why AI visibility (GEO, or generative engine optimization) is now showing up next to SEO in marketing tools.

The new playbook

If the buyer is an agent working for a person, marketing will evolve into the following.

• Agents can read your product. Write and explain every feature of the product in a clean readable format, add structure data, add comparison tables, write clear pricing, make your whole page crawlable, explain shipping, explain return and refund policy. If an agent is not able to find the basic information, it might skip you all together.

• Reviews are back. Agents will analyse thousands of reviews in seconds from multiple sources. Invest in verified reviews early on, show crawlable reviews beside every product.

• Show, not just say. Agents don't mind reading terms and conditions, return policies, warranty claims, support

• Make everything verifiable. If you are saying your product is sustainable, ethically made, approved. You will need certification to prove that. Agents can read that in secs.

• Get mentions. Agents learn from community threads, expert comparisons, and credible publications. You need to be everywhere.

• Allow agent crawling. Ensure your website is crawlable by agents and lets agents read, scan, sign up, and purchase.

Who owns the agent matters too

If you are buying something low stakes, low ticket, the agent can make a decision. Also, agents will mimic your behavior, if your history shows the same store, same milk, same eggs, same chocolate, it will keep doing it until you ask it to stop.

If you are buying something high stakes, high ticket, the agent will take your opinion. But it will gather all the information I described above. It will read reviews, terms, comparisons and put forward recommendations.

Lastly, which brand assistant you use matters too. Is it by a marketplace or a platform (think Meta, Amazon, OpenAI), or Instinct or Apple Intelligence, each will have its own algo and incentives. A marketplace agent may favor its own product.

Just like in the early days of SEO, marketers will need to know which agents their customers use and what each one optimizes for.

What won't work

Some of the marketing playbooks will stop working.

• Attribution. Clicks, touchpoints will not matter because can agent

• Ads aimed at humans. The ads will shape-shift and will be optimized for agents. Ads will be shown where actions are.

• Gaming the system. Some sellers will hide text or instructions in listings to manipulate agents. Agents will get better at spotting it.

• Cold start problem. Agents will lean more towards established brands with a long review history, new companies will have to catch up fast!

Marketing to two audiences

If you are in marketing, your job is now to impress humans and their agents both. Your target humans will tell agents what brands they trust but agents will increasingly decide what gets finally bought.

How can you check if your brand is ready?

• Ask the assistant. Put your customers' questions to the major AI assistants and see how they describe you and your competitors.

• Fix your product data. Complete specs, structured feeds, and clear policies, everywhere you sell.

• Build a review engine. Get verified reviews from customers. Add them to review platforms.

• Track AI visibility. Measure mentions and citations in AI answers, not just search rankings.

• Pick a stance on agents. Decide whether AI agents can browse and buy on your site, and make it work if they can.

Sixteen years ago, I spent weeks on forums deciding whether a website deserved my savings, and still paid cash on delivery to buy a smartphone. Last week, I trusted an AI assistant with my card to buy groceries and help me buy a birthday gift for my best friend.