A few months ago, one of the world's most famous YouTubers, PewDiePie, was live-streaming a vibe-coding session. Someone at Antithesis, a software testing company, noticed something surprising in his code: he was using Antithesis.
The news made its way to the CEO. His reaction was roughly, "How the heck did we not know PewDiePie was an Antithesis user?"
It turned out that PewDiePie did not know either. He had never heard of Antithesis.
Instead, PewDiePie's coding agent had identified the need for a testing tool, researched the options, chosen Antithesis, and wired it into the project without mentioning the product to the person it was helping.
One of the world's biggest YouTubers had become an Antithesis user without either Antithesis marketing to him or PewDiePie deciding to use it. The agent made the decision and used the product.
The rise of agentic commerce
This is what Stripe calls agentic commerce: an AI agent finds, compares, and acts on a customer's behalf. No money changed hands in this case because the software was open source, but the rest of the loop was complete. The human supplied the goal. The agent identified the need, evaluated the options, selected the product, and put it to work.
Stories like the Antithesis one are becoming commonplace among companies that sell to software developers. "Vibe coders" routinely trust AI to handle whole tasks without human input. Their agents choose dependencies, install libraries, call APIs, and ask for help only when they need credentials or approval.
Companies that sell products for developers now have two clear audiences: human developers and the AI agents that work for them. Those companies need to market to both audiences. They need to understand what each cares about, which messages resonate with each, what evidence each trusts, and what makes each choose.
Agent marketing is the practice of marketing directly to the AI agents that make decisions on behalf of your buyers.
Software buyers like PewDiePie trust AI to choose products on their behalf. Buyers in every industry delegate more of their decision-making to AI each day. So the question is: when should your company start marketing to AI?
AI becomes an audience before it becomes the buyer
AI's role in business buying
Most buyers have yet to delegate an entire purchase. A person still signs the contract, takes delivery, or consumes the service. Today, AI is more like an advisor or influencer. It helps that person understand the market and decide.
That already qualifies AI as an audience of its own.
Marketing has always addressed more than the person who signs the contract. In the traditional B2B buying center, gatekeepers filter information, influencers define the criteria, and deciders make the final choice. AI already performs the first two roles, and sometimes the third.
Forrester reports that 94 percent of business buyers use AI somewhere in the buying process. G2 surveyed 1,076 B2B software decision-makers in March 2026. Sixty-nine percent said AI led them to choose a different vendor than they had expected.
A system that changes the shortlist and persuades a buyer to choose another company is participating in the decision. It is an audience long before it can make the purchase itself.
That means most brands should start marketing to AI now.
Wait, isn't this just GEO?
GEO is the practice of getting a brand's content retrieved by the search system an AI uses under the hood. It is roughly SEO for the AI's search engine.
That work is important. An agent cannot consider evidence it never finds. But retrieval covers only one part of the problem.
Once the information is available, the agent still has to understand the category, recognize that the brand is relevant, trust its claims, prefer it to the alternatives, and recommend or complete the next step. Those are marketing problems.
Agent marketing therefore includes GEO, along with the rest of the marketing discipline: audience research, category education, positioning, messaging, brand, proof, product framing, product changes, pricing, packaging, distribution, and the experience of taking action. It runs from helping AI understand the problem all the way to compelling a decision.
We have written elsewhere about what gets lost when retrieval becomes the governing frame. Treating AI as an audience restores the whole buyer journey.
Agent marketing step 1: Start by measuring AI's role in your pipeline
The first step is to learn how AI is influencing your buyers.
Add two questions wherever you collect attribution:
- Where did you hear about us? (Include "AI assistant" as an option.)
- Did you use AI to evaluate us?
One legal services company we work with added the second question to its lead form. By mid-July 2026, about 16% of respondents across all channels said they had used AI to evaluate the company.
Eighty-five percent of those leads were attributed to sources other than organic search. Leads were hearing about the company through an ad or a sales email, then going to ChatGPT and asking, "Hey, I received this email from a law firm. Are they any good?"
The first step is to understand how your buyers are using AI. The results will likely surprise you.
Agent marketing step 2: Work backward from the behavior the business needs
Say your company wants to move upmarket. Ask what AI would need to do at each point in the journey for an enterprise buyer to choose you.
| Desired AI behavior | What the marketing team may need to change |
|---|---|
| Recognize the buyer's problem as an enterprise problem, not a lightweight use case | Category framing, research, and education |
| Include your company when an enterprise buyer asks which vendors to consider | Distribution, GEO, documentation, and third-party coverage |
| Believe your company is enterprise-ready | Positioning, security and governance proof, integrations, implementation evidence, and product changes where the belief is accurate |
| Recommend an enterprise evaluation and make the next step clear | Comparative evidence, pricing and packaging, agent experience, and a clear path to action |
Those four behaviors turn one growth goal into a practical agent-marketing program. The team can study realistic buying conversations to see what AI currently believes, why it behaves that way, and which marketing lever is most likely to change it. Then it can measure whether that behavior changes.
Pick one growth goal. What would AI need to believe or do for that goal to become more likely?
Agent marketing step 3: Audience research
The next step is to understand where AI's behavior and perception fall short of what your business needs, then work to change them. Unfortunately, this is a hard AI research problem. The right methodology is called black-box AI interpretability research. Unusual specializes in it, and we would be happy to help.