When agents start selling: How ad-supported AI will reshape commerce

‍If you watched Super Bowl LX, you probably remember Anthropic running ads (not so) gently poking fun at OpenAI's move toward advertising.

It was cringey and clever. But it was also strategic. It confirmed something anyone in media already knows: when ads enter a new medium, the debate gets loud fast.

Sam Altman, the CEO of OpenAI, famously called advertising "a last resort." But in January 2026, it became the plan. Sponsored messages started appearing at the bottom of ChatGPT responses for free and Go tier users in the US, clearly labelled as ads. The economics, it seems, don't care what you said in a previous interview.

I've spent more than two decades watching platforms reshape how brands reach people. The internet upended print. Social media upended the internet. Mobile upended social. But the way advertising works in an agentic world is unlike anything we've seen before.  

In the past, an advertiser targeted a person to compel a decision. Now, the target is the decision engine acting on that person's behalf. As consumers fine-tune their agents, advertising stops feeling like advertising and starts feeling like recommendation — sometimes indistinguishable from the rest of the output.  

And this is where global merchants have the most opportunity — not in mastering a new mechanic, but in recognizing that the game, fundamentally, hasn't changed.

Brand loyalty is the algorithm's input

There's a misperception spreading through marketing circles: once agents make purchase decisions, brand becomes irrelevant. Why invest in emotional connection when an algorithm decides?  

That logic isn't just wrong — it's upside down.

Agents don't invent preferences. They reflect them. They are, at their core, a mirror of what users have told them to value — the brands they trust, the products they've loved, the emotional associations built over years of brand investment.  

In the agentic era, brand loyalty doesn't disappear. It becomes the input.

The clearest illustration is Spotify Wrapped. Once a year, for 751 million people, Spotify creates a moment people share, screenshot, and talk about for weeks. What makes it work isn't the algorithm — it's the emotional storytelling, built on actual user behavior. That genuine connection shapes preference. And in an AI-mediated world, preference shapes agent behavior.

The key word is "genuine." Think back to the social era, where influencers acted as the trust layer between brands and consumers. In 2017, the FTC sent warning letters to celebrities including Kim Kardashian for failing to disclose paid Instagram endorsements. Influencer marketing didn't slow down, but the rules changed fast. The lesson: if platforms won't enforce transparency, regulators will.

In an AI-driven world, the same dynamic plays out at a deeper level.  

Agentic commerce is not the death of brand — it's brand operating at a level where authenticity is the only currency that compounds.

Transparency is the foundation

When OpenAI announced it was testing ads, researcher Zoë Hitzig resigned the same day, writing: "OpenAI has the most detailed record of private human thought ever assembled. Can we really trust them to resist the tidal forces pushing them to abuse it?" Users feared the thing they'd come to rely on would be corrupted by commercial pressure.

That fear follows a familiar arc. When TV ads appeared, people felt their private family time was invaded. When sponsored posts arrived on social media, consumers felt misled. Regulators eventually stepped in — the FTC's Endorsement Guides, first issued in 1980 and significantly updated in 2009, established that paid relationships must be disclosed. None of that accountability was volunteered. It was extracted from platforms that moved fast and hoped nobody noticed.

AI will face the same reckoning. The question isn't whether advertising in AI is ethical in principle — it can be, with clear disclosure.  

The question is whether platforms build the right guardrails before regulators force their hand.  

User controls — the ability to ask why something was recommended, the ability to opt out — are not optional features.  

In AI, they're structural requirements for trust.  

Earned trust is the new SEO

The Financial Times reported that OpenAI operated at a loss of around $8 billion in the first half of 2025, with only 5% of its 800 million users as paying subscribers. Platforms at that scale need sustainable revenue, and advertising has always been one of its most reliable engines.  

In fact, visibility might be the most underestimated dimension in discussions about the agentic world. Showing up in AI recommendations isn't just a function of paid placement or search optimization — visibility depends on the signals AI systems trust: earned media, independent reviews, credible external voices. Think G2 for software, trusted review ecosystems, earned editorial coverage.

Organic third-party endorsement may soon be more valuable than any paid placement, because it's the kind of signal an agent can rely on without questioning motive. Earned credibility, as anyone who had worked in communications instinctively knows, compounds. The merchants most at risk are those who've relied on owned and paid channels without investing in it. Those with a structural advantage have cultivated genuine quality, real advocates, and independent voices that algorithms can surface with confidence.

Agentic commerce doesn't change what credibility requires.  

But it does make the absence of it more expensive.

The future belongs to brands that don't need to shout

Five years from now, I expect consumers' wariness about AI agents — am I relinquishing control? Am I being manipulated? — will look a lot like the fear people once had about entering those three digits on the back of a credit card during online checkout.  

Today, no one gives it a second thought. The shift with AI will happen faster.

As disclosure becomes standard and regulation catches up, the relationship with AI will move from wary to trusted. People will tune their agents to reflect their values. The agent becomes an extension of the self, not a corporate intrusion.

The technologies change. The fundamentals don't.  

For payment infrastructure like Nuvei, the goal is to be invisible: reliable, secure, seamless, unremarkable. In agentic commerce, where a human isn't watching every transaction, that standard rises. A technical glitch doesn't just create friction — it sends a signal - and signals get learned. In agentic commerce, the payment layer isn't a back-office concern anymore.  

It's brand.  

For any merchant preparing for agentic commerce, I’d give the same advice I gave brands back when the internet upended print and social media upended the internet:

1. Build trust in your brand.
2. Deliver consistently.  
3. Be transparent.  
4. And make the path to purchase seamless.  

The merchants who struggle will be chasing new mechanics without anchoring to those principles.  

The ones who win understand that the new mechanics exist to serve the same old truth: people buy from brands they trust.

And soon enough, so will their agents.