In case you missed it, I have a new newsletter and podcast! It's called AdGentic, and it's all about the emerging ecosystem of ads in AI interfaces. That includes the regulars like ChatGPT and Perplexity, the AI surfaces inside of retailer apps, and – as covered in the first episode – ads in personal AI agents.
AdGentic is part of the ReFiBuy Media Network, and you can subscribe to it wherever you best like to take in this kind of content:
Here is the first edition, reproduced with permission, originally titled Enshittification is coming to personal AI agents.
Like everyone else, it seems, I recently got caught up in the euphoria of setting up my own personal AI agent and letting it loose on my busywork.
It was a magical experience. At first I chalked it up to the thrill of getting chores off my plate: email, scheduling, the usual stuff. But eventually it dawned on me how clean the form factor was, how dutiful the AI was, how it seemed to have only my best interests in mind. And then I clocked it. Part of that feeling came from what was missing. Ads.
No clutter, no distractions, no reminder that somebody, somewhere, wants to sell me something.
Enjoy it while it lasts, dear reader. Because I'm afraid it won't.
Ads are coming to personal AI agents. Here's why, and what they could look like.
Costco isn’t scaling its RMN with legacy adtech and manual solutions. It’s leaning into a cloud-centric, composable stack.
As part of the Costco Velocity Network, GrowthLoop empowers the retail media team to build audiences directly from the data cloud, enabling faster activation, greater relevance, and better performance — all while maintaining privacy and governance standards.
The result? Exceptional value and experience for Costco members and better ROI for brand advertisers.
Learn why the GrowthLoop Composable Commerce Media solution was the right choice for Costco’s Velocity Network.
A short history of broken promises
Every big consumer platform started out like this.
Google, 1998: The founders, Sergey Brin and Larry Page, published an academic paper describing their search engine. Tucked into an appendix titled "Advertising and Mixed Motives," they warned that "advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers."
Two years later, Google launched AdWords. At launch, the ads sat to the right of the results, "clearly separate," in Google's words. Over the next two decades they moved to the top of the page, and the label shrank to a small "Sponsored" tag.
Facebook skipped the no-ads promise and went straight to treating ads as inevitable. In 2008, Mark Zuckerberg told 60 Minutes: "There have to be ads either way because we have to make money." He even pitched them as an upgrade for users: "What would you rather see, a banner ad from Bloomingdale's or that one of your friends bought a scarf?"
(Remember the scarf. We'll come back to it.)
Then there's WhatsApp. Co-founder Jan Koum kept a note on his desk from his co-founder, Brian Acton: "No Ads! No Games! No Gimmicks!" Facebook bought WhatsApp in 2014. Both founders were gone by 2018, and in 2025 Meta announced ads in WhatsApp's Updates tab.
Amazon took a different road, because it started as a store rather than a media company. But in 2025, Amazon's ad business brought in about $68.6 billion in revenue. That's roughly double the combined operating profit of its North American and international segments, and that profit figure already includes the ads. It would be even for Jeff Bezos to imagine that his ‘everything store’ would eventually become the content. It’s the ads that pay the bills.
And the AI era is already following the script. In 2024, Sam Altman described ads as "a last resort" for OpenAI. But early 2026, ChatGPT was showing ads to its free users.
So that's the pattern. And it has a name.
The writer Cory Doctorow coined it in late 2022: enshittification. His definition: "Here is how platforms die: first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die." The word stuck. The American Dialect Society named it its 2023 word of the year.
And consumers resent it. Of course they do.
But despite all that, advertising is the most elegant way we have to pay for consumer technology.
Three ways to pay for an agent
I'm an advertising maximalist, and I'm with Eric Seufert on this one. As he put it, "Digital advertising is the most positive sum, economically expansionary technology summoned by humanity."
There are three main ways to pay for a consumer digital service:
- A subscription is a flat price you pay for access, whether you use the thing once a month or all day.
- A fee is a cut taken each time money changes hands: the service fee on your food delivery, the commission a hotel pays a booking site.
- Ads are paid for by a third party who wants your attention.
Ads paid for most of the internet you use today, and most consumer services end up mixing ads with one of the other two models. Mobile games: subscriptions or ads. Streaming TV: subscriptions or ads. Online travel agents like Booking.com: fees and ads. DoorDash: subscriptions, fees and ads. Ecommerce: increasingly ads, at least for the profit part. Social media: ads.
When social networks like X and Facebook have offered paid subscriptions, they have never come close to replacing ads as the core business. Subscriptions take off when a company is paying. But consumers are a different animal -- getting the mass market to dip into their post-tax spending dollars on a recurring basis is very difficult.
What today's AI companies have said
As of October 2026, the personal agents launching now are each trying something different.
At launch, Meta said its Muse agent would be free for a generous allowance, and that "over time we will profit by taking a small fee from transactions".
OpenAI's Dots launched for paying ChatGPT subscribers only, though Sam Altman added: "You should of course expect us to do a mass-market product for billions of people someday". OpenAI already runs ads in ChatGPT's free tier.
And Instinct, a startup that raised $1 billion at a $10 billion valuation, is free and hasn't said how it will make money.
Why ads will win
Here's my problem with subscriptions and fees: incentives.
Subscriptions cap out at the people willing to pay, and most people won't. When OpenAI announced ChatGPT ads, roughly 95% of its weekly users were on the free plan.
Fees are a sneakier trap. A fee pays the agent every time you buy something, which gives it an incentive to nudge you toward buying stuff.
Once merchants are competing for those transactions, they'll happily pay for a better spot in line. And paying for placement in front of a shopper kinda sounds like….an ad?
And according to research I co-authored with Bazaarvoice, 54% of AI users already believe that if they’ve seen a brand ad inside an AI tool, that brand is more likely to get recommended by the same tool.
Meta may well build its transaction business first. But notice the language: profit from fees "over time." Nobody has ruled anything out.
So ads are coming. The real question is, how bad will it be?
What the ads could look like
This is where my eternal optimism kicks in. Perhaps we aren't destined to repeat the mistakes of the past.
Think about why ads annoy you. Usually it's one of two things: it's not relevant to me, or it's getting in my way. Now remember the scarf that Zuckerberg mentioned. Plenty of us will begrudgingly admit that the ads on Instagram and TikTok are often (ok, sometimes) actually helpful.
An AI agent that really knows you has an even better shot at getting both right. It knows what kind of conversation you're having, what you're in the market for, and whether an ad would help or get in the way. That's where the real upside of ads in AI sits: in the targeting and the context. Those decide whether an ad feels annoying or helpful at least as much as the format does.
So far, though, the formats have been uninspiring, to say the least. ChatGPT's ads are a labeled sponsored box under the answer: a logo, a headline, a short description, an image and a link.
But we are starting to see formats that are native to AI. Sponsored prompts, which Amazon pioneered, put a brand-sponsored follow-up question into the conversation. Sponsored agents let people chat with an advertiser's own AI agent. And OpenAI is testing its first visual ad format in ChatGPT, which appears while someone waits for an image to generate.
Ads in AI can also take the best of the formats already out there:
- From retail media: sponsored picks in the agent's shortlist, and offers on the order confirmation.
- From gaming: rewarded ads. Watch or listen to an ad, earn more free usage.
- From streaming TV: an ad on the "working on it" screen while your agent books your flight, or a cheaper ad-supported plan.
And here's the neatest option: not showing any ads at all. Offsite advertising is where a retailer, or any company with rich user data, lets advertisers reach its users on other platforms like CTV, social or the open web. Advertisers pay for it because they can target precisely on what people have done and what they're likely to do next.
Now think about how much these chat assistants know about our behaviour and intentions. With an offsite ads program, AI companies would never need to interrupt a conversation with an ad.
There are two catches. The first is margin: offsite usually earns less, because the ad spend gets shared with more partners along the way. The second is privacy, and here the market matters a lot. Privacy laws in Europe and Australia are much stricter than in the US, so offsite would be a harder sell there. In the US, it's already part of the deal consumers strike with retailers, especially when they join a loyalty program, which is what retail media hinges on. An AI company using what it knows about you to put you in an advertiser's audience is not much different from a retailer doing the same with your purchases and searches. And the US is the consumer market most brands are trying to win.
And here's a couple more items on my personal wish list, which I'll be sending off to Ad-Santa for Christmas:
- Opt-in brand drops and sponsored announcements. My agent updates me when my favorite brands launch something new.
- My agent checking in with me about my current interests. Advertisers shouldn’t have to pay to target me for something I just bought. And continuing to see those ads are annoying. And yes, sometimes my 9 year old screws up my algo. It would be nice to be able to correct the record.
Enshittification is coming, but hopefully it's less shitty this time
The dark scenario, the one everyone is worried about, is ads that quietly steer my agent. Paid rankings I never see. A merchant paying to become my agent's default choice.
That's the real enshittification, and it's exactly the bias Google's founders warned about back in 1998.
So here's my test: visible ads beat invisible steering. Ads shown to me, relevant and clearly labeled, pass it. So does offsite, because it never touches what my agent recommends. Paid rankings dressed up as advice fail it.
I don't think ads are abuse. I think we might one day pine for the uncluttered, clean experience of today's early AI agents. But compared with the alternatives, ads are our least-worst option.
And one more thing...
I'm curious about what the possible impact of this proposed Instacar acquisition is for the brands that advertise on these platforms. Do you see it being a net positive or net negative? Answer the poll and see what your peers are thinking too.
