The AI lifestyle subsidy is going to end
Or, why your digital experiences are about to get worse.
Whether you’re enthusiastically generating or reluctantly dismissing another pop-up for a useless chatbot, right now is the best it’s ever going to be. Oh, the technology is going to get better. But what about your experience as a user?
We lose money on every sale, but we make it up in volume
If you were alive in the bygone era of 2022, you might remember Derek Thompson writing about The End of the Millennial Lifestyle Subsidy:
…if you woke up on a Casper mattress, worked out with a Peloton, Ubered to a WeWork, ordered on DoorDash for lunch, took a Lyft home, and ordered dinner through Postmates only to realize your partner had already started on a Blue Apron meal, your household had, in one day, interacted with eight unprofitable companies that collectively lost about $15 billion in one year.
Derek Thompson via The Atlantic: The End of the Millennial Lifestyle Subsidy
The lifestyle subsidy was a great deal for consumers. Venture capital and zero interest-rate policy supplied a ton of money for any startup that could demonstrate strong growth - even if they were growing by selling customers $20 of services for $10. “We lose money on every sale, but we make it up on volume” was not just a surprisingly-old joke (first recorded in 1833!), but the real pitch to investors.
Is it any different this time around? Interest rates aren’t what they used to be, but investors are still scared of missing out on the next big thing. That investor money is going straight into subsidizing the costs of LLM-based products. Today, dedicated groups are actively farming free trial accounts from AI startups and using them to resell access to Anthropic and OpenAI APIs. (Hopefully those startups realize, instead of happily watching their signup number go up).
But besides unit economics, consumer AI applications present a new opportunity to control the way that people discover things.
Search engines and social media aren’t as good as they used to be
In the original Google research paper (1998), Sergey Brin and Larry Page wrote:
The goals of the advertising business model do not always correspond to providing quality search to users… we expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers.
Sergey Brin and Lawrence Page: The Anatomy of a Large-Scale Hypertextual Web Search Engine (quote from Appendix A)
Within two years, Google had launched AdWords, and some years later they acquired DoubleClick, setting the stage for Google to be one of the world’s largest and most profitable advertisers today. Now if you search for any company name, you’ll probably see:
An ad for a competitor
An ad from the company itself, desperately fighting for space above the fold
An ad for a phishing site that steals your password and even relays your one-time passcode
And if you scroll down a bit, the actual link to the actual company website
I remember a time when you didn’t need to install an adblocker to get decent Google results without scrolling. But we don’t live in that time anymore.
As people got better at search engine optimization, it became harder to find product reviews that weren’t actually affiliate marketing. Top 5 printers in 2025: best value, best quality, best for photos, best for feng shui, best at some other arbitrary qualification so I could include one more paid link. Sure, go get your money, but I kinda wanted to hear less-biased reviews.
There was also a moment in time when you could search for “printer recommendations reddit” and hear from mostly-real people. But now there’s dozens of tools to monitor relevant social media posts, so FengShuiPrinterCo can swoop in at the right moment (and if you’re lucky, they’ll disclose their affiliation). In more proactive moves, some companies actively moderate relevant Reddit communities to secure their brand dominance.
AI discovery is only going to get worse from here
(June 2025) ChatGPT and friends today are powerful token-generating machines that can use their existing body of trained material, search the internet, and synthesize a reasonable summary to start. It’s a great tool for discovery and learning, especially in new areas. You can follow citations back to the original source (which you want to do, a lot, for when it’s wrong) and ask follow-up questions. I put a date on this claim because I think it could really change a lot in the next six months.
On the marketing side, venture capital firm Andreesen Horowitz is already pushing Generative Engine Optimization (GEO) as the successor to social media and search engine optimization. They’ve got a logo wall with dozens of companies to make sure that when you ask ChatGPT about what printer to buy, it will make sure you’ll get a ranking based on feng shui.
More worryingly, the AI giants themselves will move into ads. The recently-hired CEO of Applications for OpenAI is Fidji Simo; she led ads at Facebook for a decade and most recently led Instacart out of the “millennial subsidy” era and into profitability with… more ads. Kevin Zhang (East Wind) argues that the current valuations of Anthropic, OpenAI, Mistral, and other “frontier AI” companies can only be justified if they can monetize free users through ads. There is too much money at stake.
Generative ads are going to be ugly
At least the Google ads say “Sponsored”. If someone is astroturfing on social media, you can correlate it with their other activity to find patterns. And influencers might not always disclose their sponsors, but they’re public figures who need to stay on the good side of their platforms.
By default, every LLM chat is 1-on-1, so you can’t correlate suspicious activity. Conversation has no concrete format to put a “Sponsored” tag on. LLMs are wrong all the time already - if you get a recommendation that’s factually inaccurate, how can you distinguish hallucination from payola? In fact, I bet marketers are already doing “blackhat GEO”: detecting AI crawlers and serving wildly-exaggerated pages to boost their reach; humans get the pages with normal levels of marketing-speak, of course.
Forget about superintelligence; for ads, “agent” and “alignment” are exactly the right terms. In economics, the principal-agent problem describes a conflict of interests when you (the principal) ask the agent to do something, but the agent’s incentives are not aligned with yours. For now, the AI serves you and your needs; but in the near future, the AI will serve ads.
Any silver linings?
I’m pretty skeptical that there will be any mainstream consumer AI applications that avoid ads. Most people prefer not to pay any money if possible (and even paid products like the Amazon Kindle are ad-supported by default, with an extra fee to remove ads). The flywheel of cash is going to spin as ad revenue becomes product investment. Eventually, we might see industry standards and regulations about ad disclosures, but they’d need serious enforcement options to make any real difference.
On the other hand, there should still be two niches that remain free of first-party ad pressures. The first is fully-paid services for privacy enthusiasts and purists, like Proton Mail for email or Kagi for search today. The second is open-source models and applications for techies, like llama.cpp. Both of these will lag behind the big AI companies initially, but carve out their independent advantages over time.
—
Until then, enjoy the AI lifestyle subsidy. It won’t be around forever.
A story from 2021 or so, a great time for lifestyle subsidies: I got ads for different grocery delivery startups every day. Finally, I scanned the QR code on a postcard in my mailbox that promised $50 off my first order. A guy on a motorbike delivered a couple steaks in 10 minutes flat and the venture capitalists paid for the whole order. And then, the entire startup shut down within a year. I really don’t know what they were expecting.
I should have bought more steaks with investor money back then. What’s the equivalent today?