TL;DR: Ads are turning up in AI products—starting with Google’s AI Mode chatbot for search. The rollout is happening slowly, but the tension is already clear: AI is often marketed as a tool you can trust to cut through the noise—not another feed filled with ads. Now it’s quietly becoming just like the rest of the internet.

What happened: During a Sunday briefing, Google told advertisers they could now buy placements for personalized offers and coupons directly in the Gemini-powered AI Mode search results. Google brands these ads as “contextual”—meaning they surface when users ask for product recs—and insists they don’t influence the core answer.

Here’s how this looks in practice: AI Mode offers a few product suggestions based on your search, and when you click on one, Google lets you know that there’s a sponsored deal available—say, 20% off the washable rug that you were clearly already interested in.

Perplexity has also experimented with sponsored follow-up questions next to AI-generated answers. But Google’s making the largest splash because it’s the biggest ad seller in the world—so dominant that the Department of Justice is actively suing it for allegedly monopolizing the digital ad market.

Who asked for this: We’ll give you one guess. If C-suite execs had their way, every AI answer would probably be so stuffed with ads that it would make Times Square look austere. AI companies need to make money, desperately, and integrated ads are a great way to do that. Unfortunately for them, it’s also a great way to turn users off. OpenAI hit a nerve last month when ChatGPT randomly suggested a user buy a Peloton. (The company scrambled to clarify that it wasn’t an ad—just a nudge to showcase new third-party app integrations).

Research shows that people’s reactions to ads depend heavily on placement: TV ads when you hit pause are generally fine. But we get more frustrated with obvious, disruptive banner ads. And with AI, the stakes shift again. Companies want their chatbots positioned as trusted, unbiased helpers (despite painfully visible inaccuracies and hallucinations). Users also often ask them sensitive questions, including about health worries and personal dilemmas. But a recent study found that when hypothetical sponsored suggestions appeared in ChatGPT convos, users felt misled and trust dropped sharply—even for recommendations that weren’t ads.

What comes next: This is just the first step—more AI companies are likely to follow Google’s move (like OpenAI has floated a few times), and we’ll probably see ads not only in chatbots but also “agentic” AI shopping tools that nudge you toward sponsored products and walk you through checkout in one frictionless swoop. And while the FTC requires “clear and conspicuous” disclosure of anything sponsored—AI companies can’t shrug and say “the algorithm picked Amazon-brand boxers” if compensation or preferential placement is involved—enforcement is messy in practice. —WK

Convenience is the real addiction

Sometimes you just need to rant. Here, we give voice to a tech-related issue that’s been bugging us.

Tech makes everything so easy that we’re living in a friction-free fantasy world, and honestly… it’s ruining us. It’s so good at eliminating effort that we’ve forgotten how to tolerate a little inconvenience (and the joy that comes with learning or the unexpected).

We used to navigate our way through awkward conversations, boring errands, and getting lost because that was life. Now we outsource nearly all that discomfort to apps and algorithms—ChatGPT tells us how to respond to someone, Amazon automatically restocks our groceries, and Google tells us how to get somewhere, step-by-step. Sure, some of it is genuinely useful. But it wouldn’t hurt to add a little more friction to our lives. “Friction-maxxing,” if you will.

I’ve thought about all the ways I’ve intentionally added digital friction to my life. Years ago, I removed my credit card details from Google Chrome so it can’t autopopulate checkout fields. Yes, it annoys the hell out of me to actually dig my wallet out for every purchase, but guess what? My impulse buys have plummeted because I’ve reintroduced, gasp, effort. There’s a weird satisfaction in typing out card numbers like an analog rebel. I've also been using Opal, an app that blocks other apps (like social media) to add friction to the part of my brain that wants to pick up my phone every five seconds.

Relying less on tech and doing more for ourselves could prove to be crucial in the future. A study from MIT suggests that relying heavily on AI tools like ChatGPT can actually reduce brain engagement and weaken memory over time. Instead of letting tech do the thinking, maybe we should embrace a little friction. It hurts in the moment, but maybe that’s the point. —SM

If you have a funny, strange, or petty rant about technology or the ways people use (and misuse) it, fill out this form and you may see it featured in a future edition of the newsletter.

Chaos Brewing Meter:

If you, like me, have a bad Instagram habit, you might have done a double take at a pic of a young woman in bed with deposed Venezuelan President Nicolás Maduro. Or Lebron James. Or The Rock. In the latest mutation of the deepfake problem, a new 404 Media report says the social platform is now teeming with influencers posting deepfakes that imply they’re having sex with real celebrities. The kicker? The influencers are also not real people. They’re AI-generated and using the posts to rack up engagement—from people who either can’t tell what’s real or who enjoy watching a digital tire fire—with the end goal being to funnel eyeballs to paid adult content sites.

You might call these deepfake trends disturbing, toxic, or absurd. Another word for it: defamation. Several celebs have already sued or sent cease and desist letters for their work or likeness being used in AI, including Lebron James (whose AI deepfakes are truly unhinged). Nonconsensual intimate images are already banned on Instagram, but they keep popping up anyway.

Meanwhile, the US Senate is spinning to contain the deepfake flood. It just passed a bill that lets victims bring a federal civil lawsuit for nonconsensual AI deepfakes “depicting intimate activity or nudity.” If the bill becomes law, bots won’t just get banned; the human producers and distributors behind them will get dragged into court. —WK

• Do you really have 60,000 employees if 25,000 of them are AI agents? McKinsey's CEO thinks so.
• Critical update: The “Are You Dead?” app is changing its (objectively excellent) name. And we don’t know how to feel about it.
• OpenAI’s first AI-powered gadget could take aim at one specific Apple product.
• Curious about the new agentic Slackbot? Here’s what it’s good at.
• If you’ve ever been to a T-Mobile store and thought, “This experience would greatly improve with a hologram,” you’re in luck.
• Here’s how much time you’re actually spending reworking AI content at your job, according to a new survey from Workday.
• File this under least surprising things we’ve read this week: Roblox’s age verification system is a mess.