From Amanda
Apple's new era: New CEO, new revenue problem, ads and a smarter Siri
There's a lot happening at Apple right now. The company just unveiled its first foldable iPhone, Siri AI rolled out in beta, we've officially entered the John Ternus era, and regulatory changes to Apple's App Store business model are starting to show up in the numbers.
App Store revenue declined for the first time in years with Apple's U.S. commission revenue falling 18% since January. Eric Seufert has argued that advertising is the only realistic lever for offsetting some of that pressure.
Cue ads in Apple Maps, which officially started rolling out in the U.S. and Canada. Businesses can now pay to appear in Suggested Places before a search, or alongside results after one, but the jury’s still out on how users will respond. There's also some whispering around Visual Intelligence. Code discovered in iOS 27 by developer Aaron Perris suggests Apple is laying the groundwork for sponsored results there too, though Apple hasn’t confirmed it.
These placements share a common approach to intent, which tracks with how Apple is approaching AI more broadly: keep as much intelligence as possible on the device.
Ads in Maps already do this. They're contextual and use signals from the current session, like your search, approximate location, and what part of the map you're viewing. Apple says it doesn't track users across third-party apps and websites or associate Maps ad activity with an Apple Account. Visual Intelligence could take that same idea even further, using what your phone already understands about what you're looking at to make the results, including sponsored ones, more relevant.
Siri AI shows the same instinct at work outside of ads. It can now use personal context and what's on your screen to understand what you want and take actions across apps on your behalf.
That might sound like Apple inching toward the same agentic AI model we're seeing with Meta and Muse, but Apple's approach is still very much device-is-king. Siri is deeply integrated into the operating system, and the iPhone itself is providing much of the context and orchestration behind what it can understand and do.
Eric Seufert noted an interesting consequence of that: If Siri learns which app you typically use for a certain task and starts choosing it automatically, it could decide which apps get the user's attention — and which don't.
Apple needs new revenue streams, and advertising is an obvious one. But Apple is also getting better at understanding what you want, when you want it, while keeping the device at the center.
Tech Giants
Meta just paid $17 billion over teen safety… then it launched Muse
Meta agreed to a landmark settlement of roughly $17 billion over teen safety , and now it wants the rest of social media to follow.
As part of the agreement, Meta is adopting new protections for teens, including default time limits, overnight restrictions, and stronger parental controls. It then publicly called on TikTok and YouTube to adopt similar standards, positioning the settlement as an altruistic blueprint for the entire industry.
I'm firmly in favor of stronger protections for kids and teens online, and whatever the motivations, there's something promising about Meta trying to make those protections the baseline.
But Meta’s timing is a bit ironic. Just as it tightens guardrails around teen social media use, it’s rolling out Muse, a new personal AI assistant that can take actions on your behalf, including sending emails, booking travel, and making purchases. Meta is putting more restrictions on one version of engagement while simultaneously building an entirely new one.
It’s worth noting that Meta has already been under fire for how younger users interact with AI. As AI assistants like Muse become more capable and more embedded in daily life, questions about healthy engagement, privacy, and protection aren't going away.
Advertising
Amazon and ChatGPT partner up for ads
Amazon Ads and OpenAI have launched a pilot that lets select U.S. advertisers extend their Amazon campaigns into ChatGPT Ads, with Delta Vacations among the first brands testing it. Advertisers will buy the ads through Amazon DSP, but OpenAI controls how and where they appear within ChatGPT Ads.
ChatGPT has access to incredibly rich commercial intent signals in the conversations themselves, and Amazon brings years of first-party data about what people actually browse and buy. Put those together, and you have the potential for a pretty powerful advertising combination.
Amazon has experimented with bringing its data to other platforms like Meta and Snap before, but those integrations never really scaled. ChatGPT gives it a very different kind of signal to work with. A conversation can reveal exactly what someone wants, what they care about, what they’re comparing, and how close they are to making a decision. That makes ChatGPT a meaningful testing ground for whether Amazon's model can work.
As conversational ads become a bigger part of the channel mix, marketers will need to know how those high-intent interactions translate into measurable outcomes across the rest of the customer journey, and how that spend performs alongside Google, Meta, TikTok, CTV, and everything else competing for their budget.
That's where an independent measurement layer becomes increasingly important. Mobile measurement partners (MMPs) like Branch already support measurement for Amazon and ChatGPT Ads, giving marketers the infrastructure to measure these emerging surfaces alongside the channels they already run.
Adam's Take
Today’s tl;dr on what’s working for LLM optimization
The new growth vector for discovery is chatbots. Duh. We researched what’s going on with LLM optimization, so you don’t have to:
- Self-promotion is starting to backfire. A recent study shows that brands that self-promote are left out of the actual recommendation two-thirds of the time.
- Third-party mentions still work. This is especially true for trust-specific verticals like consumer finance.
- Reddit is getting played out. “Reddit-maxxing” was the early breakout for juicing brand reputation, but the jig is up. Reddit is frantically fighting spam, but it’s too late. Share of ChatGPT citations are down 8X since July.
- Fresh content makes an outsized impact. Newer content and sources see outsized preference, which makes sense given how a model is trained at a point in time but needs to stay relevant for as long as possible. Models will continue to optimize towards fresh content to avoid becoming stale.
- Rewriting history has emerged as a new tactic. I'm getting emails about this already: vendors are asking/paying publishers directly to edit existing highly-ranked content to mention their product. We’ll see how long that lasts.
- Data structure is (still) important. LLMs won’t show what they can’t read. Your data needs to be in machine-readable formats. Think Q&A, competitive breakdowns, and "authoritative" takes.
- ChatGPT Shopping results now require vendor proactivity. Not pushing product feeds to ChatGPT? You’re now left out of results (see next article).
ChatGPT Shopping is increasingly leaning on integrated products for recommendations
ChatGPT Shopping mode — launched last year and coined “the future of agentic commerce” — is the feature that allows you to view, compare, and purchase products within your chat.
For ChatGPT’s near-billion users, this “feature” is a seamless part of their chat experience. Ask a shopping question, and ChatGPT will guide you through product recommendations. Most may not even know it functions as a separate technology.
To retailers, however, this is anything but simple technology. They first need to ensure their products are showing up in ChatGPT. At launch, this was easy: 100% of surfaced products were sourced from the web, meaning product description pages and Google Shopping results. Retailers didn’t need to do anything.
Last month, that changed dramatically. Now the majority of products surfaced come from Product Feeds, which are structured files that vendors provide directly to ChatGPT.
Recommendation bias towards Product Feeds is a rapidly increasing trend. Some analysts are already seeing 100% of results come from Product Feeds.
If you’re a retailer and you’re not uploading product feeds into ChatGPT, this is your sign.
Podcast
The End of False Choices: Putting Privacy First to Unlock Performance Marketing in Healthcare
Healthcare marketers are often presented with a false choice: protect patient privacy or run effective performance marketing. Jessica Holton believes they can — and should — do both.
In this episode of How I Grew This, Amanda and Adam sit down with Jessica to unpack the complex privacy landscape facing healthcare marketers, why consent management needs to be part of the marketing conversation, how brands can connect measurement across web and app, and why a privacy-first data foundation is becoming even more important as AI enters the marketing stack.