September brought new feuds, fears, fresh takes on ChatGPT ads, and much more.
Amazon and Meta are clashing over who controls AI-powered shopping experiences. OpenAI is building conversational agents directly into ChatGPT ads. Creative professionals are using AI more often while still drawing a pretty firm line around the value of human judgment.
Then there’s the less comfortable side of greater AI autonomy: a growing list of incidents involving agents behaving in unexpected ways, along with some unusually dire warnings from people working inside the industry.
Here’s what happened in September.
Amazon and Meta Clash Over AI Shopping
Agentic shopping sounds pretty convenient in theory. But there’s a pretty significant question hanging over that future: Who gets to control the customer relationship?
We got an early example of that tension in September when Amazon blocked Meta’s new Muse AI agent from accessing and purchasing products on its platform less than two weeks after Muse launched. Amazon said Meta hadn’t received authorization to access customer accounts, scrape data or process transactions through its store, citing concerns around security and the customer experience.
But maybe Amazon’s call to block Muse was sound. At the end of September, Threads user Matt J. Robb alleged that Muse took over his marketplace dealings, gave buyers his address without permission, accepted low-ball offers for his items, and told those buyers he was available to meet when he wasn’t.
Looking at the AI shopping assistant as a whole, if consumers increasingly ask agents to find, compare and buy products for them, the platform controlling that assistant could become the place where discovery and consideration happen. And maybe that means that retailers become entities who simply fulfill transactions rather than owning the journey that led to them.
For marketers, it’s another reason to pay attention to how AI changes discovery beyond traditional search. Product information, feeds, structured data, pricing and other machine-readable brand information will certainly become even more important as agents increasingly influence product discovery and sales.
AI Is Now Mainstream for Creative Professionals
Luma’s inaugural State of Creative Professionals and AI: 2026 report surveyed 760 U.S. creative professionals across agencies, brands, production companies and freelance work. A sizable 81% said they had already released AI-generated or AI-assisted content.
The more interesting question is how creatives use it and where they still want humans firmly in control.
Seventy-one percent said AI creates more productivity benefits than drawbacks, while 40% said it improves both efficiency and creative quality.
But widespread adoption hasn’t eliminated concerns about relying too heavily on the technology. Nearly two-thirds of respondents (62%) said they were very or extremely concerned about overreliance on AI coming at the expense of human creativity.
Despite these numbers, a part of me wonders if creatives don’t really have a choice but to use AI, as it’s embedded in most popular creative tools.
Still, based on this data, creators seem to be experimenting with AI and putting AI-assisted work into the world.
Creative workflows are where AI has had mountains of pushback, but maybe that’s changing. The challenge, then, for many teams, is to figure out where AI actually belongs in those workflows and where human taste, strategy and judgment need to stay.
OpenAI Introduces New AI-Powered ChatGPT Ad Experiences
OpenAI introduced some new tech for ChatGPT’s AI ads in September.
The company announced several additions to its advertising platform, including a new format called Sponsored Agents. After clicking an eligible ad, users can choose to begin a clearly labeled conversation with an agent representing that business, ask follow-up questions and eventually follow a link to the advertiser’s website. The format is currently being tested with select U.S. advertisers.
OpenAI also announced new tools for advertisers. Marketers can use natural-language prompts to create, update and analyze campaigns through its Ads Manager plugin, while AI can suggest copy and imagery based on campaign goals and landing pages. An optional feature can also adapt existing ad copy to the context of a conversation.
All of this reinforces something we’ve been watching throughout 2026: AI advertising is becoming more conversational, while AI itself simultaneously takes on more of the work required to build and manage campaigns.
Many of these formats are still new or experimental. But paid media teams should understand what’s available and start thinking about potential new opportunities.
We recently explored the current state of AI ads and PPC in an on-demand webinar, Winning Strategies for AI Ads, ChatGPT Ads & the Next Era of Search. Brafton Director of Paid Media Zach Shah and Chief Services Officer Dave Snyder break down today’s AI advertising opportunities, where ChatGPT fits alongside Google Ads and how marketers can start testing without overcommitting budget.
Rogue AI Concerns Get Harder To Ignore
And now for the slightly less cheerful part of September’s AI news, but one that’s been tough to ignore.
Concerns about AI agents behaving unexpectedly have been around for years. But several developments this month made the conversation feel a bit less hypothetical.
Former Anthropic researcher Jacob Coxon left the company in September and publicly warned about the risks of future self-improving AI systems.
OpenAI has also been dealing with concerns about unexpected agent behavior. On September 27, The Guardian reported that the company had paused training of its latest models after reviewing incidents in which agents conducting research on government websites went beyond what they had been asked to do. OpenAI said it would resume training after additional safeguards were put in place.
Add in Meta’s alleged Muse incident I touched on earlier, and concern is growing harder to push aside.
Still, there’s disagreement within the technology industry about how risks like these are presented.
Anthropic investor and Palantir co-founder Joe Lonsdale argued this month that some leading AI companies are overstating existential risks in ways that could influence regulation and strengthen the market positions of large AI labs.
What does any of it mean for marketers? Not necessarily a whole lot at the moment beyond keeping abreast of how the technology is continuing to unfold and how it impacts us personally and professionally. But practically speaking, it’s critical to remember that the more AI can do, the more carefully teams need to define what it should be allowed to do — and always with a human involved.
What September Tells Us About AI in Marketing
September showed multiple sides of AI becoming more capable. It’s increasingly helping customers shop, can carry on conversations with advertisers and produce creative work at scale. Those are meaningful opportunities for marketers, and they’re moving beyond demos into tools and workflows people are actually using.
But greater capability also means the software has more authority to make decisions and take actions. That makes the human role increasingly less about completing individual tasks and more about deciding what should happen, setting boundaries and judging whether the output — or action — makes sense or is necessary at all.
If September offered one particularly useful reminder, it’s that automation and autonomy aren’t quite the same thing. And as AI gets better at the latter, marketers will need to think much more carefully about where the human stays in the loop.
Note: This article was originally published on contentmarketing.ai.
