HAPPY FRIDAY,
September already! Was it just me or did that seem short? Let's did into this week's chaos starting with OpenAI managing to both release what their president called "a new era of AGI" and have their CEO on record calling AGI an "irrelevant marketing term," all within a few days of each other. Impressive, in its way.
In This Week's AI Fridays:
- OpenAI declares AGI while Sam Altman calls the term meaningless
- Nvidia buys Hugging Face and becomes the most important company in AI infrastructure
- Product pages are beating Reddit in AI citations by a wide margin, and the SEO playbook needs updating
- What high-citation brands do differently in AI search
- Plus: AI food imagery, the sameness problem, and more in quick hits
OpenAI Hails 'New Era of AGI' With Astra, While Altman Calls AGI Irrelevant
GPT-6 Astra is out, and OpenAI president Greg Brockman is calling it the dawn of a new era of artificial general intelligence. Sam Altman, the CEO, said earlier this week that AGI is a marketing term that does not mean much. What you make of the model itself almost does not matter at this point, the internal messaging is its own story.
The Highlights:
- GPT-6 Astra launched September 3, with Brockman calling it a milestone into "a new era of AGI"
- Altman had described AGI as an "irrelevant marketing term" just days before the launch
- The rollout itself was chaotic enough that Altman issued a public apology within hours, paying subscribers were locked out
- OpenAI is now in the position of having to define a term its own CEO just dismissed, in real time, to the press
The Takeaway:
The contradiction is not a gaffe; it is what happens when a company has outgrown its ability to control its own narrative. Whether Astra is AGI depends entirely on who at OpenAI you happened to ask that day.
Nvidia Acquires Hugging Face in a Significant Infrastructure Power Move
While we've been watching the model wars between OpenAI, Anthropic, and Google, Nvidia went and bought the place where most of the open AI development lives. Hugging Face is where researchers share models, datasets, and tools; it is the connective tissue of the open AI ecosystem, which makes this acquisition considerably more interesting than another chip announcement. Nvidia now has a meaningful stake in the layer of AI that sits between the frontier labs and everyone trying to build something with what those labs produce.
The Highlights:
- Nvidia acquires Hugging Face, the largest open-source AI model and dataset repository in the world
- Hugging Face hosts over 500,000 models and is used by most serious AI development teams outside the frontier labs
- The deal shifts Nvidia's position from hardware supplier to infrastructure owner, a meaningful distinction
- Jack Clark at Import AI flagged concerns about what this means for open AI development
The Takeaway:
The model makers get the headlines, but whoever owns the infrastructure tends to win the long game. Nvidia just made a credible argument that they understand this better than most.
AI in Marketing
Product Pages Get 24% Of AI Citations, Reddit & YouTube 4%
The conventional wisdom for the past couple of years has been that Reddit owns AI search, that the forums and communities with authentic user voices are what AI models cite when someone asks a buying question. Ten Speed's citation study suggests this is considerably overstated, at least for B2B buyers. Product pages account for 24% of AI citations in that context; Reddit and YouTube together get 4%. Which means a lot of people have been optimizing for Reddit while their own product pages sat there doing the actual work. The implications for content strategy are fairly direct: if you want to show up in AI-generated answers, your product content needs to be written for how AI reads it, not just how humans skim it.
What High-Citation Brands Do Differently in AI Search: The 2026 AEO Playbook
Answer engine optimization has been a phrase kicking around for a while, but this HubSpot piece gets into the specifics of what separates brands that get cited from brands that do not. The short version: structured, direct, authoritative content written around real questions tends to get picked up; content written around keyword density and thin coverage does not. If your SEO strategy is still built on the logic of ranking pages rather than answering questions clearly, the gap between where you are and where AI search wants you to be is probably wider than you think. Worth reading alongside the Ten Speed citation data above, as the two pieces cover the same terrain from different angles.
The Sameness Problem Behind Those Unappetizing AI-Generated Menus
Restaurants started using generative AI for food imagery because it was fast and cheap, which are two reasons that tend to override other considerations. The results have been recognizable enough that customers are picking up on them, not because they can explain what a diffusion model is, but because the food looks wrong in a way they cannot quite name. Hyperglossed surfaces, anatomically questionable shrimp, pasta that seems to be enjoying itself too much. The broader point here is not about food photography. It is about what happens when any brand optimizes for production speed without anyone in the room asking whether the output is actually doing what they need it to do. The restaurant industry just happens to be the most visible example right now because bad food photography is immediately obvious in a way that bad AI copy is not.
People Aren't Leaving Google For AI. They're Using Both
The panic about Google dying has been running for a while now, and the data keeps refusing to cooperate with the narrative. Users are not abandoning search; they are using AI tools alongside it, often for different kinds of questions. What is changing is how they use search when they do use it: fewer queries, fewer clicks, less time on pages. For marketers whose organic traffic strategies depend on volume at the top of the funnel, this is the more uncomfortable truth. The channel is not disappearing; it is just becoming less generous. Fewer people arriving at your site from search, even if search itself stays busy, is a meaningful shift in how the top of the funnel behaves.
via Search Engine Journal
Quick Hits
- OpenAI launched GPT-6 Astra as a generational leap and then Altman spent the morning apologizing for locking paying users out, a bold two-act structure. Read more
- Jack Clark's Import AI takes a worried look at the Hugging Face acquisition and what it means for open-source AI development, worth reading if you want a perspective that is not a press release. Read more
- Google released Gemini 3.8 with a focus on long-horizon reasoning and a new cyber performance program, signals that frontier AI competition is moving well past the chatbot phase. Read more
- The Verge goes deep on why AI food imagery looks the way it does, donut shrimp, luminous noodles, produce that has clearly never met a vegetable, and the backlash from actual customers is real. Read more
- Product pages account for 24% of AI citations in B2B search; Reddit and YouTube together get 4%, which is a useful number to have ready the next time someone asks why you are spending budget on product content. Read more
- HubSpot's AEO playbook lays out what high-citation brands actually do differently in AI search, more practical than most pieces on this topic, fewer frameworks, more specifics. Read more
This article was obviously generated with AI but curated by a human, don't be weird about it.
Compiled by
Pete Bishop
Chief Innovation Officer, ZGM Modern Marketing Partners
Pete Bishop has spent the last two decades helping brands adopt new technology without losing the plot. He hosts Artificial Breakdown, a podcast and weekly newsletter that translates AI news into practical marketing decisions, and scans 50+ sources each week to write this issue.
Get this in your inbox
One email a week with the AI and marketing news worth your time.