The marketing industry is stuck in an awkward adolescence: AI is being treated like a finished product when it’s barely a reliable workflow tool. This week made that contradiction unavoidable. OpenAI is expanding its ad business across Europe while simultaneously on track to miss its own revenue forecast by 90%. Brands are automating creator networks of 300,000 people, yet marketers still won’t trust AI for CTV or influencer buys. And while platforms scramble to prove AI can scale without producing generic slop, publishers are building defensive moats against the very scrapers fueling the hype. The pattern? Infrastructure is being rebuilt in real-time, but the narrative is running six months ahead of reality.
The AI Ad Bubble Meets Arithmetic
OpenAI wants $100 billion in ad revenue by 2030. eMarketer thinks the entire chatbot ad market might hit $5.41 billion. That is not a rounding error; that is a hallucination. Adweek reports that OpenAI’s ad business is on pace to miss its own forecast by roughly 90%, which raises an uncomfortable question: are VCs and trade press writing checks with numbers that the actual market can’t cash?
The expansion playbook isn’t slowing down, of course. Digiday notes that OpenAI is now pushing ads into France, Germany, and Ireland, effectively covering every major global advertising region in less than a year. Speed is not strategy. Chasing geographic coverage while unit economics remain unproven is a classic land-grab mentality, and land grabs work until the capital dries up.
Meanwhile, the people actually spending media budgets are telling a more cautious story. Digiday research shows advertisers are comfortable using AI for social and retail media, but they’re dragging their feet on influencer and CTV applications. That hesitation is rational. Social and retail media have closed-loop attribution. CTV and influencer marketing don’t. Throwing opaque AI into already murky measurement environments isn’t innovation; it’s compounding risk.
CTV Is Growing Up—and It’s Messy
If you want proof that a channel has matured, look for the standardization fights. The IAB is now trying to force ad tech to agree on what words like “CTV” actually mean, because five people in the industry currently give five different answers. This is not bureaucratic trivia. When buyers and sellers lack shared definitions, fraud hides in the gaps and measurement vendors sell incompatible metrics. The IAB’s push for digital video taxonomy is long overdue, and its success or failure will determine whether CTV keeps growing or stalls under its own complexity.
Netflix, for its part, seems to be eyeing a future that looks suspiciously like the past. AdExchanger reports that Netflix is considering live TV channels, a move that would blur the line between streaming and linear in ways that matter for ad inventory. If Netflix goes linear-ish, it’s not just a product decision—it’s an admission that scheduled programming can still deliver audiences at scale, and that the purely on-demand model may have limits for ad-supported revenue.
Even out-of-home is getting into the act. LinkNYC kiosks started airing World Cup games, TV ads included, turning street furniture into broadcast endpoints. It’s a neat stunt, but it also signals that video ad inventory is increasingly platform-agnostic. The buy side doesn’t care if an ad runs on a smart TV, a phone, or a sidewalk screen—as long as the impression is measurable and the audience is real.
Publishers Draw the Line on Scraping
While AI platforms beg for ad dollars, they’re simultaneously hoovering up publisher content to train the models that power those ads. Publishers have noticed. Digiday explains SPUR’s Content Telemetry Framework, a publisher-run initiative designed to turn opaque AI scraping into transparent, usage-based licensing. It’s a direct response to the status quo, where AI companies treat the open web as a free training buffet.
The scraping economy is more layered than most coverage suggests. Digiday’s media briefing breaks down the difference between declared “good bots,” mixed-use crawlers, and gray-market scrapers that ignore robots.txt entirely. Cloudflare’s new AI settings are helping compliant crawlers behave better, but the real damage comes from the gray economy that never bothered to ask permission. Publishers can deploy telemetry frameworks all day; if enforcement stays toothless, the frameworks become compliance theater.
Commerce Media and the Infrastructure Layer
Not every platform bet is paying off. Intuit, which launched SMB MediaLabs in 2023 to retarget QuickBooks customers across the web, is shutting that network down and reshaping its media strategy entirely. The lesson: slapping an ad network on top of first-party data sounds compelling in a pitch deck, but execution is harder. Small-business data is valuable, but activating it at scale without alienating the user base requires more than ad tech plumbing.
On the programmatic front, AdExchanger covers how containerization is moving AI decisioning models into SSP environments, fundamentally altering the supply chain. For years, algorithmic buying lived inside DSPs with limited inventory visibility. Containerization pushes that intelligence closer to the supply source, which could reduce latency and improve yield—but also concentrates power in the hands of the SSPs hosting the containers. Expect this to become a consolidation battleground.
The Bottom Line
This week’s stories don’t paint a picture of an industry being transformed by AI. They paint a picture of an industry retrofitting its plumbing while vendors shout about revolution. The real work happening right now is standardization, automation, and content protection—not generative creative breakthroughs. The marketers who win the next 12 months will be the ones who treat AI as infrastructure rather than magic, who demand measurable CTV outcomes before the taxonomy is perfect, and who recognize that publisher consent is going to become a line item on every media plan. The bubble hasn’t popped yet, but the hissing sound is getting louder.