The week of February 7–14, 2026, delivered one of the most consequential stretches in recent advertising technology history. OpenAI formally launched advertising inside ChatGPT. Google pushed deeper into AI-powered commerce with new shopping ad units in AI Mode. A Discover-specific algorithm update landed on publishers like a cold wave. And Spotify revealed that its senior engineers have not manually written code since December, raising hard questions about what software companies will look like in the months ahead.

None of these developments happened in isolation. Together, they signal an accelerating convergence between artificial intelligence platforms and the advertising infrastructure that funds them—a convergence that is redrawing how brands reach consumers and how publishers sustain their businesses.

0:00
/0:08

ChatGPT now shows ads, and not everyone stayed to watch

On February 9, OpenAI announced it had begun testing ads in ChatGPT for logged-in adult users on the free and Go tiers in the United States. The ads appear at the bottom of ChatGPT responses, matched to conversation topics, past chats, and previous ad interactions. Plus, Pro, Business, Enterprise, and Education subscribers do not see them. The Go subscription tier, priced at $8 per month, had expanded globally in mid-January before arriving in the U.S. alongside the ad test.

The mechanics are straightforward on paper. Advertisers pay a $60 CPM—a rate comparable to premium streaming and NFL inventory, as media buyers told Digiday, and well above Meta's typical CPMs, which often come in under $20. Early participants include Target, Ford, Mrs. Myers, and Adobe, with major holding companies WPP Media, Omnicom, and Dentsu among confirmed partners. The minimum commitment stands at $200,000.

But the rollout has been anything but smooth. Only a subset of eligible users are currently seeing ads. AdExchanger reported on February 13 that when asked where ads would appear, ChatGPT itself gave incorrect answers—an OpenAI spokesperson confirmed the chatbot's explanation was entirely untrue. Within the next several weeks, ads will expand to all qualifying users, including logged-out ones.

The fallout extended beyond technical bugs. Former OpenAI researcher Zoë Hitzig resigned on February 11, publishing a New York Times commentary titled "OpenAI Is Making the Mistakes Facebook Made. I Quit." Hitzig, who spent two years working on AI models and safety guidelines, warned that economic pressures from an upcoming IPO—planned for the fourth quarter of 2026—could create strong incentives to override the company's own privacy commitments. OpenAI CEO Sam Altman had previously called advertising a "last resort" and "dystopian," language that has since softened considerably.

Then came the question of agencies. Search Engine Roundtable reported on February 12 that Asad Awan, OpenAI's head of monetization, suggested in a company interview that advertisers could eventually prompt ChatGPT to create ads and campaigns directly—without hiring agencies or performance marketers. The reaction from the agency community was immediate and skeptical.

OpenAI has been tightly scripting how partners talk about the test, positioning it as a learning exercise rather than a formal product launch. Partners cannot discuss pricing, performance data, or internal results, and must route sensitive questions back to OpenAI. Materials require pre-approval. As eMarketer's principal analyst Nate Elliott pointed out, OpenAI is terrified of scaring away users.

The Super Bowl became an AI battlefield

The advertising launch arrived with dramatic timing. Super Bowl 60 on February 9 had already become a proxy war between AI companies, with AI-related spots accounting for 23% of all Super Bowl ads, a concentration that would have been unimaginable two years ago. Thirty-second spots cost up to $10 million. NBC sold out five months early—the fastest sellout in Super Bowl history.

Anthropic made its Super Bowl debut with "A Time and a Place," a campaign created by agency Mother. The four spots dramatized personal moments—asking AI about health, relationships, work—then interrupted them with jarring sponsored responses from fictional ad-supported chatbots. The tagline: "Ads are coming to AI. But not to Claude."

OpenAI fired back before kickoff. In a 420-word post on X, Altman called the ads "deceptive" and dismissed Anthropic as serving "an expensive product to rich people." OpenAI aired its own 60-second spot during the first quarter, positioning its Codex coding assistant as the latest chapter in a long history of human innovation through code. Since August, Codex has grown 20x, with a million people using it last month to build apps and websites.

ADWEEK's analysis noted a fundamental tension: the marketing industry preaches digital, fragmented, programmatic futures, yet advertisers paid up to $20 million for minute-long spots during a linear broadcast consumed in real time with no skip button. Nike bought 90 seconds. Anheuser-Busch took 2.5 minutes across three brands.

Beyond the spectacle, new data revealed something important about how AI systems remember advertising. ADWEEK partnered with Emberos to create the first real-time AI Influence Index, testing how ChatGPT, Claude, Gemini, Perplexity, and Grok responded to identical post-Super Bowl questions. Rankings varied by 20% to 30% across models. ChatGPT favored ads that could be explained clearly in conversation. Claude prioritized emotional resonance and purpose. Perplexity leaned on citations and press coverage. Gemini focused on mainstream visibility and tech relevance. Neither OpenAI nor Anthropic cracked the top 20 in AI influence despite spending millions—Xfinity, Bud Light, Squarespace, Ramp, and Dove dominated instead.

The creator economy also found its footing at the Big Game. Digiday reported on February 13 that brands turned in-person activations into a new playground for creators. The NFL moved the Pro Bowl closer to the Super Bowl for the first time, extending the week and increasing deal opportunities. Jason Tartick, founder and CEO of Rewired Talent Management, noted that creator activations and IRL experiences have deeply evolved.

Google pushes AI Mode shopping and agentic commerce forward

While OpenAI tested its first ads, Google was expanding a much more mature advertising ecosystem into AI territory. On February 11, Google Ads announced new shopping ad formats for AI Mode, giving retailers a way to appear naturally within conversational search moments. The same announcement revealed that Veo 3 video generation support arrived in Google Ads Asset Studio. In 2025, Gemini-generated asset creation grew 3x, and in Q4 alone, the system produced nearly 70 million creative assets in AI Max and Performance Max.

Perhaps more consequentially, Google's Universal Commerce Protocol-powered checkout went live in AI Mode. U.S. shoppers can now purchase items from Etsy and Wayfair directly inside AI Mode in Search and the Gemini app, with Shopify, Target, and Walmart joining soon. The protocol, first unveiled at NRF 2026 in January, carries endorsements from more than 20 industry leaders and is compatible with A2A agent communication standards.

Vidhya Srinivasan, who leads Google's ads and commerce division, published her third annual letter on February 11, outlining a vision where AI reshapes shopping from discovery through purchase. The letter detailed how advertisers need AI-powered targeting solutions—broad match on Search campaigns, AI Max for Search, Performance Max, and Google Shopping Ads—to show ads in AI-powered search experiences.

For publishers and advertisers tracking the intersection of AI and commerce, the landscape just became considerably more complex. Google is building checkout infrastructure inside conversational AI. OpenAI is selling ads adjacent to conversational responses. The two approaches are architecturally different—one embeds commerce inside an AI shopping experience with product feeds and merchant data, while the other places traditional display-style ads below chatbot answers—but both are competing for the same dollars.

Google's Discover update hits publishers hard

On February 5, Google released the February 2026 Discover core update, its first algorithm adjustment of the year. The update targets only English-language users in the United States and introduces three changes: prioritizing locally relevant content from domestic websites, reducing sensational content and clickbait, and surfacing deeper expertise from specialized sections within broader publications.

PPC Land reported that Google's internal testing demonstrates users find the Discover experience "more useful and worthwhile with this update." But the impact on non-U.S. publishers could be severe, with geographic filtering potentially reducing traffic from international sources. The update is rolling out over two weeks.

Industry reaction was swift and critical. PPC Land covered on February 7 how Vic Daniels, Co-founder and Executive Chairman at GRV Media Ltd, published a LinkedIn analysis on February 6 characterizing the geographic filtering as "The Invisible Border"—a form of digital discrimination that contradicts decades of internet philosophy. The concern is that Google's algorithm creates geographic echo chambers on what was designed to be an open, borderless web.

Google's update frequency throughout 2025 established patterns continuing into 2026. Three core updates deployed last year—March, June, and December—alongside continuous smaller refinements. The March update required 14 days to complete, June took 16 days, and December needed 18 days. Meanwhile, Search Engine Roundtable noted on February 10 that unconfirmed Google search ranking volatility heated up again, adding to persistent instability that plagued publishers throughout January.

Separately, Google AdSense added new triggers for vignette ads effective February 9. There are now six triggers instead of three, designed to "unlock incremental revenue by identifying additional high-value impression opportunities." Publishers have a one-month review period before the new triggers activate automatically on March 9, but opting out disables all existing triggers as well—a choice that effectively penalizes publishers who decline the expansion.

Spotify stops writing code, raises questions about AI in production

Spotify disclosed during its Q4 2025 earnings announcement on February 10 that senior engineers stopped writing code manually in December 2025, using only AI-generated code through an internal system called Honk built on Claude Code. Co-CEO Gustav Söderström acknowledged that the transformation "will be painful for many companies" but emphasized that Spotify had been preparing for AI-driven development "for at least one and a half years."

The numbers behind the shift are substantial. Spotify achieved record quarterly user growth of 38 million monthly active users, reaching 751 million total. Premium subscribers hit 290 million. Revenue reached €4.5 billion, growing 13% year-over-year on a constant currency basis. Operating income hit €701 million for the quarter. The company paid $11 billion to music rights holders during 2025.

For the first quarter of 2026, Spotify forecasted 759 million monthly active users—net additions of 8 million—and 293 million premium subscribers. Total revenue guidance of €4.5 billion represents approximately 15% year-over-year growth. Currency headwinds remain significant, with approximately 670 basis points of headwind compared to prior-year exchange rates.

A study published January 29 by Anthropic found that developers using AI assistance scored 17% lower on coding comprehension tests despite completing tasks slightly faster. Söderström dismissed concerns, suggesting that productivity gains enable Spotify to increase software output until "the limiting factor is actually the amount of change that consumers are comfortable with" rather than engineering capacity. Co-CEO Alex Norström framed 2025 as the "Year of Accelerated Execution" and 2026 as the "Year of Raising Ambition."

Criteo faces a $75 million problem

On February 11, Criteo reported Q4 2025 earnings that sent its stock dropping sharply. Revenue declined compared to the prior year, net profit fell to $46 million from $72 million, and the take rate dropped. Most critically, two large retail media clients pulled back from previous agreements, creating a $75 million shortfall in 2026 projections, mostly frontloaded to the first half of the year.

CEO Michael Komasinski told AdExchanger that without the forecast downgrade, Criteo's core business would be growing at high single digits. But investors were not reassured. Connected TV represents one of the fastest-growing online ad markets, and Criteo has relatively little business there. The company is moving into CTV and social media, with partners including Meta adding video formats to its Criteo integration in 2026. Currently, only 37% of its Commerce Go! AI-based dynamic allocation product customers extend to social media.

The earnings serve as a cautionary tale for the retail media sector. The channel has been one of the advertising industry's strongest growth stories, but client concentration risk remains real. When large retailers pull back, the impact cascades through dependent ad tech companies.

The IAB pushes measurement reform and publisher protocols

The IAB was active across multiple fronts this week. On February 2, the organization unveiled Project Eidos, uniting its numerous measurement initiatives under one program. The name comes from the Greek word meaning "to see"—an apt metaphor for an industry that has relied on what Angelina Eng, VP of the IAB Measurement Center, called "digital duct tape" to fix its campaign measurement problems for years.

On February 10, AdExchanger published a detailed Q&A with Anthony Katsur, CEO of IAB Tech Lab, who laid out what he called the "Publisher Protocol for 2026." Publishers at the IAB Annual Leadership Meeting had been openly frustrated about too much future-casting and not enough focus on how to grow real revenue this year. Katsur pointed to sell-side decisioning as the major shift—after more than a decade where most decisioning power sat on the buy side, smarter sell-side partners are using AI to curate audiences, content, and supply closer to the user.

On agentic AI workflows, Katsur was cautiously optimistic. Very early tests are underway, but nothing scales until 2027 or 2028. He flagged a practical concern: buyer agents and seller agents need to speak a common language, or hallucinations could turn impression goals into budgets—an outcome he described as "a very bad phone call."

Amazon opens its MCP server, GEO startups chase AI citations

Amazon Ads opened a beta program on February 2 for a new Model Context Protocol server connecting existing software systems and their APIs to agentic AI systems. The MCP acts as a translation layer between advertising software and AI agents, allowing advertisers to bring their own LLM or agentic solution while using Amazon's infrastructure. Paula Despins, Amazon Ads' VP of ads measurement, said the server will be natively integrated with Amazon campaign management tools.

The MCP announcement arrived alongside a separate development in generative engine optimization. On February 11, AdExchanger reported that GEO startup Evertune launched a feature letting advertisers run programmatic ad campaigns directly on the sites and pages most often cited by AI chatbots. Evertune has partnerships with The Trade Desk on the buy side and Index Exchange on the sell side—notably, much of Evertune's team previously worked at The Trade Desk.

The convergence of GEO and programmatic targeting represents a meaningful evolution. Rather than merely optimizing content so that AI models cite a brand favorably, Evertune is placing traditional programmatic ads on the destinations AI chatbots send users to. It is a pragmatic acknowledgment that influencing AI models directly remains difficult and unpredictable, while reaching users who follow AI recommendations through established advertising channels offers measurable, immediate returns.

Lego and other brands deepen in-house programmatic teams

Digiday reported on February 4 that Lego is building an in-house programmatic media buying team, hiring five programmatic roles during January including a global programmatic media manager and associates in the United States and Denmark. Publicis One handles Lego's global media account, but the Danish toymaker reportedly spends $469 million a year on media and already manages programmatic and search spending for its direct-to-consumer business.

Lego is not alone. UK retailer Boots has built a 27-person-strong team handling programmatic, CTV, affiliate PPC, and audio, and was hiring a digital media manager to bring paid social in-house as the final piece. Paper manufacturer Georgia-Pacific began in-housing in 2019 and now handles everything except TV.

The broader trend suggests that programmatic—once considered the final frontier of in-housing—is increasingly within reach for large advertisers. Digiday's follow-up analysis on February 9 noted that the promise of agentic media buying is adding fuel, with CMOs quietly watching efforts from DSPs like Yahoo to develop AI systems that automate spending decisions.

OOH gets a programmatic upgrade

On February 10, the Out of Home Advertising Association of America introduced an updated version of its OpenOOH venue taxonomy, first launched in 2020. The update addresses a longstanding problem: generic categories like "retail" or "transit" do not give programmatic buyers enough information about what they are bidding on. When publishers and SSPs lump together media of varying quality, it undermines transparency and makes buyers think twice about digital out-of-home. The OAAA is folding its taxonomy working group into its broader OAAA Taxonomy Committee, which will develop and maintain the standard going forward.

iSpot launches SAGE for agentic measurement

Ad measurement firm iSpot introduced SAGE on February 10, an agentic AI platform with a conversational interface that media buyers can use to generate campaign planning ideas. SAGE includes separate agents for creative planning and performance analysis, along with competitive intelligence features. iSpot joins a growing list of ad tech companies launching conversational AI dashboards that let clients query campaign data directly—part of a wider effort to make media planning and measurement accessible to nontechnical users.

Publishers see Q4 gains but face an uncertain road

Digiday's Media Briefing on February 5 explored how a strong Q4 2025 gave publishers their first significant win in a long stretch of belt-tightening. Digital advertising revenue was up for many, buoyed by a more stable economy and advertisers feeling more confident to spend. But traffic erosion, a volatile open-web programmatic market, and AI-driven shifts in content discovery mean growth has to be earned differently.

People Inc. CEO Neil Vogel singled out InStyle's social video franchise "The Intern"—launched last February, now in its seventh season—as a model, with advertisers paying between $500,000 and $700,000 to sponsor the show, which costs next to nothing to produce. That kind of diversification away from Google referral traffic dependency is becoming essential.

Meanwhile, the Winter Olympics in Milan-Cortina opened February 6, and Digiday reported that programmatic access is drawing smaller advertisers to what used to be the exclusive preserve of the industry's largest spenders. NBCU sold out its Olympics ad inventory with a month to spare, setting a new Winter Olympics ad sales record. ADWEEK noted that 70% to 75% of Super Bowl advertisers will also appear in the Winter Games.

The week that redefined how ads meet AI

The past seven days compressed what might normally take months into a single week. OpenAI crossed the Rubicon into advertising—a move that already cost it a researcher and raised uncomfortable questions about trust, privacy, and the commercialization of intimate AI interactions. Google expanded its AI commerce infrastructure in ways that make the company even more central to how people discover and buy products. A Discover algorithm change reminded publishers that their traffic exists at the discretion of a single company's ranking decisions. And Spotify demonstrated that AI-driven software development is no longer theoretical.

U.S. ad revenue growth for 2026 is projected at 6.6% excluding political ads, per Madison & Wall's forecast featured in Digiday on January 8—roughly half of 2025's 11% surge. WPP projects 7.4%, Morgan Stanley analysts 10%. The midterm elections will drive significant political spending, and the FIFA World Cup in Mexico, Canada, and the United States will provide another boost.

But the big story is structural, not cyclical. The advertising industry is reorganizing itself around AI interfaces—some that show ads, some that do not, and some that are still deciding. The choices being made this week will shape what that industry looks like for years.

Timeline

Share this article
The link has been copied!