OpenAI confirmed plans to test advertisements within ChatGPT during the week of January 20, marking a shift from CEO Sam Altman's previous characterization of ads as a "last resort" for the platform's business model. The announcement, made on a Friday before a holiday weekend, left advertisers with limited information about implementation details, pricing structures, or timeline specifics.

Michelle Merklin, VP of search at independent performance agency Tinuiti, told AdExchanger the company has "more questions than answers at this point." OpenAI representatives informed the agency there is currently no wait list for advertisers, but the company has not provided direct access to its newly formed advertising team. Merklin noted OpenAI has been "very noncommittal about everything" regarding the ads rollout.

Within days, The Information reported OpenAI established a baseline CPM of approximately $60 for ChatGPT advertising placements. The pricing rivals inventory costs for the Las Vegas Sphere and NFL game broadcasts. Meta's advertising CPMs typically range from $10 to $20, while YouTube campaigns often reach the high end around $60. OpenAI's pricing structure stands out because the platform lacks attribution systems or pixel networks to track conversions that advertisers use to measure campaign effectiveness.

The cost-per-impression model announced by OpenAI charges advertisers based on impressions or views rather than clicks. This payment structure aims to push advertisers toward meaningful organic engagement over direct link performance. However, early advertiser requirements suggested initial campaign spending would remain below $1 million during the launch phase.

OpenAI CFO Sarah Friar defended the timing during a Davos appearance on January 22, stating "early is a weird word" when discussing advertising models. Friar emphasized that successful ad models require scale, adding "when you have 800 million weekly active users, you're really far beyond many of the companies who started in that model." Demis Hassabis, CEO of Alphabet-owned DeepMind, questioned whether OpenAI moved too quickly, telling journalist Alex Heath it was "interesting they've gone for that so early" and suggesting "maybe they feel they need to make more revenue."

The contextual targeting capabilities distinguish ChatGPT from traditional advertising platforms. Yang Han, CTO of DSP StackAdapt, explained the platform can target users at precise moments based on conversation content. ChatGPT maintains memory of user interactions, potentially knowing when someone last ran a race and related queries they have made. This enables product recommendations tailored to individual consumer needs beyond basic demographic targeting.

The advertising implementation faces technical challenges. Becoming fully contextually relevant requires a broad advertiser base comparable to Meta's millions of active advertisers. Until ChatGPT achieves similar scale, the platform risks showing mediocre advertisements that could alienate users. Han characterized this as "a bit of a chicken-and-egg problem" where limited advertiser participation could deter new advertisers from joining.

Tyler Holoubek, assistant director of investment at independent agency KSM Media, observed varying client reactions to ChatGPT advertising opportunities. Some clients approached testing with enthusiasm, seeking early learnings. Others in regulated industries including pharmaceuticals and financial services showed greater hesitation about the untested channel.

The advertising development arrives as OpenAI explores new products distinct from traditional social formats. A company spokesperson clarified that separate AI systems will evaluate whether conversations have commercial intent before surfacing relevant advertisements in ChatGPT responses. OpenAI maintains that ads will likely appear only after users show clear interest or purchase intent, ensuring relevance and maintaining user trust.

Meta completes Threads advertising expansion

Meta announced on January 21 the global expansion of advertising on Threads to all users, completing a testing program that began in January 2025. The rollout, starting the week of January 26, extends advertising availability to all markets where Threads operates, reaching the platform's more than 400 million monthly active users.

The expansion represents the final phase of Meta's monetization strategy for the text-based social platform. Meta launched Threads in July 2023 without advertising capabilities, prioritizing user growth before introducing commercial elements. The company initiated testing with limited advertisers on January 24, 2025, when the platform had accumulated 300 million monthly active users. Testing expanded to all eligible advertisers on April 23, 2025, though delivery remained restricted to select markets.

Advertisements on Threads utilize Meta's AI-powered advertising system, offering the same personalization level users experience on Facebook and Instagram. Available formats include image ads, video ads, carousel ads, Advantage+ catalog ads, and app ads. Technical specifications support 4:5 aspect ratio rendering, with assets taller than 4:5 cropped and vertically centered for format consistency.

The Threads feed placement activates by default for new campaigns using Advantage+ or Manual Placements in Meta Ads Manager. Advertisers seeking to exclude Threads from their campaigns must manually opt out through placement settings. Meta emphasized that ad delivery will initially remain low as the company reaches global user availability through a gradual rollout during the coming months.

Technical capabilities evolved substantially throughout 2025. The platform initially supported only single image and video advertisements when testing began in January. Meta introduced carousel ad support on October 6, 2025, followed by Advantage+ catalog ads on October 28. The catalog functionality currently supports images and image carousels, while slideshow, video, and entry cards remain unsupported. Aspect ratio specifications expanded on October 28 to accommodate the 4:5 rendering format.

The timing for the expansion demonstrates Meta's measured approach to platform monetization. Meta stated during the initial testing phase that it would monitor the limited test closely before broader implementation. The company emphasized that advertisements help people discover businesses while enabling advertisers to extend existing Meta campaigns to the conversation platform without bespoke creative or additional resourcing requirements.

Microsoft Prebid caching termination threatens publishers

Microsoft announced it will stop caching Prebid video files on April 30, shortly after the company shuts down the Xandr DSP. The change affects publishers who rely on Prebid and Google Ad Manager for video advertisement delivery, potentially creating immediate and costly disruptions.

When advertisers win video ad impressions in Prebid auctions and publishers use Google Ad Manager, Google runs a secondary auction comparing the Prebid bid against bids from its own demand stack. If the Prebid bid wins, GAM calls for the video creative to be served. During the brief moment before GAM makes the final auction decision, an intermediary must host the video ad file from Prebid. Microsoft Advertising has provided this storage service since acquiring Xandr in 2021, following AppNexus and AT&T's Xandr.

Patrick McCann, SVP of research at Raptive and Prebid board member, estimated monthly caching costs between $50,000 and $250,000. These figures reflect current usage levels, but costs could reach seven figures monthly within a few years as Prebid adoption grows and video ad products develop. Around four in five US sites that allow ad serving use GAM, meaning if 60 percent of Prebid video ad coverage in that supply chain is affected, a vast number of actual ads could fail to fire on pages come February.

Microsoft already extended the deadline from January 31 to April 30, providing additional transition time. However, McCann noted that advertisers and publishers often wait for absolute deadlines, potentially leaving many unprepared. Publishers who fail to implement alternative solutions will see video ads result in VAST errors, meaning no video will appear.

Technical workarounds exist through multiple SSPs and video ad networks. Magnite, PubMatic, Mediavine, and JW Player already offer Prebid video caching as a paid option for clients. Publishers must evaluate these alternatives and implement solutions before the April 30 deadline to maintain video advertisement delivery.

The caching termination aligns with Microsoft's broader strategic shifts in advertising technology. The company announced in May 2025 it would discontinue Microsoft Invest (formerly Xandr) with an effective date of February 28, 2026. Corporate Vice President Kya Sainsbury-Carter cited incompatibility between traditional demand-side platform models and Microsoft's vision for conversational, personalized, and agentic advertising futures.

Prebid launches publisher-side AI agent framework

Prebid announced on January 29 it is taking ownership of code developed using Ad Context Protocol that will power publisher-side AI agents. The group that established open-source standards for header bidding software now provides publishers with tools to build autonomous AI agents for selling ad inventory.

The offering, named the Prebid Sales Agent, is available to all publishers regardless of Prebid membership status. Publishers can download the software from GitHub and provide feedback on its development. Emerging frameworks like AdCP are establishing groundwork for AI-to-AI transactions, but expecting publishers to build and maintain their own autonomous agents represents a stretch without standardized tools.

The development arrives as the advertising industry races toward a future where AI agents negotiate programmatic deals autonomously. Prebid's involvement ensures publishers have access to standardized tools rather than requiring individual development efforts for agentic capabilities. The open-source approach mirrors Prebid's successful header bidding standardization, which reduced publisher reliance on proprietary solutions.

AdCP framework faces implementation skepticism

Dr. Aaron Andalman, Chief Science Officer at Cognitiv, published analysis on January 26 questioning whether Ad Context Protocol delivers meaningful improvements beyond workflow standardization. Since its launch several months ago, AdCP has been positioned as a foundational step toward agentic advertising where AI systems plan, execute, and optimize media buys without human intervention.

AdCP builds on Model Context Protocol, an open standard introduced by Anthropic and broadly adopted across the industry. MCP addresses limitations of large language models in taking reliable actions inside software. Andalman argues that advertising's agentic future represents a compelling vision relying on capabilities that do not exist yet. AdCP may eventually make buying easier, but it will not make buying better by itself.

The protocol standardizes how AI systems interact with advertising platforms and tools. However, standardized workflows do not automatically translate to improved media outcomes. Advertisers feeling pressure to implement AdCP should evaluate whether the framework addresses actual business needs or represents premature adoption of incomplete technology.

Meta's Andromeda update accelerates creative exhaustion

Meta's latest Andromeda update, the fastest and most advanced iteration of Meta's ad-retrieval system, has changed the rhythm of Meta Ads delivery. Monica Shukla, VP Biddable COE at an unnamed agency, wrote on January 28 that delivery cycles are faster, creative gets picked up and exhausted with remarkable speed, and optimization feels more dynamic than ever.

The increased velocity has introduced confusion among advertisers. Many assume that because Andromeda adapts quickly, they can supply massive volumes of creative, rely on automation, and let the system handle optimization. Others expect the algorithm will surface insights, fix signal issues, or compensate for budget constraints. Shukla emphasized that Andromeda is powerful but far from omniscient, relying on signal strength, strategic clarity, and creative asset quality.

One persistent misunderstanding stems from the belief that uploading 20 or more creatives per ad set is universally beneficial. When creative supply dramatically outweighs available budget, learning slows, delivery fragments, and the system struggles to distinguish signal from noise. Many smaller or emerging brands fall into this trap, assuming they must keep pace with global advertisers' creative velocity despite insufficient budgets to support meaningful learning across numerous assets.

A more effective approach aligns creative volume with budget reality and campaign goals. Upper-funnel campaigns can support a broader range of concepts, but only at budgets that allow each asset to gain traction. Brands achieving the greatest lift from Andromeda have adjusted not by outsourcing judgment to the algorithm but by refining their creative strategy. Match creative volume to budget and campaign goals rather than defaulting to maximum creative uploads.

Federal immigration enforcement seeks ad tech surveillance

US Immigration and Customs Enforcement published a request on January 23 at 1:00 PM Eastern Standard Time seeking information from advertising technology companies about location data and surveillance capabilities. The request for information, posted on SAM.gov, asks companies to demonstrate how their platforms track population movements through mobile applications, advertising platforms, and data broker networks.

ICE seeks operational capabilities for location tracking, geofencing, and behavioral analysis tools used in commercial advertising. The agency specified it works with "increasing volumes of criminal, civil, and regulatory, administrative documentation from numerous internal and external sources." Companies must respond by February 2, 2026 at 10:00 AM Eastern Standard Time, with the request becoming inactive February 3.

The filing states it represents market research rather than an immediate procurement, emphasizing "this is not a solicitation announcement" and "does not commit the Government to contract for any supply or service." However, the request signals federal law enforcement examining commercial advertising technology for investigative deployment. Companies submit responses through a Microsoft Form submission portal with written documentation limited to 10 pages.

ICE indicated it will select several respondents to present live demonstrations of operational capabilities, platforms, and data services. The filing does not specify demonstration locations. The commercial surveillance tools ICE seeks operate globally through mobile applications, advertising platforms, and data broker networks assembling comprehensive population monitoring capabilities.

The timing coincided with November 2024 investigative reporting exposing how data brokers sell location coordinates tracking US military and intelligence personnel movements, demonstrating commercial surveillance capabilities relevant to government interests. California approved regulations for a data broker deletion platform in November 2025, taking effect January 2026, while Texas filed a lawsuit against Hisense over smart TV surveillance affecting 1.27 million residents in November 2024.

The request raises questions about advertising technology's role in government surveillance programs. Location data collected for advertising targeting purposes can be repurposed for tracking specific individuals or groups. The advertising industry has faced increasing scrutiny over data collection practices and potential misuse of consumer information.

Timeline

  • January 21, 2026: Meta announces global expansion of advertising on Threads to all users starting the week of January 26, completing yearlong testing program
  • January 22, 2026: OpenAI confirms plans to test advertisements within ChatGPT, revealing limited details about implementation timeline or pricing
  • January 22, 2026: OpenAI CFO Sarah Friar defends advertising timing at Davos, citing 800 million weekly active users as justification for scale
  • January 23, 2026: ICE publishes request for information on SAM.gov seeking advertising technology location tracking and surveillance capabilities
  • January 26, 2026: Microsoft confirms it will stop caching Prebid video files on April 30, threatening publisher video ad delivery
  • January 26, 2026: Meta begins rolling out advertisements to all Threads users globally across all markets
  • January 26, 2026: Dr. Aaron Andalman publishes analysis questioning Ad Context Protocol's ability to improve media outcomes
  • January 27, 2026: The Information reports OpenAI set baseline CPM of approximately $60 for ChatGPT advertising placements
  • January 28, 2026: Monica Shukla writes about Meta's Andromeda update accelerating creative exhaustion and delivery cycles
  • January 29, 2026: Prebid announces it is taking ownership of Ad Context Protocol code to power publisher-side AI agents
  • January 29, 2026: AdExchanger reports on advertiser hesitation toward ChatGPT's untested channel lacking attribution systems
  • February 2, 2026: Response deadline for companies to submit information to ICE about advertising technology surveillance capabilities
Share this article
The link has been copied!