Industry expert questions AdCP media buying protocol for ad automation

Ad tech veteran Ari Paparo published a detailed analysis of the Ad Context Protocol on November 3, 2025, expressing support for certain automation capabilities while raising significant concerns about media buying applications. The protocol, launched on October 15, 2025, aims to establish technical standards for AI agents operating across advertising platforms.

"AdCP seeks, in its own words, 'to create an open source specification for the use of MCP for advertising use cases,'" according to Paparo. The protocol attempts to establish infrastructure allowing AI agents to execute advertising tasks across fragmented platforms. Scope3, Yahoo, PubMatic, The Weather Company, and numerous other vendors back the initiative, which is closely associated with Brian O'Kelley and Scope3.

Model Context Protocol provides the technical foundation. MCP enables AI models to connect securely with external data, tools, and APIs in real time. According to Paparo's description, MCP differs from REST APIs through richer context handling, permissions, and semantic awareness specifically designed for AI-to-system communication rather than general-purpose software interfaces.

The creative specification represents what Paparo characterizes as a "brilliant" application of the protocol. The specification provides detailed structure for creative asset handling, addressing complexity that advertisers encounter regularly. According to technical documentation Paparo cited, the protocol defines parameters including asset type, dimensions, acceptable formats, maximum file sizes, and animation permissions. A sample specification shows requirements for a 300x250 image accepting JPG, PNG, or GIF formats with a 200KB maximum file size and animation permitted.

"If I had to really, really, really dumb this down, I would say something like, 'You know how AI hallucinates all the time? Imagine if you prevented that by telling it exactly what you wanted instead,'" Paparo explained. The creative protocol addresses time-consuming, error-prone workflows currently handled by humans or unreliable point-to-point integrations. Both platforms involved in creative exchanges benefit from standardization, as each party seeks efficient delivery of appropriate creative assets.

The signals protocol addresses data broker and targeting parameter discovery. Advertisers scrolling through thousands of poorly organized segments in demand-side platforms experience significant friction that this protocol attempts to reduce. The specification theoretically enables agents to query available segments across multiple providers supporting the protocol, then activate selected segments in preferred platforms.

However, Paparo expressed reservations about the signals protocol's business viability. "I worry that the business reality of the data world is that the parties often don't have incentives to trust one another, and AI is not going to solve that problem. It might make it worse, actually," he wrote. Data brokers describing segments through MCP servers face incentives to present segments as appealing and powerful as possible, potentially exaggerating targeting capabilities.

The media-buy protocol generated Paparo's strongest criticism. According to his analysis, the protocol intends to enable buyers and sellers to transact through agents, but faces substantial obstacles that automation cannot overcome. Historical failures of programmatic direct in various forms inform this skepticism. Paparo attempted building similar functionality at Google in 2010, encountering hostility and indifference from both buyers and sellers.

"Buyers don't want to be 'price takers,' and sellers don't want to reveal their rate cards. Buyers don't want to scale downwards to smaller sites, and publishers don't have the scale to offer performance or data," Paparo wrote. He questioned whether agents solve these fundamental business problems.

The size objection appears frequently in industry discussions. Buyers resist overhead for small media buys. Paparo acknowledged that AI agents could theoretically overcome this barrier, but only if many small publishers implement the agent framework—a significant implementation hurdle. Supply-side platforms or sales representatives might implement protocols on behalf of publishers, but this intermediation raises questions about why these entities don't already offer such capabilities. Independent self-serve booking systems like DanAds and BuySellAds operate as small businesses despite existing opportunities.

Pricing transparency represents another obstacle. Publishers typically resist exposing inventory packages and pricing to buyers. "I don't see publishers putting sensitive pricing data into an MCP server, even if only customers are allowed to query it," Paparo stated. This impediment affects larger publishers participating with larger advertisers, as publishers only expose rate card pricing to smaller advertisers rather than negotiated rates for major customers who would benefit most from automation.

Performance-based buying examples highlighted by AdCP supporters face similar constraints. According to Paparo's assessment, most sellers lack sufficient valuable data to deliver performance to buyers at scale. Vendors monetizing publisher first-party data typically operate through ad network models or curation to achieve scale across publishers. "So again, the use of agents ends up being more appropriate for reinforcing the role of intermediaries, rather than enabling a breakthrough to empower buyers and sellers," he argued.

The fragmentation challenge confronting open web publishers has always limited their collective power. Part of the optimism surrounding AdCP stems from beliefs that agents and automation could reduce fragmentation pain. Paparo dismissed this view based on factors he outlined. Long-tail publishers lack technical sophistication and data scale necessary to differentiate themselves to advertisers without aggregation.

The protocol may still deliver value, but likely to different constituencies than intended. "It's just that the value will likely accrue to very large publishers and cross-publisher ad networks, in much the same way it has with programmatic more generally," Paparo wrote. This dynamic mirrors existing programmatic advertising patterns rather than transforming them.

Cross-publisher entities—essentially supply-side platforms—already collect signals and inventory at scale. These entities could package inventory for performance delivery to buyers and negotiate with buyside agents while assuming delivery and outcome risks. Paparo described this as "curation by wire" rather than prepackaged curation through deal IDs. This model reduces publisher risk by allowing yield maximization in prebid auctions with CPM-based payment.

Another application combines signal agents with media-buy agents for valuable, well-defined audiences. When buying agents seek tightly defined signals widely available in ad servers or curation platforms, agents could pre-negotiate data-driven programmatic guaranteed deals across many publishers. "In this example, the agent has real value because it reduces the per-publisher cost of doing business, and the execution path can be more efficient than using programmatic auctions," Paparo explained.

The broader context includes mounting skepticism about whether another protocol is needed. Augustine Fou, a fraud researcher and marketing consultant, cautioned that agentic AI does not eliminate bad actors in the supply chain. "More automation means less transparency," Fou stated. Agents can still act on behalf of people with bad incentives.

Industry analysis from July 2025 suggested that agentic AI poses existential threats to traditional demand-side platforms. Paparo previously argued that autonomous AI systems could automate campaign setup, targeting, and optimization functions currently handled by DSPs. The modern DSP represents "one of the most complex categories of software ever invented," but technological shifts threaten this established model through AI-driven alternatives.

McKinsey data indicates substantial investment momentum. The consulting firm reported $1.1 billion in equity investment flowing into agentic AI during 2024. Job postings related to the technology increased 985 percent from 2023 to 2024, according to McKinsey's Technology Trends Outlook 2025.

Multiple platforms have announced AI agent capabilities throughout 2025. LiveRamp introduced agentic orchestration capabilities on October 1, enabling autonomous AI agents to access identity resolution, segmentation, and measurement tools. Adobe launched its Experience Platform Agent Orchestrator on September 10 for managing agents across Adobe and third-party ecosystems.

The timing coincides with significant market disruptions. Microsoft announced on May 14 that it would discontinue Microsoft Invest (formerly Xandr) effective February 28, 2026. According to Microsoft Advertising Corporate Vice President Kya Sainsbury-Carter, the company cited incompatibility between traditional demand-side platform models and their vision for "conversational, personalized, and agentic" advertising futures.

Industry veteran David Kohl argued that AdCP represents "the tail wagging the dog", warning that protocol-first development creates predictable winners and losers rather than optimal outcomes. Kohl called for shared goals, structured innovation approaches, and measurable success metrics before committing to specific technical implementations.

However, Ben Kahan, senior director of programmatic at Brainlabs, defended the standardization effort. "The industry goes through eras, and now we seem to be entering an agentic era. But everything is siloed or fragmented. There hasn't yet been a push for standardization across any of it," Kahan stated.

Jason Widup, senior vice president of marketing at Pixis, described the protocol as "exactly the kind of infrastructure-level shift the advertising industry needs." Pixis built its operating system on Model Context Protocol as a shared intelligence layer. Widup sees the protocol as "a way to reduce friction across fragmented tech stacks, unify signals between partners, and enable more automated, intelligent decision-making."

The protocol does not address all advertising automation challenges. Fraud detection, viewability measurement, and attribution modeling remain platform-specific capabilities. Payment processing, invoice reconciliation, and financial reporting operate outside protocol scope.

Google Ads marked its 25th anniversary on October 23, 2025, emphasizing transformation from manual optimization to AI-powered automation serving over one million active advertisers globally. The anniversary reflection noted that automation means jobs have "become more strategic and focused on providing the business-informed guidance and data foundation that AI systems need to succeed."

Single-platform automation represents where most ad tech vendors currently focus. When thousands of people type into user interfaces, using AI to eliminate typing becomes competitively vital. This differs from truly agentic workflows, though overlapping investment exists between building chat interfaces and enabling MCP access.

Use cases where customers remain consistent across different platforms probably offer the easiest AdCP adoption paths. An optimization startup might build an agent to optimize DSP campaigns for common customers. That DSP might offer its own chat-based AI wizard while also providing MCP-based hooks for trusted third parties.

Paparo concluded with tentative assessments rather than definitive judgments. The protocols receive welcome recognition. Immediate use cases could deliver substantial benefits. Media-buying use cases remain problematic, especially around outcomes. Publisher intermediaries merit attention regarding how they enable buying through private marketplaces, programmatic guaranteed deals, and data-driven media buys.

The analysis matters for the marketing community because it provides practical assessment of technical capabilities versus business realities. While protocol development represents important infrastructure work, success depends on whether it addresses actual market needs rather than assumed problems. The divergence between creative automation potential and media buying skepticism suggests that technical standards alone cannot overcome fundamental business model misalignments.

Timeline

• October 15, 2025: Ad Context Protocol launches with backing from Scope3, Yahoo, PubMatic, and other vendors
• October 23, 2025: Google Ads celebrates 25 years marking shift from manual campaigns to AI automation
• October 1, 2025: LiveRamp introduces agentic orchestration capabilities for identity resolution and measurement
• November 2, 2025: Industry skepticism emerges about AdCP protocol necessity
• November 3, 2025: Ari Paparo publishes detailed analysis expressing support for creative protocol while questioning media buying applications
• July 21, 2025: Ari Paparo argues agentic AI threatens traditional DSP business models
• September 10, 2025: Adobe launches Experience Platform Agent Orchestrator for managing AI agents
• July 27, 2025: McKinsey reports $1.1 billion investment in agentic AI during 2024

Summary

Who: Ari Paparo, founder and CEO of Marketecture Media, published analysis on November 3, 2025, examining the Ad Context Protocol. The protocol initiative includes Scope3, Yahoo, PubMatic, The Weather Company, and Brian O'Kelley among its backers.

What: Paparo provided critical assessment of AdCP, expressing strong support for creative automation specifications while raising substantial concerns about media-buy protocol viability. He questioned whether agents can overcome fundamental business problems including pricing transparency resistance, scale economics, and data fragmentation that have historically prevented programmatic direct success.

When: The analysis appeared on November 3, 2025, following the October 15, 2025 protocol launch and amid broader industry developments including Microsoft's announcement to discontinue Microsoft Invest by February 28, 2026.

Where: The analysis addresses the broader programmatic advertising ecosystem, particularly focusing on relationships between buyers, sellers, intermediaries, and technology platforms operating across the open web market.

Why: The analysis matters because it distinguishes between technical capabilities and business viability in advertising automation. Paparo argues that while creative and potentially signals protocols offer genuine value, media buying automation faces obstacles that technical standards cannot overcome—including misaligned incentives between buyers and sellers, pricing transparency resistance, and the enduring role of intermediaries in achieving scale. The assessment provides practical perspective for marketing professionals evaluating whether to invest resources in implementing AdCP standards.