Brian O'Kelley, co-founder and CEO of Scope3, published a perspective on March 30, 2026, at the Agentic Advertising Organization's website arguing that ads.txt - the decade-old supply-chain transparency standard - lacks the vocabulary to describe how modern media is actually sold. The piece, published at agenticadvertising.org, introduces adagents.json as a structured alternative capable of expressing scoped authorization, delegation hierarchies, and country-level restrictions in a single machine-readable file.

The argument is not simply that ads.txt is old. It is that the format was designed to answer a narrower question than the industry now needs answered.

What ads.txt actually does

ads.txt, short for Authorized Digital Sellers, launched roughly a decade ago as a lightweight text file that publishers host on their domains. Each line declares a seller - typically an exchange or SSP - along with an account identifier and a relationship type: DIRECT or RESELLER. The format was intentionally simple. According to O'Kelley, that simplicity was deliberate: "The preference was to keep it simple 10 years ago, and keep extending."

The basic format looks like this:

ssp.pinnacle.example, pub-123, DIRECT
novaexchange.example, acct-456, RESELLER
riverline.example, acct-789, RESELLER

That structure answers two questions cleanly. Is a seller declared at all? Is the relationship labeled direct or reseller? It does not, according to O'Kelley, answer which property the authorization applies to, which placement, which country, which dates, or which delegated commercial path. Those omissions were tolerable when publisher monetization was simpler. They are less tolerable now.

sellers.json was introduced later to add a complementary layer, hosted by SSPs rather than publishers, listing the publishers and intermediaries they work with. The SupplyChain Object extended transparency further by letting buyers trace the full chain of participants in a bid request. But even combined, according to O'Kelley, these mechanisms "are still partial signals spread across multiple layers."

Where ads.txt breaks down for CTV

Connected television is the sharpest stress test. A single publisher may have a direct path for premium homepage video, a separate programmatic channel for banner and native units, a third path active only in certain countries during a specific rights window, and a fourth path representing a network-managed slot on the same page. According to O'Kelley, "Those are not edge cases. They are normal commercial conditions."

CTV inventory's commercial complexity has been a persistent challenge for supply-chain transparency frameworks. Sara Sinclair, PhD, VP of Ad Platforms and contributor to the AdExchanger piece that O'Kelley references in his post, wrote that the nuance of how CTV inventory is sold requires fundamental changes to how the industry models the supply chain. O'Kelley updated the adagents.json specification partly in response to Sinclair's analysis.

The problem is concrete: a buyer looking at an ads.txt file for a major streaming publisher cannot determine from that file alone whether the network slot at the bottom of a page is the same commercial product as the publisher's premium pre-roll. Without that distinction, supply path optimization cannot function cleanly - buyers attempting to consolidate paths may accidentally exclude valid premium inventory while retaining low-value placements.

The IAB Europe Guide to Programmatic for CTV, published in April 2026, identified similar gaps. Programmatic CTV budgets reached 26% of media spend in 2026, up three percentage points year-over-year, and more than half of marketers expect over 60% of CTV buys to occur programmatically during the year. The scale of that reallocation makes the absence of a clear publisher-declared inventory namespace more consequential, not less.

What adagents.json adds

The adagents.json format starts from a different premise. Rather than declaring who is present, it lets publishers declare what each path is authorized to make available. According to O'Kelley, the specification introduces several new structural elements.

delegation_type distinguishes between three relationship categories: direct, delegated, and ad_network. That single field replaces the binary DIRECT/RESELLER distinction that ads.txt has used since launch.

placements create what O'Kelley calls "a publisher-governed inventory namespace with stable placement_id values." A publisher can define a placement called hero_video and associate it with specific properties, tags, and authorized agents - separately from article_banner or a managed network feed. Each placement_id is stable and publisher-controlled, not inferred from platform-specific conventions.

placement_tags let publishers define commercial groupings. Tags like programmatic, direct_only, or managed_by_riverline can then be referenced in authorized_agent entries to scope which agents can access which types of inventory. A buyer agent, in theory, can determine before bidding whether a given placement is programmatically available at all.

countries and the pair effective_from / effective_until let publishers attach geographic and temporal qualifiers to authorizations. A rights window that expires at the end of a quarter can be expressed directly in the file rather than inferred from platform-level signals.

exclusive signals whether a path is sole or concurrent with other authorized paths. signing_keys provide a publisher-attested trust anchor for signed agent responses, connecting the format to the emerging infrastructure of agentic advertising protocols.

The same commercial reality that requires three identical-looking ads.txt RESELLER lines to describe can be expressed in adagents.json with explicit placement scoping:

{
  "authorized_agents": [
    {
      "url": "https://ssp.novaexchange.example",
      "authorized_for": "Programmatic supply",
      "authorization_type": "property_ids",
      "property_ids": ["pinnacle_news"],
      "placement_tags": ["programmatic"],
      "delegation_type": "delegated",
      "countries": ["US", "CA"]
    },
    {
      "url": "https://riverline.example",
      "authorized_for": "Bottom-of-page recirculation inventory",
      "authorization_type": "property_ids",
      "property_ids": ["pinnacle_news"],
      "placement_ids": ["riverline_feed"],
      "delegation_type": "ad_network"
    }
  ]
}

According to O'Kelley, that level of specificity tells a buyer something much closer to commercial reality than the flat seller list ads.txt produces.

The accretion problem

O'Kelley raises a structural risk with any transparency standard that evolves through incremental extension. "The risk with any transparency standard is that it accumulates vague labels faster than it accumulates clarity," he writes. "That is how systems become technically richer while becoming harder to interpret."

Neal Richter, Vice President of Advertising Science and Engineering at Amazon, commented directly on O'Kelley's LinkedIn post to acknowledge the limits of ads.txt while defending its original design rationale. "Happy to discuss ways to fix/evolve/replace ads.txt Brian," Richter wrote. "It stands for Authorized Digital Sellers. A Seller is an agent in the prior sense of the word. Your 'which' questions are good ones, and ones that were considered." Richter added that he "always appreciated the depth of thought on data models/schemas at AppNexus" - a reference to O'Kelley's earlier company, the exchange that became Xandr.

O'Kelley responded by inviting Richter to join the AAO Slack working group, signaling that the specification is under active collaborative development rather than unilateral publication. The involvement of Richter, whose role spans Amazon's advertising science and engineering functions, is notable given Amazon's participation in the broader agentic advertising infrastructure being developed in parallel at IAB Tech Lab.

Industry context: agentic pressure

The timing of O'Kelley's piece is not coincidental. The advertising industry has spent the past several months building out the infrastructure for agentic media buying - systems in which AI agents autonomously discover inventory, negotiate terms, and activate campaigns. That infrastructure requires machine-readable publisher authorization files that go considerably beyond what ads.txt can express.

The Ad Context Protocol, launched in October 2025 by a coalition including Scope3, Yahoo, PubMatic, Swivel, Triton Digital, and Optable, defined a framework for AI agents to communicate across advertising platforms. For that framework to function, buyer agents need a reliable, structured way to determine what a publisher has actually authorized. An agent that can only read a flat ads.txt file lacks the resolution to make those determinations without platform-specific inference.

Prebid.org took stewardship of the Prebid Sales Agent on January 29, 2026, in collaboration with AgenticAdvertising.org, extending open-source header bidding infrastructure to support agentic direct buying. The AdCP's signing_keys field in adagents.json is designed to connect publisher authorization with the trust verification layer those agents need.

IAB Tech Lab formally named its parallel initiative AAMP - Agentic Advertising Management Protocols - on February 26, 2026, covering execution, protocol, and an Agent Registry that reached 10 active entries by March 11. Amazon's MCP server is among the listed entries. Industry debate over which standards frameworks will achieve sufficient adoption to become genuine infrastructure, rather than parallel experiments, remains unresolved.

The IAB Europe report on scaling agentic advertising, published in March 2026, noted that agents need "the right kind of scaffolding" to be made competent, and that the conditions for success are specific to each use case. Publisher authorization files are part of that scaffolding. If the file format cannot express what a publisher actually authorizes, no amount of agent intelligence compensates for that gap.

Publisher perspective

Mykyta Plastomak, from TeqBlaze, commented on O'Kelley's LinkedIn post to argue that publishers can already move toward cleaner supply chains through demand path optimization - owning demand relationships, optimizing paths, and reducing dependency on intermediaries - without waiting for a new standard. That perspective reflects a practical reality: publishers have been adapting supply-path relationships through existing mechanisms for several years.

But O'Kelley's argument is not that adagents.json replaces those commercial relationships. It is that those relationships need a structured representation publishers can declare once and buyers can read reliably. The distinction between a delegated SSP path limited to programmatic inventory in the US and Canada, and a network-mediated recirculation feed with no programmatic access, cannot be expressed in ads.txt without combining multiple external files and platform-specific inference. adagents.json attempts to express that difference directly.

Sara Sinclair, VP of Ad Platforms and the AdExchanger contributor whose article on CTV complexity prompted O'Kelley's spec update, is credited in O'Kelley's post as an AdCP contributor. "I look forward to learning more about adagents.json," Sinclair wrote in the LinkedIn thread, "and am happy to see so many in alignment that the current setup does not transparently state the variety of relationships that exist in the CTV ecosystem."

What this means for buyers and measurement

Supply path optimization has become a baseline expectation in programmatic buying. Three-quarters of buyers surveyed in early 2026 said that curated marketplaces, private marketplaces, or deal ID-based premium supply would be important or very important for their 2026 programmatic strategies. Buyers approaching curation as a layered control system need more than a seller list - they need placement-level clarity about what is actually available and through which path.

The adagents.json format, if adopted, would give buyers a publisher-declared model against which bid requests could be validated. A bid request claiming to represent a publisher's premium pre-roll placement could be checked against the publisher's adagents.json to verify whether the submitting agent is authorized for that placement, in that geography, during that date window. That is a materially different verification capability than checking whether a seller ID appears anywhere in a flat text file.

According to O'Kelley, "The future of supply-chain transparency is not a slightly longer text file. It is a publisher-declared model that reflects how inventory is actually sold." Whether the industry moves in that direction depends on publisher willingness to maintain more complex authorization files and on buyer-side tooling capable of reading and validating them. The specification is open for contribution through the AAO working group.

Timeline

Summary

Who - Brian O'Kelley, co-founder and CEO of Scope3 and a founding member of the Agentic Advertising Organization, published the adagents.json perspective. Neal Richter (VP, Advertising Science and Engineering, Amazon), Sara Sinclair PhD (VP of Ad Platforms), and Mykyta Plastomak (TeqBlaze) contributed to the LinkedIn discussion. Sara Sinclair is credited as an AdCP contributor whose CTV analysis prompted a spec update.

What - O'Kelley argues that ads.txt, the industry's primary supply-chain transparency standard since roughly 2014-2015, cannot express the commercial complexity of modern publisher monetization, particularly in connected television. adagents.json introduces structured authorization with delegation_type, placement_ids, placement_tags, country scoping, date windows, exclusivity flags, and signing_keys - allowing publishers to declare what each path is authorized to make available rather than simply listing who is present.

When - The perspective was published on March 30, 2026, at agenticadvertising.org. The LinkedIn discussion attached to the post occurred approximately three weeks before the publication date based on the timestamps visible in the document.

Where - The specification and perspective are published at agenticadvertising.org, the website of the Agentic Advertising Organization. The LinkedIn discussion took place on Brian O'Kelley's public LinkedIn profile. The standard applies globally wherever programmatic advertising inventory is transacted.

Why - The emergence of agentic advertising - systems in which AI agents autonomously discover, negotiate, and activate media - requires machine-readable publisher authorization files that go beyond what ads.txt can represent. An agent reading a flat ads.txt file cannot determine placement-level authorization, delegation type, geographic scope, or temporal validity without platform-specific inference that varies by system. adagents.json attempts to provide that structured layer so that buyer agents and seller agents can interact with publisher authorization that reflects actual commercial terms.

Share this article
The link has been copied!