Seller-defined audiences (SDA) are groups of readers that a publisher, or a data company acting for it, builds from its own first-party data and describes to buyers with numeric codes from a shared industry taxonomy. The codes travel in programmatic bid requests in place of cookies or device identifiers, so a buyer learns that an impression belongs to a segment such as "Interest | Sports | American Football" without receiving anything that points to the person behind it. IAB Tech Lab finalised the specification in February 2022 to keep audience targeting workable on the open web as cross-site identifiers disappeared. Since December 2024 the standards body has promoted it under a new name, Curated Audiences, although the mechanism is unchanged.

How a segment reaches the bidder

A publisher first decides who its visitors are, using registration data, subscriptions, reading behaviour or survey answers. Each attribute is mapped to a node in the IAB Tech Lab Audience Taxonomy 1.1. The copy IAB Tech Lab maintains on GitHub lists 1,558 nodes numbered up to 1,679 across three top-level branches: 197 demographic, 497 interest and 864 purchase-intent entries. The node IDs are then written into an OpenRTB 2.x bid request, where a demand-side platform (DSP) reads them and decides whether, and how much, to bid.

Placement inside the request separates people from pages. Audience signals sit in user.data, while contextual signals describing the page or video go in site.content.data or app.content.data and draw on the Content Taxonomy. Each Data object carries three elements: name, the domain of the entity that made the classification; ext.segtax, an integer identifying the taxonomy; and a segment array holding the IDs. A publisher that has classified a reader as a woman aged 30 to 34 with an interest in American football would send its own domain as the name, segtax 4 and segment IDs 49, 6 and 608.

The segtax register does the translation. Value 4 denotes Audience Taxonomy 1.1, while 5, 6 and 7 denote Content Taxonomy 2.1, 2.2 and 3.0. Values 1 and 2, older content taxonomies without sensitive-category flags, are marked deprecated. Everything from 500 upward is reserved for proprietary vendor taxonomies, approved through a GitHub pull request rather than a working group vote. As of September 2026 the register holds roughly 90 vendor entries or reserved ranges, including Magnite at 507 and 508, Optable at 5000 to 5002, Permutive at 6000 to 6010 and a long block held by Sirdata for private taxonomies built for individual supply-side platforms (SSPs).

Segment descriptions do not fit in a bid request. Data providers instead publish a label under the Data Transparency Standard (DTS), which carries more than 20 disclosures on provenance, recency, modelling and segmentation criteria, and upload it to the IAB Tech Lab Transparency Center, where buyers retrieve it separately. The standard predates SDA: IAB Tech Lab completed version 1.0 in 2019 and tied its compliance seal to an annual audit.

Delivery usually runs through Prebid, the open-source header bidding framework, whose first-party data mechanism passes publisher values to bidder adapters, with data management platforms plugging in through real-time data modules. IAB Tech Lab's implementation guide, written by Hillary Slattery and published in October 2023, adds a rule for exchanges. Where a cookie or other identifier is present, SDA signals should be folded into deal IDs with no evident meaning rather than passed next to the identifier.

Google's variant

Google adopted the taxonomy approach under its own label. Peentoo Patel, director of product management for Google Ad Manager, announced publisher provided signals on September 13, 2022, with a beta sharing taxonomy-coded segments with Google Ads and Display & Video 360. The announcement quoted Anthony Katsur, chief executive of IAB Tech Lab, saying SDA was built "to help advertisers and publishers responsibly and reliably share first-party data at scale without the need for user identifiers."

According to Google's Authorized Buyers documentation, the signals arrive in user.data with a segtax value, cover Audience Taxonomy 1.1 and Content Taxonomy 2.2, and reach Google demand, Authorized Buyers, Open Bidding and SDK Bidding. The feature is limited to Ad Manager 360, and Google withholds some taxonomy values it treats as sensitive. PPC Land described the Ad Manager configuration in May 2024, when it still carried a beta label.

Origin and evolution

The specification grew out of Project Rearc, unveiled at the IAB Annual Leadership Meeting in February 2020, weeks after Chrome said on January 14, 2020 that it would phase out third-party cookies within two years. SDA was first proposed in March 2021, and a draft stayed open for public comment through May 2021. Testing began before the text was final: a Magnite-led group of more than 30 publishers, agencies and platforms, including News Corp, Conde Nast and PubMatic, said in June 2021 it had completed a first phase of cookieless transactions against seller-defined segments.

Version 1.0 arrived on February 24, 2022, drafted by the Rearc Addressability Working Group of 295 individuals from 146 organisations. It shipped alongside DTS version 1.1 and the Transparency Center. Benjamin Dick, senior director of product at IAB Tech Lab, told AdExchanger at the time: "We're trying to make sure publishers have firmer controls over how their audiences are used."

Uptake stayed thin. By December 2024 Katsur conceded the framework had been "sitting underappreciated in the corner", according to AdExchanger, and blamed the name: "Maybe we shouldn't have branded it 'seller-defined,' because a lot of buyers are like, 'I want to define the audience.'" IAB Tech Lab confirmed the switch to Curated Audiences on December 16, 2024, arguing that parties other than sellers can build segments, and scheduled a 2025 curation workstream on data leakage and on disclosing which enrichment partners touched a segment.

Why it matters

The specification gave publishers a way to sell what they know about readers without exporting the readers. That became raw material for curation, and IAB Tech Lab lists SDA among the four components of its curation framework, next to the taxonomies, the DTS and the SupplyChain object. Its ID-less guidance, open for comment from November 21 to December 19, 2024, counts publisher audience labels in bid requests among its methods for reaching people without identifiers.

Taxonomy IDs also pass through every hop in the supply chain unchanged, a property an IAB Tech Lab paper argued matters more once AI agents write campaign briefs. On the transparency side, Yahoo became the first DSP to show DTS labels inside its interface, announced in January 2025.

Limitations and disputes

Self-attestation is the central weakness, and the 2022 specification admits as much. Its authors warned that a provider could label a games enthusiasts cohort as high net worth individuals to lift prices, with the voluntary DTS audit as the main check.

Cohort size is not fixed either. The document called static minimums impractical and technically unenforceable, citing Google's refusal to run Ads Data Hub queries on groups of 50 or fewer users only as a directional reference. Other privacy systems set explicit thresholds, from 50 browsers in Protected Audience to 2,000 in the FLoC trial.

Leakage is a commercial risk too. If segment IDs travel next to device identifiers at scale, the specification noted, an audience built by premiumpublisher.com could be re-identified by an advertiser and bought on bobsblog.org, turning a publisher asset into a commodity.

Compliance has been poor. RTB House, a buy-side vendor with its own interest in cleaner signals, analysed about 135 billion SDA-enriched bid requests from 20 SSPs during the first two weeks of December 2022. According to its 2023 findings, about 33% of contextual signals and 75% of user signals failed the specification, mostly through a missing taxonomy or missing segment IDs. A separate RTB House survey found readers tagged with a cooking segment answered yes to cooking questions about 19% more often, in relative terms, than a control group, suggesting well-formed signals do carry information.

Not the same as

Topics API. Chrome assigned topics of interest inside the browser from its own taxonomy, registered as segtax 600 and 601. Neither the publisher nor its data partner chose the labels, and Google retired the API in October 2025.

Curated deals. A curator packages inventory and data into a deal ID. SDA is one possible input to that package, not the package itself.

DTS labels. A label describes how a segment was built. It holds no user data and cannot be used for activation.

Agentic Audiences. Formerly LiveRamp's User Context Protocol, it uses the same user.data container but carries embeddings, vectors of floating-point numbers, rather than taxonomy node IDs.

Recent developments

The pressure that produced SDA eased when Google said on April 22, 2025 that Chrome would keep third-party cookies, then retired Topics and most other Privacy Sandbox APIs on October 17, 2025, citing low adoption. Taxonomy-coded segments survived as a curation currency. IAB Tech Lab's final ID-less guidance, released on July 17, 2025, describes Curated Audiences as publisher segments built from first-party and contextual data. Its Deals API, which names the curator domain and fee type, was marked final on February 6, 2026.

Commercial activity has shifted towards pooled publisher data. Mediavine brought audience data from 18,000 publishers into Index Marketplaces in April 2026. In September 2026 BCN, the sales house of Hubert Burda Media, Funke and Klambt, detailed an agent that matches a plain-language brief against more than 900 audience segments and 600 contextual categories, then has PubMatic build the deal. LiveRamp, meanwhile, transferred its User Context Protocol to IAB Tech Lab on November 3, 2025. Its successor now fills the same bid request slot with embeddings, a sign that the taxonomy node may not remain the only way publishers describe an audience.

Timeline

  • January 14, 2020: Chrome announces it will phase out third-party cookies within two years
  • February 2020: IAB Tech Lab unveils Project Rearc at the IAB Annual Leadership Meeting
  • March 2021: Seller-defined audiences first proposed; public comment runs through May 2021
  • June 2021: A Magnite-led group of more than 30 companies completes a first phase of cookieless transactions against seller-defined segments
  • February 24, 2022: IAB Tech Lab releases SDA 1.0 with DTS 1.1 and the Transparency Center
  • September 13, 2022: Google announces publisher provided signals in beta
  • December 2022: RTB House samples about 135 billion SDA-enriched bid requests from 20 SSPs
  • March 8, 2023: RTB House publishes its first compliance results, finding 75% of user signals non-compliant
  • October 2023: IAB Tech Lab publishes the SDA implementation guide
  • December 16, 2024: IAB Tech Lab renames SDA as Curated Audiences and plans a 2025 curation workstream
  • January 2025: Yahoo DSP starts displaying DTS labels
  • April 22, 2025: Google says Chrome will keep third-party cookies
  • July 17, 2025: IAB Tech Lab releases final ID-less guidance
  • October 17, 2025: Google retires Topics and most other Privacy Sandbox APIs
  • November 3, 2025: LiveRamp transfers the User Context Protocol to IAB Tech Lab
  • February 6, 2026: The Deals API, with curator domain and fee fields, is marked final
  • April 14, 2026: Mediavine and Index Exchange announce audience data from 18,000 publishers in Index Marketplaces
  • September 21, 2026: BCN details an agent that converts plain-language briefs into PubMatic deals

Summary

Who. Publishers, sales houses and their data partners build and label the segments. IAB Tech Lab wrote the specification through its Rearc Addressability Working Group and maintains the taxonomies, the segtax register and the Transparency Center. SSPs and header bidding wrappers relay the signals; DSPs and their clients decide whether to buy. Google runs its own version for Ad Manager 360 publishers.

What. A seller-defined audience is a publisher-built segment expressed as a taxonomy node ID and sent in an OpenRTB bid request without a user identifier. The Data object names the classifier, a segtax integer identifies the taxonomy, and a separate DTS label describes how the segment was made.

When. The idea was proposed in March 2021 and finalised on February 24, 2022. Google followed with publisher provided signals in September 2022. IAB Tech Lab renamed the specification Curated Audiences in December 2024.

Where. The signals travel in user.data and site.content.data fields of programmatic bid requests across web, app and connected TV, most often through Prebid, SSP marketplaces and curated deals.

Why. The specification was meant to let first-party audience data travel without cross-site identifiers and without leaking to other sites. Chrome kept its cookies, but the format survived as a common vocabulary for curation, even as self-attestation, low compliance and the absence of a minimum cohort size continue to limit buyer trust.