What are curated audiences, and how do they reach a bid request?
Part of Curated deals and PMPs
The short answer
Curated Audiences is IAB Tech Lab's addressability specification for first-party audience data, called Seller Defined Audiences until December 2024. A publisher or data partner groups visitors by attribute, maps each group to a taxonomy ID with a Data Transparency Standard label, and sends that ID in the OpenRTB bid request instead of a user identifier.
Key takeaways
- IAB Tech Lab renamed Seller Defined Audiences to Curated Audiences in December 2024 because audiences can be defined by multiple parties, not only sellers.
- The signal is three data points in the bid request: the provider's domain, the segment IDs and a segtax number naming the taxonomy those IDs come from.
- Audience segments ride in user.data and context segments in site.content.data, and the specification expects them to travel without user IDs.
- A curated audience describes people a provider has grouped; a curated site list describes pages and publishers a buyer can open and review.
- Segment labels are self-attested, so the questions a buyer asks about provenance, age, modeling and cohort size matter more than the label's name.
What are curated audiences?
Definition
- Curated Audiences (formerly Seller Defined Audiences)
- An IAB Tech Lab addressability specification, incubated within Project Rearc, in which a publisher or its data partner signals audience or context groups in OpenRTB using standard taxonomy IDs and disclosure labels rather than cookies or device IDs.
IAB Tech Lab's Curated Audiences page (opens in a new tab) says the specification allows publishers, DMPs and data providers to scale first-party data responsibly and reliably without data leakage or reliance on deprecated IDs. It does not invent a new protocol. It combines pieces that already existed: Prebid integrations, OpenRTB 2.6 community extensions, IAB Tech Lab's Transparency Center and the Data Transparency Standard.
The document behind it is the February 2022 version 1.0 specification (opens in a new tab) from the Rearc Addressability Working Group. In it, publishers or their data partners determine audience attributes based on customer interactions on their properties, map those attributes to standardized taxonomy descriptions and data transparency disclosures, and relay those anonymized taxonomy IDs within OpenRTB.
Why was Seller Defined Audiences renamed?
In December 2024, Seller Defined Audiences was renamed Curated Audiences. Tech Lab explained the change in a December 16, 2024 post (opens in a new tab) that also started a Curation Framework working group. The post was candid: it said the specification had not gained the adoption it deserved, and that Tech Lab had asked the industry why.
We’re kicking off a Curation Workstream in 2025 and rebranding Seller Defined Audiences as Curated Audiences, as audiences can be defined by multiple parties and not only sellers.
The post listed topics for the working group: more ways to prevent data leakage for publishers, visibility of curation or enrichment partners, and plans for testing in 2025. The Curated Audiences page, last updated August 25, 2025, said work on a Curation Framework starting with standards to streamline the Deals ID process was underway, with a release expected in Q4 2025. Tech Lab's Deals API version 1.0 was in public comment until January 31, 2026 and was finalized on February 6, 2026.
How do curated audiences travel in a bid request?
The specification asks the provider to relay three data points through the OpenRTB Segment Taxonomies community extension, known as segtax: the provider name, meaning the unique domain of the entity making the cohort determination; the segment IDs; and the taxonomy those IDs belong to. Together they let a buyer look up the full disclosure label for that segment in Tech Lab's Transparency Center without sending the whole label in every request.
| Data point | Audience signal | Context signal | What it tells a buyer |
|---|---|---|---|
| Provider | user.data.name | site.content.data.name | Which company decided the visitor or page belongs in the group |
| Taxonomy | user.data.ext.segtax | site.content.data.ext.segtax | Which list the IDs come from, for example 4 for Audience Taxonomy 1.1 |
| Segments | user.data.segment.id | site.content.data.segment.id | The group or category IDs that apply to this impression |
The segtax extension (opens in a new tab) keeps the list of taxonomy numbers. Standard values include 4 for IAB Tech Lab Audience Taxonomy 1.1 and 7 for Content Taxonomy 3.0; values of 500 and up are vendor specific, registered by pull request. The extension also says segment IDs should still be considered fully proprietary when no segtax is provided. Our Audience Taxonomy guide covers what the standard labels mean.
The 2022 document states that integration with Prebid.js is a prerequisite to adopt SDA signaling, and that requests to DSPs are expected to be devoid of any user id, cookie, IP or deal ID information. So the audience travels as data in the request, and no deal ID is needed to bid on it.
How is a curated audience different from a curated site list?
Both are called curation, and both can be sold through private marketplace deals, but they select different things. A curated audience selects impressions by who a provider says the visitor is, or what a provider says the page is about. A curated site list selects impressions by where they appear: named domains a buyer has reviewed. Each has real strengths and real limits.
| Question | Curated audience | Curated site list |
|---|---|---|
| What is selected | Visitors or pages grouped by a provider's attribute | Specific domains chosen against a brief |
| What the buyer can inspect | A self-attested disclosure label, not the members | The sites themselves, page by page |
| Reach across sites | Can span the sites where the provider's data runs | Limited to the listed domains |
| Signal in the request | user.data or site.content.data with segtax | site.domain matched to the agreed list |
| Personal data involved | First-party data processed by the provider, sent as group IDs | None needed to define the list |
| Main risk | Labels that overstate what a segment contains | A list too narrow to deliver, or sites that stop selling |
The two can also be combined: a deal can be built on a reviewed site list and also carry a provider's audience data. IAB Tech Lab's Deals API has an auxdata field for exactly this, indicating whether non-publisher data with an associated fee is applied to a package and whether it may change after the deal starts. When a proposal mixes both, ask which part is the site list and which part is the audience layer, and who is paid for each.
What does the specification ask of audience providers?
The 2022 specification names four areas where data security and accountability expectations need to be set. They are a practical part of the document for a buyer, because each one becomes a question to ask a seller.
| Area | What the specification says | What a buyer can ask |
|---|---|---|
| Accuracy of self-attested labels | Self-attestation opens financial incentives to misrepresent an attribute; its example is a games enthusiasts cohort relabeled as high net worth individuals | Has the provider completed the Data Transparency Standard compliance program? |
| Consumer privacy preferences | Cohort developers are expected to build groups in line with regional legislative requirements on transparency and choice | Which consent signals govern who enters the segment? |
| No commingling with identifiers | Cohorts should travel in isolation from user agent data, pseudonymous IDs or encrypted user provided IDs | Does the request carry any user ID next to the segment? |
| Sufficiently large cohorts | Groups should be large enough that no individual can be identified; Tech Lab cites a 50 user query limit in one Google product as a directional starting point only | What minimum size does the provider apply, and how is it counted? |
What should a buyer ask about a curated audience?
Questions to put to any curated audience proposal
- The provider's domain, as it will appear in user.data.name or site.content.data.name, is named in writing.
- The segtax value is stated, and a vendor value of 500 or more is matched to its entry in the public segtax list.
- Each segment ID has a Data Transparency Standard label the buyer can read before the flight starts.
- The label states where the data came from, how old it is and how much of it is modeled.
- The provider states how a visitor qualifies, in business rules rather than a segment name.
- The proposal says whether any user identifier travels in the same request as the segment.
- The proposal says who is paid for the audience layer, separately from the media.
What does a curated audience review look like on a real brief?
Fictional example
A heat pump installer weighs two curation proposals
A fictional heat pump installer serving New York, Massachusetts and Connecticut wants display beside home energy and home improvement content in the fall, when homeowners plan heating upgrades. Its agency receives two proposals. One is an audience package described as home improvement intenders, sold on a deal. The other is a site list of home renovation, energy efficiency and regional homeowner sites.
For the audience package, the planner asks for the provider's domain, the segtax value and the disclosure label for each segment. The label shows the segment is partly modeled from a provider's own sites, refreshed monthly, with qualifying rules based on article reading. The planner records that the label is self-attested and asks whether any user ID travels with it.
For the site list, the planner opens sample articles from each site, removes two online stores, and checks who sells each site on the chosen exchange. The client sees both, with what each selects and who is paid, and chooses.
How does PlacementGraph help when a buyer prefers context to audience data?
Frequently asked questions
Do curated audience signals carry user IDs?
The specification says they should not. Its design principles rule out relying on third-party cookies, mobile or OS IDs, user-provided IDs or user agent information, and its exchange guidance expects requests to DSPs to be devoid of user ID, cookie, IP or deal ID information. Whether a given request follows that is something to check in a sample.
Who maintains the list of segtax taxonomy numbers?
IAB Tech Lab keeps it in the segtax community extension on GitHub. Standard taxonomies have low numbers, such as 4 for Audience Taxonomy 1.1. Any company can request a vendor value of 500 or more through a pull request, which Tech Lab approvers review for ID conflicts rather than through a formal working group review.
Can two providers' segments with the same name be treated as equal?
No. The specification's own example is an auto intender cohort from one publisher that will likely differ from an auto intender cohort from another, despite the same standard name. Business rules, data age and modeling differ, which is why the disclosure label for each provider's segment has to be read separately.
Does Curated Audiences work outside web browsers?
IAB Tech Lab lists that it can be used for inventory across all browsers, apps and OTT or CTV environments. The 2022 specification's implementation steps are written around Prebid.js, so how a given app or CTV seller passes the same signals is a question to ask that seller directly.
Sources
- Curated Audiences (formerly Seller Defined Audiences) (opens in a new tab)IAB Tech Lab
- Curated Audiences and launching the Curation Framework Working Group (opens in a new tab)IAB Tech Lab
- Taxonomy and Data Transparency Standards to Support Seller-defined Audience and Context Signaling, v1.0 (opens in a new tab)IAB Tech Lab
- Segment Taxonomies (segtax) community extension (opens in a new tab)IAB Tech Lab
- OpenRTB 2.6 specification: User, Content, Data and Segment objects (opens in a new tab)IAB Tech Lab
- Deal Sync API specification 1.0 (opens in a new tab)IAB Tech Lab
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.