Amazon Ads on April 15, 2026, launched automatic deal selection inside Amazon DSP, a machine-learning feature that removes the need for advertisers to manually browse and activate programmatic deals for Streaming TV campaigns.

Amazon DSP has introduced automatic deal selection, a feature that uses machine learning to build and continuously update a curated group of programmatic deals based on campaign objectives and targeting criteria. According to Amazon Ads, the feature is available for Streaming TV campaigns where the Goal is set to Awareness and the KPI is set to Reach. It launched on April 15, 2026, in the United States, accessible to self-service advertisers through both the ADSP Self Service interface and the Amazon Ads Public API.

The announcement adds another layer of automation to a platform that has been accumulating machine-learning-driven tools since at least October 2024. For marketing teams running lean operations, the practical implication is direct: the system handles deal curation tasks that previously consumed significant manual effort.

How automatic deal selection works

During line item setup, advertisers navigate to the Inventory section and activate the feature by checking the box labeled "Automatic deal selection." From that point, the system takes over. According to Amazon Ads, it regularly monitors and updates eligible deals throughout the campaign flight - adding high-performing deals and removing under-performing ones without manual intervention.

The mechanism is a departure from the traditional deal management workflow, where buyers review lists of available deals, assess their suitability against campaign goals, and manually assemble inventory groups. That process can involve hundreds of options. Amazon's documentation describes the feature as reducing the need for "manually navigating through hundreds of available deals to find opportunities that align with your campaign objectives."

Machine learning drives the continuous optimization cycle. The feature analyzes performance data on an ongoing basis and adjusts the composition of the deal group accordingly. There is no stated threshold for how frequently updates occur - the documentation describes the cadence only as regular monitoring throughout the campaign flight.

The API implementation

The feature extends beyond the user interface. According to Amazon Ads, automatic deal selection is also accessible through the Amazon Ads API via the Inventory Group resource under the Deal Management API, currently in beta.

When automatic deal selection is enabled on a line item, the system creates and manages an inventory group with the curationSourceType parameter set to AMAZON_CURATED_AGENT. This distinguishes machine-curated groups from the two other existing types: advertiser-curated groups and Amazon-curated groups. The distinction is made at the API level, meaning that developers and technical buyers integrating with the platform can identify and query these groups programmatically.

Key fields within the Inventory Group resource include inventoryGroupId, name, inventoryType, advertisingDealIds - which holds the collection of deal IDs that the system has assembled - and createdInRegion.

To create an automatic deal selection line item programmatically, advertisers must create an inventory group with curationSourceType set to AMAZON_CURATED_AGENT and inventoryType set to STREAMING_TV. Amazon Ads specifies that automatic deal selection is only supported for line items in streaming TV awareness campaigns. Other campaign types are not eligible at launch.

Advertisers can retrieve the inventory group and its associated deals through the API, and can monitor deal-level delivery metrics via two dedicated entities: DealPerformance and DealPlanningMetrics. Both provide programmatic access to performance data that would otherwise require manual review inside the interface.

This API architecture is relevant in light of recent changes to Amazon's inventory management API framework, documented in March 2026. Amazon at that point launched unified inventory management APIs in open beta, consolidating deals, proposals, and inventory groups under the API v1 model. The automatic deal selection feature slots into that updated framework using the same consolidated endpoint structure.

Who built it

Nicole Goksel, Senior Product Manager - Technical at Amazon Ads, led the feature from concept to launch. According to her LinkedIn post announcing the release, she spent the first decade of her career on the publisher side, building deals that went unactivated. "The ecosystem rewarded relationships and manual workflows more than performance," she wrote.

Goksel, who joined Amazon in May 2022 and moved into her current ADSP Deals role in December 2024, described the feature as an attempt to change that dynamic. "This feature automatically selects, activates, and optimizes programmatic deals based on campaign goals and targeting criteria," she said in the post. "It makes the deal ecosystem more meritocratic: if your deal performs, it gets selected. No more sitting dormant because no one found it in a sea of deals."

The engineering team acknowledged in the announcement included Kiran Anantha, Abhilash Karanam, Rujuta Deshpande, Mei Leng, Riya Patni, Abhishek Mugal, Rahul Shukla, Saket Relan, Pratik Kamanboina, Venkata Sairaj Vuriti, and Jason Lin.

Kartal Goksel, Chief Technology Officer at Seedtag, commented on the LinkedIn announcement: "Great deals getting no traction because no one found them in a sea of options has been one of AdTech's open secrets for years. Activating on performance signals instead of relationship inertia flips the incentive in exactly the right direction."

Scope at launch

The feature is limited at launch. It covers only Streaming TV campaigns configured with an Awareness goal and a Reach KPI, and it is available exclusively in the United States. Amazon Ads indicates that expansion is planned across additional campaign goals and formats, though no timeline for that expansion was provided.

Self-service advertisers are the designated audience at launch. Managed service advertisers and other buyer categories are not addressed in the initial announcement.

Context: the deal discovery problem

The problem the feature addresses has been a structural friction point in programmatic advertising for years. Publishers create deals and make them available inside DSP interfaces. Buyers, faced with long lists of available inventory sources, tend to activate a subset they already know or have worked with previously. New deals, deals from smaller publishers, and deals that lack an existing relationship to support activation often go unused - not because of poor performance potential, but because of low visibility.

This dynamic creates a compounding disadvantage for publishers who lack existing agency relationships. Deals that perform well but are never activated produce no revenue signal. Amazon Ads launched Supply Desk in June 2024 as a consultative service to address part of this friction, offering pre-curated deals and prescriptive packaging recommendations for buyers. Automatic deal selection takes a different approach - instead of making curation available through a service layer, it embeds curation directly into the line item setup workflow.

Amazon Publisher Cloud, launched in June 2024, addressed a related problem from the supply side - enabling publishers to use first-party signals and Amazon audience data to build more targeted deals before making them available to buyers. Automatic deal selection sits on the opposite end of that pipeline: it concerns how deals already in the marketplace get discovered and activated by buyers.

Implications for streaming TV buyers

Streaming TV has become a significant focus area for Amazon's advertising infrastructure. The platform added Disney's Real-Time Ad Exchange in June 2025, Spotify's inventory in October 2025, and Netflix's programmatic access beginning in Q4 2025. The expansion of available inventory has made the deal discovery challenge more acute, not less. With more publishers available, the volume of deals buyers need to evaluate grows accordingly.

Amazon's exclusive reach measurement for streaming TV inventory, announced in December 2024, provided advertisers with tools to understand unduplicated audience reach across streaming channels like Twitch and Prime Video. Automatic deal selection can be understood as a paired capability: where reach measurement tells advertisers how well their current deal selection is performing, automatic deal selection adjusts the composition of that deal set without requiring manual input.

For buyers running awareness campaigns with a Reach KPI, the primary optimization objective is maximizing unduplicated audience exposure within budget constraints. The machine learning system is designed to pursue that objective by continuously assessing which available deals contribute to reach and adjusting accordingly. How quickly the system reacts to changes in deal performance and what data lookback window it uses are technical details not disclosed in the current documentation.

Smartly's integration with Amazon DSP, announced in March 2026, brought AI-powered creative optimization to streaming TV buying through a third-party layer. Automatic deal selection operates at the inventory layer, addressing a different part of the campaign assembly workflow. Together these tools reflect a broader push toward reducing manual setup requirements for streaming TV campaigns.

What is not yet available

The feature is not available for performance-oriented campaign goals such as conversions or consideration at launch. It is also restricted to the United States, leaving international markets without access for the time being. Amazon Ads does not specify whether expansion will occur on a market-by-market basis or through a broader international rollout.

The beta status of the Deal Management API means that API-based implementations may be subject to change. Advertisers building programmatic workflows around the AMAZON_CURATED_AGENT curation type should account for potential modifications to the API specification before it reaches general availability.

Timeline

Summary

Who: Amazon Ads, operating through Amazon DSP. The feature was led by Nicole Goksel, Senior Product Manager - Technical for ADSP Deals, with an engineering team of more than ten contributors.

What: Automatic deal selection is a machine-learning feature that automatically curates, activates, and continuously updates a group of programmatic deals for Streaming TV campaigns. The system uses the curation source type AMAZON_CURATED_AGENT in the Deal Management API, and surfaces deal-level performance data through the DealPerformance and DealPlanningMetrics API entities.

When: Announced and launched on April 15, 2026.

Where: Available in the United States only at launch, accessible to self-service advertisers through the ADSP Self Service interface and the Amazon Ads Public API.

Why: Manual deal management in programmatic advertising requires buyers to navigate hundreds of available deals, a process that disadvantages smaller publishers and slows campaign activation. The feature replaces that manual workflow with continuous machine-learning optimization, removing the dependency on pre-existing relationships or manual browsing. Amazon Ads describes the target audience as advertisers with lean in-house teams focused on maximizing reach efficiently.

The launch reflects a broader direction inside Amazon's advertising infrastructure. As the platform has expanded its streaming TV inventory - adding Netflix, Spotify, Disney, and other premium publishers over the past 18 months - the number of deals available to buyers inside the DSP has grown substantially. Larger inventory pools make manual curation harder to scale, particularly for smaller teams. Automatic deal selection responds to that scaling problem directly. Whether similar automation eventually extends to non-awareness campaign goals, or to non-streaming formats, will depend on how the machine learning models perform against the constrained initial scope. The current launch is narrow by design, and Amazon has left the expansion timeline open.

Share this article
The link has been copied!