Taboola opens DeeperDive monetization to third-party AI chatbots and agents

Taboola has opened the monetization engine behind its DeeperDive AI answer engine to any generative AI company, enabling chatbots, virtual assistants, and other AI products to convert user queries directly into advertising revenue without requiring a subscription model. The move, announced on June 16, 2026, extends Taboola's DeeperDive infrastructure, powered by NVIDIA accelerated computing and its Realize demand network, to third-party AI platforms that have no existing publisher relationship {Taboola turns 7 million DeeperDive users into ad inventory for any chatbot}.

For AI builders, this creates a potential zero-integration revenue layer for conversational products: user queries themselves become targeting signals, and Taboola's existing advertiser demand handles the rest.

What happened

Taboola (Nasdaq: TBLA) originally launched DeeperDive in June 2025 as a publisher-embedded generative AI answer engine that draws responses exclusively from a publisher's own editorial archive. By April 2026, DeeperDive reached nearly 7 million monthly active users and expanded into six additional languages beyond English, processing tens of millions of AI-powered answers per month across thousands of publisher sites including USA TODAY Network, Bangkok Post, HuffPost UK, TIME, Weather Channel Digital, Gannett, Slate, NBC News, and Yahoo.

The monetization stack Taboola is now exporting rests on three components: large language models for generating responses, retrieval systems for pulling in supporting content, and Taboola's intent graph, which maps what a user is likely to want next based on the specific query language used {Taboola turns 7 million DeeperDive users into ad inventory for any chatbot}. The entire pipeline runs on NVIDIA accelerated computing to maintain low latency and high throughput as query volumes scale.

The commercial logic is straightforward: if a user asks about buying a home, the AI platform can surface a relevant mortgage ad using the query itself as the targeting signal, without manual campaign setup or keyword bidding. Publishers integrating DeeperDive receive a share of the ad revenue generated through it.

Why AI builders should care

This announcement directly addresses a structural problem in the generative AI industry: how to sustain products with strong user adoption when consumers are unwilling to pay for every AI service they use. Taboola CEO Adam Singolda framed the move as building the "economic layer of the AI internet."

For AI builders, three implications stand out:

• No-publisher-relationship integration: The monetization layer can be integrated by any AI platform with no publisher relationship required, which means general-purpose chatbots, agent frameworks, and virtual assistants can all use it.
• Existing advertiser demand: Taboola's Realize platform has direct relationships with tens of thousands of advertisers, giving AI products access to a demand base that would otherwise take years to build independently.
• NVIDIA-backed scalability: The infrastructure already handles millions of monthly AI interactions, and the same GPU-driven inference architecture is being extended to third-party platforms to keep latency acceptable.

Taboola's own engagement data supports the viability of the model: approximately 50% of DeeperDive user questions concern the last 24 hours of news, entertainment, and sports, and integrating publishers have seen engagement rates of up to 17%, which Taboola notes is far above historical recirculation rates in the low single digits.

Practical implications

For development teams building AI-powered products, integrating Taboola's monetization layer means adding a JavaScript snippet or API call that enables ad insertion into AI-generated response pages. The query itself becomes the targeting signal, so no separate campaign setup or keyword bidding is required from the integrating platform.

Hardware dependencies are a key consideration. Taboola relies on NVIDIA accelerated computing for inference, and the cost per query must stay low enough to leave a viable margin once revenue is split among publishers, advertisers, and Taboola itself. Extending this model to general-purpose AI platforms multiplying query volume will place additional weight on the GPU infrastructure.

Publishers using DeeperDive reported that the tool draws from more than 9,000 publisher partners globally, and the first-party behavioral data from those partners feeds Taboola's intent graph. Whether that depth of signal transfers cleanly to a third-party AI platform with no publisher relationship, one that answers questions using a general web index rather than a single trusted archive, is a technical distinction that remains unresolved.

Caveats

The evidence for this announcement comes primarily from Taboola's own press communications and industry coverage, which may reflect company-prioritized messaging and selective milestones. Independent verification of DeeperDive's 7 million MAUs and 17% engagement rate has not been published by a third party.

The key open question is whether the same monetization model will preserve user trust and engagement when applied to non-publisher AI apps that pull from broader web indexes rather than a single publisher's editorial archive. DeeperDive's original design draws on trusted editorial content, and it is uncertain whether users will accept ads in responses from general-purpose chatbots in the same way.

Additionally, the announcement does not specify pricing or revenue share terms for third-party AI platforms integrating the system.