Taboola turns DeeperDive into an ad network for AI apps and agents
Taboola just made a pretty significant move in the digital advertising market: the company opened up the monetization infrastructure of its DeeperDive product to any AI application that wants to generate revenue with native ads inside conversational interfaces. The announcement was made in June 2026 and marks a major shift in the company’s strategy, which now aims to position itself as the economic layer of what CEO Adam Singolda calls the AI internet.
In practice, this means chatbots, virtual assistants, and LLM-based services can now connect directly to Taboola’s advertiser inventory without having to build their own advertising infrastructure from scratch. It sounds simple when you read it like that, but the impact on the AI ecosystem is much bigger than it seems. We are talking about an open door for any developer or company running language models to gain access to a well-established advertising network, with the entire technical layer already built and tested at global scale.
And this is no small move. The company already has a base of roughly 20,000 performance advertisers and integrations with approximately 9,000 publishers around the world, which puts DeeperDive in a very different position from the generic solutions floating around out there. None of the direct competitors come close to that volume of established commercial relationships, which turns this infrastructure opening into something with real market weight — not just another product announcement.
The timing makes perfect sense too: keeping an LLM running at scale is expensive, and subscriptions alone don’t balance the books for most AI companies. Native advertising inside conversations is increasingly becoming the most viable path forward, and Taboola wants to be the infrastructure behind it. 🎯
What is DeeperDive and how does it work inside conversations
DeeperDive is described by Taboola as an answer engine that sits embedded within publisher pages and transforms long-form editorial content into an interactive chat experience. Users can ask follow-up questions about an article, explore topics in more depth, or get summaries, with responses generated by a combination of LLMs, information retrieval systems, and Taboola’s proprietary intent graph.
The core idea was never to interrupt the conversation but rather to complement it in a way that makes the ad feel natural within the flow of messages between the user and the AI system. That is what separates native advertising from traditional advertising: it does not show up as a visual block disconnected from context but as a recommendation that feels natural within the experience the user is already having.
In the product’s current architecture, when a user interacts with a chatbot or assistant that has integrated DeeperDive, the system analyzes the conversation context in real time and, based on that, decides which sponsored content can be inserted in a relevant way. If someone asks about buying a house, for example, the system might present mortgage or home financing offers as sponsored results right within the conversation flow. The process uses semantic signals generated by the LLM itself to understand user intent, not just isolated keywords. This represents a significant evolution from the traditional contextual advertising model, which relied heavily on matching specific terms and could not capture nuances of meaning the way a language model can.
Singolda actually made a clear statement during the announcement: he believes traditional display formats are on borrowed time. Native ads that integrate visually and behaviorally into AI-generated responses would be, in his view, the future of online advertising.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
By opening this infrastructure to third parties, Taboola transforms DeeperDive from a closed product into a platform. Any AI-based application can now plug in this monetization layer without having to negotiate individual contracts with advertisers, without building auction systems, and without assembling an entire team dedicated to ad operations. All that work is already done, and the developer just needs to integrate the API and configure the display rules that make sense for their product experience.
The business model behind the AI ad network
The business model powering this new ad network for LLMs mirrors Taboola’s traditional marketplace. Advertisers pay on a cost-per-click basis, and the revenue generated is split with distribution partners. In an interview with Digiday, Singolda detailed the revenue-sharing structure: Taboola typically retains about 35% of the ad spend, while the remaining 65% goes to publishers or distribution partners. The exact terms can vary depending on the size and duration of each specific deal.
This is a point worth paying attention to because it puts Taboola in a different position from many traditional ad networks, where the split tends to be less transparent and, in many cases, far less favorable for the publisher or partner side.
The company also indicated that publishers already using DeeperDive are seeing a monetization increase of approximately 10% on top of existing revenues from Taboola. However, Singolda made sure to clarify to Digiday that the company is not offering minimum revenue guarantees to publishers adopting DeeperDive, which means results can vary significantly depending on each partner’s audience profile and traffic volume.
From an ad performance standpoint, the numbers are interesting. Singolda told Digiday that, out of all Taboola’s ad formats, DeeperDive already has the highest conversion rate for advertisers. This makes sense when you think about the context: a user who is actively asking questions about a specific topic inside an AI interface demonstrates a much clearer intent than someone who is simply scrolling through a news page. The logic is similar to search ads, which have historically always had the best conversion rates in the digital advertising ecosystem. 📈
Why LLM monetization became urgent right now
There is a very concrete financial reality pushing the AI market in this direction. Running an LLM in production with high availability, acceptable latency, and the capacity to serve millions of simultaneous users is extremely expensive. Infrastructure costs for compute — especially GPUs and high-performance storage — remain steep even with the price drops that have happened over the past few years. And when you add development costs, security, regulatory compliance, and ongoing model maintenance, the number gets even more staggering.
The subscription model was the first obvious answer, and it worked well for a few big players. But for the vast majority of companies building LLM-based products, subscriptions alone do not solve the problem sustainably. As Singolda himself argued, consumers are unlikely to maintain dozens of separate subscriptions for different AI tools. The conversion rate from free to paid users is historically low in software products, and conversational AI tools are no exception. This creates a massive gap between the cost of serving free users and the revenue generated by subscribers, and that gap needs to be covered somehow.
This is exactly where native ads inside conversations come in as a viable — and in many cases complementary — alternative to the subscription model. Monetization through advertising allows AI products to maintain a robust free tier sustained by ad revenue while offering premium ad-free experiences for those willing to pay. This hybrid model has already worked very well in other verticals like music and video streaming, and the AI market is clearly headed in the same direction. Taboola recognizes this window of opportunity and is positioning itself to be the default infrastructure for this transition. 🚀
What changes for AI product developers
For developers and companies already operating or building LLM-based products, the opening of DeeperDive represents a very practical shift when it comes to business viability. Before this move by Taboola, the options for monetizing a conversational interface through advertising were quite limited: you either tried to adapt traditional display solutions that clearly were not built for that context, or you entered long and complex negotiations with ad networks that had no infrastructure optimized for the conversation format. In both cases, the result was usually a degraded user experience and pretty modest revenue.
With DeeperDive as an option, the developer gets direct access to an inventory of 20,000 advertisers already accustomed to native and performance advertising formats — exactly the type of ad that makes the most sense inside a conversation. Performance advertisers are the ones who measure results by clicks, conversions, and concrete user actions, which creates a much better alignment with conversational AI contexts than traditional branding ads. When a user asks a chatbot about a specific product and receives a contextualized recommendation, the likelihood of real engagement with that ad is substantially higher than with a random display banner.
The new ad network for LLMs is open not only to publishers but to a much broader spectrum of partners. Taboola is inviting utility apps, finance tools, media apps, virtual assistants, and any LLM-powered service to integrate its ad API and place native units within their own chats and interfaces. This considerably expands the platform’s potential reach and diversifies inventory sources for advertisers.
On top of that, the technical integration with DeeperDive was designed to be compatible with the most common LLM architectures on the market, which significantly reduces implementation time. Taboola built this infrastructure with developers in mind who already have their models in production and need to add a monetization layer without refactoring their entire product architecture. That is a real differentiator, especially for smaller teams that cannot afford to dedicate months of engineering work to a problem that should be solved in days of integration. 💡
The evolution of targeting: from first-party data to contextual intent
One important technical aspect worth highlighting is how DeeperDive determines which ads to show each user. Currently, the system relies heavily on Taboola’s first-party audience and interest data, derived from the code-level integrations the company maintains with thousands of publisher sites. This data asset allows the platform to understand not only what the user is asking at that moment but also their long-term interests — like following sports, researching real estate, or being into tech.
But the company has already signaled that the plan is to progressively evolve toward more contextual, query-intent-based targeting. In other words, ads will increasingly resemble search-style units triggered by what the user is asking in real time, rather than relying exclusively on browsing history. This transition is relevant both from a privacy standpoint and an effectiveness standpoint, since intent demonstrated in the moment of conversation tends to be a much stronger signal than behavioral data aggregated over time.
Singolda also mentioned plans to evolve toward more agentic experiences, where the system does not just answer questions but proactively suggests plans and purchases based on the interaction context. Imagine a scenario where you are chatting with an AI assistant about trip planning and, beyond answering your questions, it builds a complete itinerary with lodging and activity suggestions — all transparently sponsored. That is the direction Taboola is aiming for in the next product development cycles.
Native ads in conversations: the user experience challenge
Every time advertising inside conversational interfaces comes up, the first concern is user experience. And it is a legitimate one. A user interacting with an LLM expects useful, honest, and contextual responses. If they sense the system is steering answers to favor advertisers at the expense of information quality, trust breaks in a way that is very hard to recover. The line between a well-placed native ad and a response that feels bought is very thin, and crossing it carries an enormous cost in terms of product reputation.
The DeeperDive approach tries to address this with a clear separation between LLM-generated content and sponsored content. Instead of blending the two so the user cannot tell the difference, the platform works with insertions that are identified as commercial recommendations, maintaining the transparency users need to keep trusting the tool. This approach aligns with regulatory guidelines being developed in several regions around the world, especially in Europe, where digital advertising transparency requirements are growing and becoming increasingly specific for new formats like AI interfaces.
From Taboola’s perspective, prioritizing this transparency makes total sense because the company’s business model depends on the long-term credibility of the native ad format. If native ads in LLMs start being associated with manipulative experiences, the entire market suffers — not just individual players. That is why the company has a structural incentive to establish healthy standards for ad placement in this new channel, and opening DeeperDive as a platform is also a way to influence how this market will develop in the coming seasons. 📊
A rival to Google and Facebook in AI advertising?
Strategically, the ad network for LLMs fits into a broader ambition that Singolda has been articulating for a while now. The idea is to position Taboola as the economic layer for everyone outside of Google and Facebook, aggregating enough intent and performance data across the open web to offer advertisers a value proposition comparable to that of the tech giants.
Taboola’s first-party integrations with approximately 9,000 publishers help build what is known as the intent graph, which tracks what users are reading and researching over time — from sports fandom to mortgage research. This data feeds the recommendation units at the end of articles and in other native placements, and now it also feeds targeting within DeeperDive’s conversational interfaces.
Singolda argued to Digiday that this gives Taboola an edge over general-purpose LLM providers that lack direct visibility into users’ broader browsing behavior. Connecting conversational intent to a broad, cross-site understanding of user interests will become increasingly important as AI agents handle a larger share of the consumer journey, from discovery all the way to the final purchase decision.
It is a bold bet, no question about it. Competing with Google and Meta in any digital advertising vertical has never been easy. But Taboola is targeting terrain these giants have not fully conquered yet: monetizing conversational AI interfaces at scale. And having the infrastructure ready before the market reaches full maturity could be the differentiator that defines who leads this segment in the years ahead.
