Taboola expands DeeperDive into an ad network for AI apps and agents

Related Insights

The Business of AI

Taboola’s next act: an AI answer engine for publishers Read More

The new offering extends the same monetization stack Taboola uses to fund DeeperDive, its AI “answer engine” embedded across major publishers, to external AI players. Generative AI companies will be able to insert sponsored recommendations directly into AI-generated responses, with Taboola handling ad sourcing, targeting, and billing.

The business model mirrors Taboola’s core marketplace, with advertisers paying on a cost-per-click basis, and subsequent revenue then shared with partners. Taboola CEO Adam Singolda further explained the proposed revenue sharing to Digiday, noting that Taboola typically retains roughly 35% of ad spend, with the remaining 65% going to publishers or distribution partners, though exact terms can vary by deal size and duration. 

Taboola has further indicated that publishers already using DeeperDive are seeing monetization lift of roughly 10% on top of existing revenues from the company, although Singolda noted to Digiday that it is not currently offering minimum revenue guarantees to publishers for DeeperDive.

Ultimately, the company is positioning the product as a turnkey way for AI companies to convert user queries into commercial opportunities without building their own ad stack. “Today’s news is about helping build the economic layer of the AI internet,” said Singolda in a press release accompanying the announcement.

How DeeperDive works

DeeperDive is described by Taboola as an “answer engine” that sits on publisher pages and turns 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 answers generated by a combination of LLMs, retrieval systems, and Taboola’s proprietary intent graph.

Taboola then brings a monetization layer, using native ad units — Singolda alluded to his opinion that traditional display ad units are on the way out — that visually and behaviorally resemble the surrounding AI-generated responses. If a user asks about buying a house, for example, the system can surface mortgage or home-finance offers as sponsored results within the conversational thread.

DeeperDive taps into Taboola’s base of roughly 20,000 performance advertisers. It currently leans heavily on Taboola’s first-party audience and interest data — derived from its code-level integrations across thousands of publisher sites — to determine which advertisers to show to which users. Over time, the company plans to shift more toward contextual and query-level intent, making the ads resemble search-style units triggered by what the user is asking in real time, and eventually toward more agentic experiences that proactively propose plans and purchases.

From an advertiser perspective, Singolda told Digiday the format is already performing at the top of Taboola’s stack. “Out of the entire Taboola universe… the number one conversion rate advertisers see is from DeeperDive already today,” Singolda told Digiday. 

Extending DeeperDive to third parties

The new LLM ad network effectively unbundles that monetization layer so that any AI application can reach Taboola’s ad demand. Rather than limiting DeeperDive’s ads to publisher-owned experiences. Furthermore, Taboola is also inviting external conversational interfaces, such as utility apps, finance tools, media apps, virtual assistants, and other LLM-powered services, to use its ad API and insert native units into their own chats.

Per Singolda, running high-quality LLMs at scale is expensive, and consumers are unlikely to maintain dozens of separate subscriptions; therefore, AI companies will have to turn to advertising and commerce for alternative revenue streams.

“The future of the internet will be increasingly conversational and agentic,” he added, noting the high-intent of AI interface is similar to a search engine — the most successful online ad environment in the history of online publishing. 

Rivals for Facebook and Google?

Strategically, the LLM ad network aligns with a broader ambition Singolda has articulated, seeking to position Taboola as “the economic layer for anyone outside of Google and Facebook,” by aggregating enough intent and performance data across the open web to offer a comparable value proposition to advertisers. 

Taboola’s first-party integrations with approximately 9,000 publishers help build its “intent graph” that tracks what users are reading and researching over time, from sports fandom to mortgage shopping, powering recommendation units at the bottom of articles and in other native placements. 

Singolda told Digiday that this gives Taboola an edge over general-purpose LLM providers that don’t have direct visibility into users’ broader browsing behavior. Ergo, connecting conversational intent to a broad, cross-site understanding of user interests will become more important as AI agents handle more of the consumer journey from discovery to decision.