USA Today, under its parent company Gannett, has launched a generative AI-chatbot named DeeperDive in collaboration with ad-tech firm Taboola, aiming to stem declining traffic caused by AI summaries like Google‘s AI Overview that bypass original publishers. DeeperDive, announced at the WIRED AI Power Summit, lets readers ask questions, get concise, citation-based answers rooted in USA Today’s own journalism across its 220+ publications, and receive suggested related content — excluding opinion pieces to maintain factual integrity. The tool replaces traditional search boxes with prompts (e.g. “How does Trump’s Fed policy affect the economy?”), uses fine-tuned open source models plus data aggregated across Taboola’s network (over 600 million daily users), and includes safeguards like avoiding outputs when sources conflict. CEO Mike Reed says DeeperDive is not just a reader engagement move but part of a broader strategy to reclaim audience, better understand reader interests, and open new revenue streams — potentially via shopping-recommendation tools built into future versions.
Sources: Wired, WebPro News
Key Takeaways
– DeeperDive is USA Today / Gannett’s counter-move to reduced web traffic resulting from AI tools summarizing news (like Google’s AI Overviews), by keeping readers on site via an AI answer engine anchored in their own journalism.
– The chatbot is built for factual reliability: it sources responses only from USA Today’s verified news content (excluding opinion), provides sentence-level citations, and has a policy to avoid producing answers when source material conflicts.
– Beyond engagement, USA Today sees DeeperDive as a vehicle for monetization (via ads, recommendations, possibly shopping tools) and for better capturing reader behavior and interests, which could influence revenue and strategy going forward.
In-Depth
USA Today’s embrace of generative AI via DeeperDive is a clear signal that traditional media sees no choice but to adapt in a fast-changing digital ecosystem. The challenge is stark: search engines and AI tools are increasingly answering user queries directly (for example via Google’s AI Overviews), meaning readers often get what they need without ever visiting the original news source. That strikes at the heart of publishers’ business models — particularly those dependent on traffic, ad impressions, or subscriptions tied to reader engagement.
DeeperDive works as an answer engine built around USA Today’s own journalism, deliberately excluding opinion content, and leaning on strong citation practices. Such a model tries to preserve trust and accuracy — two areas where generative AI tools have often been critiqued. The system replaces the search box with more guided prompts, letting readers ask substantive questions, and then delivers short, factual answers plus links to related stories across Gannett’s network. The partnership with Taboola gives DeeperDive access to large-scale user behavior data (from 600+ million daily readers across many publishers), which can inform both content recommendation and reader interest insights.
But deploying AI in this space is not without risk. Publishers must ensure that the AI does not mislead, that it does not propagate errors, and that editorial integrity is kept intact. Excluding opinion content is one safeguard; another is refusing to generate responses when conflicting sources are detected. On the business side, success will depend on whether DeeperDive can meaningfully retain readers (keeping them on site longer, reducing bounce, increasing repeat visits), and whether that engagement can be monetized — whether via ads, shopping integrations, or other services. For USA Today and Gannett, DeeperDive isn’t just a new feature — it’s a strategic bet: that ownership over content + AI tools + trusted journalism can counterbalance the threat from big-tech AI aggregators. If this works, it could reshape how media companies think of search, discovery, and audience relationships.