AI-Powered Personalization in News Delivery
Signals That Shape Your Feed
We look at dwell time, scroll depth, article completions, saves, and shares, alongside recency and diversity of sources. These signals help the model predict stories you’ll value next. Want sharper recommendations? Adjust your topic preferences and follow trusted outlets.
From Cold Start to Familiar Companion
On day one, the system asks a few interest questions and leans on broad, high-quality editorial picks. As you read and react, it learns your rhythms and refines suggestions. Jump in, rate a few topics, and watch your feed evolve meaningfully within days.
Serendipity by Design
To avoid filter bubbles, we actively introduce carefully chosen out-of-comfort-zone stories. These explorations stretch perspectives without overwhelming your feed. If a discovery resonates, follow that topic or publication—your choices help the model broaden your horizons intelligently.
Trust, Transparency, and Control
Every recommended story includes a short reason, such as matching your followed topics, timeliness, or local relevance. We keep language human, not technical. If an explanation ever feels opaque, let us know so we can make it clearer and more useful.
Inside the Engine: Real-Time Personalization at Scale
Reading and interaction events stream into feature stores, where candidate articles are retrieved by topic, locality, and freshness. A ranking model scores these candidates against your profile. Guardrails then filter low-quality items before the final, diverse slate is delivered instantly.
Inside the Engine: Real-Time Personalization at Scale
We blend short-term behavior with longer-term interests using sequence models and embeddings that capture semantics beyond keywords. Context like time of day and device helps match format and depth. If you prefer morning briefings, the model prioritizes concise over deep dives at dawn.
Editors + Algorithms: A Collaborative Newsroom
Critical topics—elections, public health, emergencies—receive elevated visibility and credible-source requirements. Personalization never buries need-to-know news. If you think something vital is missing, flag it. Your reports guide editors and keep essential coverage prominent for every reader, every time.
Editors craft collections and context cards that algorithms can boost when relevant to your interests. This hybrid approach preserves nuance and accountability while scaling delivery. Tell us which curated explainers helped you most; we’ll expand the formats that truly clarify complex stories.
A small coastal newsroom used our tools to surface hyperlocal flood updates during a storm. Engagement doubled, but more importantly, residents found evacuation guidance quickly. Share your own experience of timely alerts—your stories help shape our playbook for public-interest personalization.
Multimodal Summaries and Audio Briefings
Concise, citation-backed summaries help you scan confidently, while optional audio briefings fit commutes or workouts. If you save longer features for evenings, we’ll queue them there. Tell us which voices, speeds, or styles you prefer so we can refine listening experiences.
Locality-Aware Alerts
Get smarter notifications for school closures, transit disruptions, or neighborhood safety updates without constant pings. You choose alert categories, quiet hours, and radius. If you move or travel, update your location preferences so alerts remain relevant without becoming distracting noise.
Deep Dives vs. Quick Takes
Some moments call for a 60-second explainer; others deserve a weekend-long read. We estimate reading time and recommend accordingly. Use “Save for later” to build a queue. Tell us when we misjudge your mood, so the system learns your pacing better.
Measuring What Matters, Together
Beyond click-throughs, we track completion rates, source diversity, topic breadth, satisfaction, and long-term retention. We also monitor overload and fatigue. Tell us when the balance feels off—your feedback ensures metrics reward depth, learning, and trust, not empty distraction.
Measuring What Matters, Together
We run careful A/B tests with caps to avoid over-testing any reader. You can opt in to early features or stick with stable experiences. Share how variants affect your experience; your comments directly influence which changes we adopt or retire.