Google News Innovation Forum
Pitched a project that received funding from Google News
Led technical implementation spanning four engineering teams
In 2025, the Washington Post was invited to join the Google News Innovation Forum, a funding opportunity where publishers pitch AI-driven products that create personalized experiences for our audiences.
I pitched, then designed the technical approach for an interactive newsletter that provides a snapshot of a recommended story. Every send includes:
What to Know: a concise, editor-reviewed AI summary of the top recommended story tailored to you
What Our Readers Are Saying: featured community comments that add depth and perspective
Ask The Post: a direct way for readers to ask any question and get a fast, AI-powered response (via text or voice), sourced solely from our journalism
More For You: nine personalized article recommendations on various topics
This newsletter is sent to hundreds of thousands of users. Our challenge was to automate personalization at scale while maintaining the quality of the journalism. We knew readers wanted articles summarized in their inbox and to see community discussion about a topic. AI can help produce this type of content at scale, but we set clear ethical guardrails–no content will be sent to readers without review by a human editor.
We had to tap into existing editorial workflows ot make this work. Article summaries, drafted by AI in the Post’s CMS, save metadata when journalists edit and approve the summary. Comment editors moderate and select featured comments using an internal tool that flags comments with AI. Together, these tools ensure our most-read articles have approved summaries and moderated comments.
The automated workflow builds the newsletter starting with the Post’s recommendation algorithm. It selects articles based on reading history, topics and authors a reader follows. An article is only used if it has a human-reviewed summary and comments featured by our comments editors. If no articles meet our requirements, we don’t jam in content that hasn’t been reviewed. We fall back to a version with just headlines.