We listen to how nonprofits ask.

• 392organizations in 33 countries
• 332emails kept whole since September 18, 2026
• 505news articles from 35 sources
• 102,000+videos listed from 158 YouTube channels
• 7,200+podcast episodes listed from 27 shows

1. 01

   Collect.

   Every organization gets its own email address, so we know exactly who sent each message. If mail from one organization arrives at an address only ever given to another, we have found list-sharing that nobody publishes. Around the mail we collect the news, each organization’s YouTube channel and the sector’s podcasts, every day. December 2026 can only be collected in December 2026. Nobody can go back for it later, including us.

2. 02

   Read.

   Every email is kept whole, with its headers and its links. Then software reads it, and AI models do the first pass: who is speaking, what they want, why it matters, how they say it, and what they ask for. Articles, video transcripts and podcast episodes get the same questions. We say which parts a machine read, and how, on the Method page.

3. 03

   AI Research.

   This is where the record turns into answers. AI models put thousands of small readings side by side, from every email, article, video and episode, until patterns no single inbox could show come to the surface: the ask that arrives on day two, the phrase several charities adopt in the same fortnight, the story a sector tells on video while it says something else by email. People check what the machines find. What holds up goes into the blog and Giving Radar Daily, every number with its sample size and its date range.

How we collect and read →