Your AI gives you the same advice it gives everyone else. Here’s how I feed Claude the newsletters, podcasts and expert threads it can’t see, so it gives me answers my competitors don’t get.
Most people don’t know where their AI gets its information. It matters more than you think.
ChatGPT and Claude learned from billions of pages of the public internet, like Wikipedia, news sites, blogs and forums. When they need something newer, they search the web and read the top results.
And one site shows up in those answers more than almost any other: Reddit. PR firm 5W combined six studies of AI search answers and found Reddit made up roughly 40% of citations across AI engines, more than any other site. (5W, 2026) The numbers swing fast. One study saw ChatGPT’s Reddit citations drop from about 60% to 10% of answers in about six weeks in 2025. (Semrush)
Reddit also licenses its data to AI companies. Google reportedly pays about $60 million a year, and OpenAI signed a deal in 2024 with undisclosed terms. (Engadget)
The Problem
Everyone asking AI for advice is getting the same internet consensus. The newsletters you pay for, the podcasts you love and the threads from real experts mostly aren’t in there. That’s the layer where the real insight lives.
Build a private knowledge file: the best sources in your field, collected in one place Claude reads before it answers. Pick 3 to 5 types of sources, like a handful of newsletters you subscribe to, the podcasts you listen to every week and the 10 to 15 people on X who actually know your space.
There are two ways to set it up.
01
Make a Claude Project
In Claude, click Projects, then + New Project, and name it after your field.
02
Add your sources
Upload transcripts, saved newsletters and notes to the Project’s knowledge with the + on the right. You can also connect Google Drive and add Google Docs, which stay synced as you update them. On paid plans, Claude can handle up to 10x more project knowledge than usual.
03
Tell Claude to use it
Click Set project instructions, paste this and save.
This is what I run. It collects everything automatically every week.
01
Collect with Apify
Apify has ready-made tools (called actors) that pull Substack posts, YouTube and podcast transcripts and X threads. Set one up per source and put it on a weekly schedule. They’re community-built and priced per use, and they only get what you can already access, like free posts or paid newsletters you subscribe to. Podcasts work when the show has a public feed. Spotify exclusives don’t.
02
Save it into a folder
Apify stores results in its cloud, so add one step that saves them into a folder on your computer called knowledge, with a subfolder for each source type. Claude Code can write that small script for you.
03
Point Claude Code at it
Open Claude Code in that project and paste this. It sets up a CLAUDE.md file, which Claude Code reads at the start of every session.
The Edge
Everyone has the same Claude. Three months in, yours has read the newsletters, podcasts and threads the best people in your field publish. Same model, very different answers.