How to Take Control of Your News Algorithm (Instead of Being Controlled by It)

Every news feed you scroll is ranked by an algorithm — and almost none of them work for you. Engagement-optimized feeds surface whatever keeps you scrolling. Chronological feeds bury the important under the recent. Either way, the ranking is a black box: you can add and remove sources, but the thing that decides what rises to the top is not something you can touch.

Feedpik takes the opposite position: the ranking algorithm belongs to the reader. Here's what that means in practice, and how to use it.

Start with sources you choose

Feedpik speaks RSS, Atom, and OPML, so your feed is built from sources you picked — tech blogs, Substack newsletters, podcasts, subreddits, Mastodon and Bluesky accounts, YouTube channels. Paste a site address into Discover Sources and Feedpik looks for its feed; a direct RSS or Atom URL always works. If you're coming from another reader, one OPML import brings everything over.

That's the foundation: no algorithm can be yours if the inputs aren't. It is also the easy half, and every reader can do it. The hard half is what happens next — something decides the order, and in most products that something is not available to you.

Tune it by talking to it

Click AI Assistant and tell it what you want in plain language:

• "Less politics."
• "More history stories."
• "Why am I seeing so much Apple coverage?"

The assistant works out what that means for your algorithm — topics to block, interests to weight up, how hard to favour fresh items — and comes back with the concrete edits it wants to make. Nothing applies until you accept them, so a request it read too broadly is something you catch before your feed changes. Every accepted change becomes a new numbered version of your policy.

Prefer sliders and forms? Everything the assistant can do also lives in Algorithm & Agents → Recommendation Settings, the ground truth for your ranking policy.

Roll back anything

Because every change is a version, there's no fear in experimenting. Cranked freshness too high and your feed turned into a firehose? Open your policy history and roll back to the version that worked. The audit trail records what changed, when, and who (or what) changed it — including changes made by an AI agent you've connected.

Save whole configurations as personas

A ranking policy tuned for stock trading makes a lousy physics research. Feedpik lets you save each configuration as a persona and switch between them in one click.

The takeaway

"Controlling your algorithm" isn't about becoming a machine-learning engineer, and it isn't about being shown the maths. It's three capabilities no mainstream feed gives you:

1. Authorship — change the ranking by describing what you want, in a sentence.
2. Consent — see the concrete edits that sentence produced, and approve them before they touch your feed.
3. Reversibility — every accepted change is a version; going back is one action.

Start free with the full reader, import your OPML, and do the thing no other reader lets you do: tell your feed what's wrong with it, and watch it change.