Filter out opinion‑centric news from Discover/News feed in app

Feature Requests
,
1

:rocket: Feature Request
Add a content filter for opinion‑centric items in the Discover/News feed in the Perplexity app, allowing users to reduce or hide stories whose main content is “X said Y about Z”.

:magnifying_glass_tilted_left: Problem Statement
In the iOS app (Discover/News feed), a large proportion of items are about what a single person said or thinks (e.g. “Politician X commented on…”, “Elon Musk said that…”).

As a user who primarily seeks evidence‑based information (data, studies, reports, analyses), I’m not interested in consuming quotes and personal commentary — even if they come from Nobel laureates, and especially when they come from politicians or celebrities.

Currently, these opinion‑centric items take up a significant share of the feed, which reduces signal‑to‑noise for users who care more about facts than about who said what.

:light_bulb: Proposed Solution

Add a content‑type preference or filter in Discover/News, for example:

  • A toggle or slider like:
    • “Fewer single‑person opinions / More data‑driven content”
  • Or explicit checkboxes:
    • Reduce stories whose main focus is “X said Y”
    • Prioritize reports, summaries, datasets, multi‑source analyses

Intended behaviour (examples):

  • De‑prioritize or hide items where the headline and body are primarily about a single individual’s opinion, quote, or tweet.
  • Prioritize items that aggregate multiple sources, show data or empirical findings, or summarize events rather than amplify one person’s statement.
  • Optionally, allow per‑topic overrides (e.g. still show expert commentary in areas I follow, but reduce political soundbites).

This could be exposed in Settings → Discover / News preferences, or as part of a more granular “Content preferences” panel.

:pushpin: API Impact

From the outside it looks like this primarily concerns search / retrieval / ranking filters for the feed generation layer, rather than the chat completions API itself.

  • API component: internal ranking / search filters used to build the Discover/News feed.
  • Model‑specific? Potentially relevant to any model used for news ranking or classification (e.g. Sonar Deep Research when curating sources).
  • Parameters: This could correspond to an internal boolean or categorical parameter, e.g.:
    • content_style: "factual_only" | "mixed" | "opinion_friendly"
    • or a numeric preference like opinion_weight in the ranking pipeline.

No changes are needed for external API users unless you want to expose similar controls on the public API later.

:counterclockwise_arrows_button: Alternatives Considered

Workarounds I’ve tried / considered:

  • Manually ignoring or scrolling past opinion‑centric items in the feed.
    • This may slightly adjust personalization over time, but does not reliably remove this type of content.
  • Using direct chat queries instead of the feed, with prompts like “summarize the facts, ignore quotes and commentary”.
    • This works for ad‑hoc queries, but does not solve the problem of the push / ambient Discover/News experience, which is where I’d like better signal.
  • Relying on topic preferences only.
    • Topic filters (e.g. tech, science) still surface many “X said Y” headlines within those domains.

These workarounds are partial and time‑consuming; they don’t give the user a clear, controllable way to say “I care about facts and multi‑source context, not isolated opinions”.

:paperclip: Additional Context

  • Use case: I’m an academic / researcher‑type user who uses Perplexity to stay up to date via the Discover/News feed, but wants to maximize fact‑rich, multi‑source content and minimize “soundbite journalism” centered on single individuals.
  • UX suggestion: even a simple first version — a single toggle like “Reduce opinion‑centric stories” — would already be very helpful. Over time, this could evolve into a more general content‑preference system (facts vs opinions, primary sources vs commentary, etc.).

Thank you for considering this request — I think it would make the Discover experience significantly more useful for users who prioritize evidence‑based information over individual commentary.

1 Like