Your Algorithm: What Viewer-Edited Feeds Mean for Creators

The feed is no longer just inferred, it's declared

On October 13, 2026, Facebook launched its own version of a feature called Your Algorithm, first built for Instagram, letting people influence their Feed and Reels by typing out what they want more or less of in plain language. Nine days later, CNN confirmed this isn't a one-app experiment. Instagram, Facebook, YouTube, TikTok, and smaller platforms like Divine are all rolling out ways for users to directly shape their main feed or build custom ones around specific topics.

Instagram VP of Product Tessa Lyons described the shift plainly: the tools now let people use human language to say 'I want to see more of this topic or less of this topic.' That's a change from relying only on passive signals like likes, follows, dwell time, and taps. Those signals still matter, but now they sit alongside an explicit, user-written preference list that the platform treats as a stronger signal of intent.

For creators, this changes what 'getting the algorithm to like you' actually means. It's not only about keeping people watching. It's about whether a viewer, at some point, chooses to keep your content's topic on their list or quietly strikes it off.

This is bigger than Instagram

The CNN feature and related coverage make clear this is a cross-platform pattern, not an Instagram quirk. Threads has a feature called Dear Algo. TikTok users have had Manage Topics sliders. Streaming services like Spotify are adopting similar prompted-preference tools. Facebook's October 13 launch is the clearest sign yet that this is becoming standard infrastructure across major platforms rather than a one-off experiment.

The practical implication: if you distribute the same show across Instagram, Facebook, TikTok, and YouTube, you're likely to face some version of this topic-declaration mechanic on every platform within the next year, even though each platform's exact implementation will differ. Treating this as an 'Instagram thing' to deal with later is probably the wrong read.

Why this especially matters for AI-generated and character-based series

The brief behind this shift points at a specific mechanism worth understanding. When a user tells the app 'more of this topic,' that's a stronger, more deliberate signal than anything inferred from behavior, because no amount of passive watching can fake explicit intent. If your show has a clearly defined topic and people genuinely want it, you now have a direct lever that didn't exist before.

The flip side is the real risk for a lot of short-form creators, especially those running AI-generated series, faceless channels, or recurring-character formats: these shows often read as a loose, shifting mix of content to a human viewer, which is exactly the kind of thing that's easy to switch off. Instagram's system is described as a classification tool that decides, often silently, whether content shows up in someone's discovery feed. If the AI can't figure out what your account is actually about, it won't get filed under a topic that stays on anyone's list.

This creates a specific failure mode worth naming: a slow, silent fade from discovery that looks and feels like an ordinary reach drop, but is actually a classification problem. A creator might spend weeks trying to fix engagement when the real issue is that their content was never legible enough to earn a stable topic slot in the first place.

Make your premise machine-legible, not just human-legible

The practical shift here is straightforward to state and harder to execute consistently: topic classification reads text. Bio, captions, alt text, and closed captions now carry the discovery role that hashtags used to play. If your recurring characters, setting, or premise aren't spelled out clearly in the text that surrounds your video, you're relying on a classifier to guess, and a guess that's wrong means your content never reaches the people who'd actually keep it on their list.

This matters more, not less, for AI-generated and faceless series, because these formats often lean on visual or tonal consistency that a human viewer picks up on instantly but a text-based classifier may miss entirely. A recurring character with a distinct personality needs to be named and described in writing, consistently, episode after episode, not just shown.

• Name your recurring characters and settings explicitly in captions, not just in dialogue or visuals
• Write alt text that describes what the series is about, not just what's literally in the frame
• Use closed captions as a discovery surface, since they're now read for topic signal
• Keep your bio focused on one clear, recognizable premise rather than a broad description of 'content'
• Avoid loosely related or shifting topics across episodes, since that mix is what gets pruned first

Niche clarity is now a survival requirement

One point from the research is worth taking seriously rather than treating as generic advice: if your content doesn't fit a clear, recognizable topic, it can become invisible the moment someone removes a loosely related category from their preferences. That's a different kind of risk than a bad week of engagement. It's a structural one, where an entire channel's visibility depends on whether it was ever filed correctly in the first place.

This doesn't mean every series needs to narrow itself into a single rigid topic forever. It means the premise needs to be stated plainly enough, consistently enough, that a classifier and a viewer arrive at the same understanding of what the show is. Ambiguity that used to read as 'creative range' to a human now reads as 'unclassifiable' to a machine, and unclassifiable content doesn't get a fair shot at staying on anyone's keep list.

If you're producing a multi-episode AI-generated series, this is also a production planning question, not just a captioning afterthought. Tools like Episodo that help creators plan and produce structured, recurring-character series can make it easier to keep premise and character details consistent across episodes, which in turn gives the text layer around each video something stable and specific to describe.