Meta published an engineering account on November 10, 2026 explaining how its Your Algorithm controls actually change what people see in Facebook Feed and Reels. For anyone whose reach depends on recommendations, the detail is more useful than the announcement: it shows where a viewer’s stated preference enters the ranking system, and where it does not.
What the feature is
Your Algorithm is a page with three parts: a one-sentence summary of recent viewing, an Interested In list pre-filled with inferred topics, and a Not Interested In list built from topics a person marked, hid or dismissed. People can add interests in their own words, move them between lists, or delete them. Meta says every change applies to both Feed and Reels, and continues to update as the person gives feedback on individual posts.
How a preference travels through the system
The engineering post describes four stages, and each one matters for reach.
- Profile generation. Large language models turn engagement signals into named interests, refreshed regularly, which is why the list is readable instead of a set of opaque IDs.
- Preference management. Each edit is validated and enriched, then written to a single store that both Feed and Reels read. That shared store is the reason a change in one surface shows up in the other.
- Ranking. Interested In topics get dedicated retrieval and ranking promotion. Not Interested In topics pass through embedding filters, demotion rules and a fine-tuned reranker.
- Delivery optimisation. Extra content fetches and cache invalidation help a change take effect sooner, instead of waiting for the next natural refresh.
The part creators will care about most
Meta says a newly added interest gets extra weight for a limited period so the person notices the change, after which it is weighted by engagement like any other topic. It also says the system spaces out exposure across topics, with daily caps, to stop one subject from taking over a feed.
Two asymmetries follow. Positive interests are a soft signal: they raise the odds of retrieval and ranking, but they do not guarantee distribution. Negative signals are harder. Quality requests, such as asking for less AI-generated or sexually suggestive content, map to dedicated demotion signals rather than topic matching. And when someone asks to see less from a named creator, that creator’s content is filtered out directly.
What this means in practice
For a brand or creator, three adjustments are worth making now. First, treat AI-generated visuals as a distribution risk rather than a neutral production choice, because viewers can now ask the system to reduce that class of content. Second, stop assuming a follower who hides your posts still sees them in Reels; the direct filter applies across surfaces. Third, watch topic labels rather than keywords, since the system matches on inferred interests and embeddings, not on exact phrases in a caption.
There is also a reporting consequence. If reach drops, the honest first question is whether your content is being classified into a topic your audience has muted, not whether the algorithm changed overnight.
What Meta did not promise
The engineering post describes mechanisms, not outcomes. It gives no measured accuracy figure for how well stated preferences match delivered content, no regional timetable for availability, and no commitment that unwanted topics disappear entirely. A separate LLM compares outputs offline, and Meta acknowledges ambiguous requests remain an area for improvement. Social Media Today’s summary of the post makes the same point in plain terms: the controls give people more input, they do not hand over the ranking.
For the commercial side of the same shift, our piece on TikTok’s agentic commerce push covers how the other major platform is changing the path from feed to checkout.

Sources
- Engineering at Meta: Your Algorithm, giving people more ways to personalise Facebook recommendations
- Social Media Today: Meta publishes explainer on Your Algorithm controls
- Meta Newsroom





