The Maze: Meta, Facebook’s owner, is rolling out a way to tell the social network what interests you in ordinary words. The controls affect recommendations in Feed, its stream of posts, and Reels, its short videos. Meta has not named rollout countries, languages or a completion date.
Source: Meta. Meta’s product demonstration shows the Your algorithm topic lists and add/edit controls.
The Details:
Tell Facebook directly. Open the three-dot menu on a post to find Your algorithm, a space for adding and managing topics you want to see. There is also a more specific route: tap Interested or Not interested on an individual post. Facebook may then ask what you liked or disliked, letting you explain what you want more or less of. The follow-up is conditional in the product flow; Meta does not promise a question after every tap. Both routes give viewers a way to express a preference without first generating a trail of likes or comments.
Words join the existing signals. Facebook normally infers interests from activity such as watching videos, liking posts and commenting. Its ranking systems use those signals to choose and order recommended content. Meta says large language models, software that interprets written language, now help those systems understand the context and intent of a request. The text joins Facebook’s existing understanding of the person. Meta claims recommendations adapt within the same browsing session. It supplies no independent timing test, signal weights or measured improvement in audience satisfaction. A faster response is the stated capability; a better commercial outcome remains unproven.
The limits differ across Meta’s apps. In February, Meta introduced Dear Algo on Threads, its text-conversation app. People could publish a request for more or less of a topic, changing recommendations for three days. Facebook’s announcement instead describes a menu-based control and leaves the duration unstated. The older Threads rule cannot fill that gap. Nor does the Facebook release explain whether written preferences feed advertising systems, how the text is retained, or whether advertisers can access it. PPC Land flags these omissions alongside the potential effect on creators, publishers and brands whose posts depend on unpaid recommendations.
MarketMaze Take: Facebook is adding a clearer expression of what a viewer wants to its distribution machinery. That matters because watching something and wanting more of it are different signals. The business opportunity is a closer match between content and stated interest, but Meta has not shown how often people will use the control.
Relevance gets another route in. For businesses anywhere that depend on social discovery, the useful distinction is between attention a post attracts and an interest someone deliberately states. Facebook’s combination of text and behavior makes that distinction operational. A viewer can ask for a different topic before building a new viewing history. Brands should read this as a change to the conditions for earning unpaid exposure, with adoption and signal weighting still unknown.
An advertising lever has not been established. A request about recommended posts does not, by itself, reveal how paid advertisements will be selected. The release offers no targeting control, sales test or return-on-advertising result for businesses. Treating the feature as a campaign advantage would skip the missing evidence. The concrete change sits with the viewer’s recommendations; its value to a merchant depends on effects Meta has yet to disclose.
Sources: Meta, PPC Land, Meta Threads
