Threads is about to turn a quiet user habit into a formal product feature, inviting people to talk directly to the recommendation system that shapes their feeds. Instead of endlessly toggling settings or muting accounts, users will be able to write natural language prompts that tell the app what they want to see more or less of, and the algorithm will respond in kind. The shift could redefine how social feeds feel, moving from opaque ranking formulas to something that behaves more like a conversational assistant.
That change is not arriving in a vacuum. It builds on months of experimentation inside Threads, a broader push at Meta to give people more visible control over ranking systems, and a wider tech industry trend toward interfaces that treat algorithms as partners rather than black boxes. I see this as a test of whether users actually want to manage their feeds more actively, or whether they simply want better defaults with the option to nudge the system when it drifts off course.
From quiet trend to official feature
The idea of addressing a social feed like a person did not start in a product meeting, it started with users. Since Threads launched, people have been posting open letters to their feed, asking the ranking system for everything from niche fandom content to more local news, effectively treating the recommendation engine as a character in the conversation. The company has acknowledged that Since Threads launched, it has noticed this trend of people posting to ask their Threads algorithm for things, and that observation is now shaping product design.
What began as a kind of meme has become a signal: users are not just scrolling, they are trying to negotiate with the system that decides what they see. By formalizing that behavior into a feature, Threads is essentially saying that the algorithm is no longer just a hidden engine, it is a participant in the social experience. That is a subtle but important shift, because it reframes personalization as a dialogue rather than a one-way calculation.
How “Dear Algo” is supposed to work
Threads is testing a feature called Dear Algo that turns those informal pleas into a structured way to steer the feed. Instead of hunting through menus, users will be able to write posts that begin with a clear invocation, then specify the kind of content they want to adjust, and the system will interpret that as a direct instruction. Reporting describes Dear Algo as a simple approach that can permanently transform your preferences, with Threads tests ‘Dear Algo’ framed as a test of how far this conversational model can go.
The mechanics are intentionally lightweight. Rather than adding a new settings page, Threads is testing a simple, yet brilliant, approach to feed personalization that lets people type natural language prompts inside the same interface where they already post and reply. The company is positioning this as a way to permanently transform your preferences without forcing you to like, hide, or mute dozens of individual posts, a direction underscored by reports that Threads is testing a simple conversational layer on top of its ranking system.
Talking to the algorithm in plain language
The most striking part of this experiment is how literal it is. Threads will let users speak to the algorithm in plain language, turning everyday posts into control levers for the feed. Instead of abstract sliders, people will be encouraged to write things like “Dear algo, show me more book recommendations” or “Dear algo, stop showing me basketball updates,” and the system will treat those sentences as structured feedback. That approach is captured in reporting that describes how Threads will let users speak to the algorithm using examples that sound like ordinary posts rather than commands.
There is a deliberate design choice in keeping the language casual. By asking people to start with “Dear algo” and then describe what they want, Threads is borrowing the tone of a letter or a diary entry, not a configuration screen. That framing lowers the barrier to participation, especially for users who would never touch advanced settings but are comfortable typing a sentence about their interests. It also makes the algorithm feel more like a character that can be reasoned with, which could increase engagement but also raises questions about how much agency users really have when they are effectively negotiating with a machine.
Examples that show what control might look like
The early examples of Dear Algo posts are revealing, because they show how granular this control could become. One scenario imagines a user asking, “Dear algo, show me more book recommendations,” which would signal that they want the system to prioritize reading-related content, from author accounts to review threads. Another example flips the request, with a user saying, “Dear algo, stop showing me basketball updates,” a clear attempt to prune a specific topic from the feed. These cases are not hypothetical marketing copy, they are drawn from reporting that explains how Threads will let users speak to the algorithm using exactly these kinds of prompts.
There is also a more self-referential twist: users are being encouraged to ask for more content from specific outlets or communities. One example highlights a post that reads, “Dear algo, show me more Mashable content,” which illustrates how people might use the feature to tune their feeds toward particular brands or creators they trust. That line appears in coverage that notes how Christianna Silva described Dear prompts that could shape what users see for the next three days, suggesting that some of these instructions may have time-bound effects rather than permanent changes.
Meta’s broader push for algorithmic transparency
Dear Algo is not an isolated experiment, it fits into a larger pattern of Meta trying to show that its ranking systems are not entirely opaque. Earlier work on Threads has focused on giving users more direct ways to tag and influence the algorithm, with tools that let people mark posts or topics in ways that feed back into personalization. Reporting on these efforts notes that Meta’s latest moves in social media transparency include developing a tool that lets users finally control feeds, with the explicit goal of less doomscroll and more delight.
In that context, Dear Algo looks like the next iteration of a strategy that treats user intent as a first-class signal. Instead of relying solely on passive behavior like likes and watch time, Meta is experimenting with explicit instructions that can override or refine what the system thinks you want. That is a meaningful shift in power dynamics, at least on paper, because it acknowledges that people sometimes know their own preferences better than the engagement metrics do. It also gives Meta a narrative to counter criticism that its algorithms are inscrutable, by pointing to visible tools that let users push back when the feed goes off the rails.
Manual guidance instead of mystery knobs
Under the hood, Dear Algo is part of a broader attempt to let users manually guide feed algorithms without forcing them into complex dashboards. Threads is trying out a new way to help users attune the feed algorithm to their preferences by treating certain posts as signals, whether or not the user likes, replies to, or reposts them. Coverage of this experiment explains that Threads is trying out a system where the act of addressing the algorithm directly becomes a kind of manual override layered on top of existing engagement signals.
That approach reflects a recognition that most people will never spend time fine-tuning dozens of settings, but they will happily fire off a sentence when something in their feed annoys them. By converting those sentences into structured feedback, Threads can capture intent that would otherwise be invisible, such as a desire to see more longform analysis or fewer short video clips. It also gives the company a new stream of data about what users say they want, which can be compared against what they actually engage with, a tension that will likely shape how aggressively the system honors these Dear prompts over time.
Why this feels different from past “controls”
Social platforms have long offered tools to mute words, hide posts, or switch between chronological and algorithmic feeds, but those controls often feel bolted on rather than central to the experience. Dear Algo is different because it lives inside the main posting flow, turning the act of asking for change into content itself. When I look at the examples and the way Threads is framing the feature, it is clear that the company wants people to think of the algorithm as something they can talk to, not just something that silently watches their behavior.
That distinction matters because it could change how users perceive responsibility for what shows up in their feeds. If you can write “Dear algo, stop showing me basketball updates” and still see nothing but sports, the frustration will land differently than if the system was always a black box. Conversely, if the algorithm responds quickly and visibly to these prompts, people may feel more ownership over their feeds and less resigned to whatever the ranking system serves up. The success of this experiment will hinge on whether the conversational veneer is backed by real shifts in what users see, or whether it becomes another settings gimmick that people try once and forget.
Parallels with conversational AI and privacy controls
The move to let people talk to a feed algorithm in natural language mirrors a broader trend in AI interfaces, where chat-style interactions are becoming the default way to manage complex systems. In the AI assistant world, companies are already experimenting with modes that give users more control over how their data is used, including features that treat certain conversations as off the record. One example is a capability called Temporary Chat, described as a way to handle one-off, sensitive, or private conversations that will not be used to personalize future interactions or train the underlying AI models, a design captured in reporting that notes Perhaps the most notable new feature in that context is precisely this kind of ephemeral mode.
Threads is not promising anything as strong as Temporary Chat for its feed controls, but the conceptual overlap is clear. In both cases, users are being invited to express their preferences and constraints in plain language, with the system interpreting those statements as configuration rather than just content. That convergence suggests a future where talking to algorithms, whether in a social app or an AI assistant, becomes the primary way people manage digital experiences, raising new questions about how transparent these systems really are about what they remember, what they forget, and how they use the instructions we give them.
The stakes for users and for Meta
For users, the promise of Dear Algo is straightforward: less irrelevant noise, more of what they actually care about, and a sense that they can correct the feed when it drifts. If the feature works as described, someone who is tired of seeing a particular topic or wants to dive deeper into a niche interest will be able to say so directly, without reverse engineering the algorithm through trial and error. That could make Threads feel more responsive and less like a slot machine, especially for people who are already comfortable addressing the algorithm as if it were another account in their network.
For Meta, the stakes are more complex. On one hand, giving users visible control can help defuse criticism that its systems are manipulative or unaccountable, and it may even improve engagement if people feel more satisfied with what they see. On the other hand, exposing a conversational interface to the algorithm invites scrutiny when the system fails to honor clear instructions, and it creates a new surface area for abuse, from coordinated campaigns to flood the feed with certain topics to attempts to game the ranking system with scripted Dear posts. How Meta balances those risks against the potential upside will determine whether Dear Algo becomes a core part of Threads or remains a niche experiment.
More from MorningOverview