CDT Review Finds Social Media Algorithm Controls Often Lack Transparency

What the CDT compared

The CDT blog post looks at the "levers of control" users can exercise over their algorithmic feeds on the five platforms it names as the most popular among teenage users: YouTube, TikTok, Instagram, Snapchat and Facebook. The organization says that, despite growing concern, "we know surprisingly little" about these controls and whether they meet users' needs.

The CDT notes that Americans use social media to engage with news, form political beliefs and look for information, while content is mostly distributed by personalization algorithms that are complex and opaque. It points to the UN's Universal Declaration of Human Rights, the EU's General Data Protection Regulation and the Digital Services Act as frameworks that give users rights to seek and receive information and to an explanation of why they see particular content. It also says teens and parents want more control, and cites Meta's recent settlement over social media addiction claims, which it says has implications for ongoing claims against TikTok, Snapchat and YouTube.

Its four takeaways: algorithmic controls are only useful when transparent; keyword and sensitive content filters are inconsistent and potentially risky; engagement metrics influence algorithms but operate in the dark; and parental controls are sweeping but may not always be effective.

Topic controls and the Instagram feature

Instagram and TikTok both offer algorithmic topic controls. On June 10th, 2026, Instagram C.E.O. Adam Mosseri announced a feature called "Your Algorithm". It gives users an AI-generated summary of the content their personalization algorithm is currently favoring. Users can then opt in or out of additional topic areas that, according to the platform, will influence what they see, mainly on the Reels page.

TikTok lets users see more or less of pre-set topics on the For You Page: creative arts, current affairs, dance, fashion & beauty, food & drinks, health & fitness, humor, lifestyle, nature, pets, sports and travel. The CDT says these categories are vague and that researchers found the content labeling process inconsistent and confusing.

The CDT describes such controls as a critical point of intervention, but says there is little understanding of how platforms label content or how the tools change feeds. At the time of writing, it says, only Instagram and Facebook give users feedback on the effect of their interventions. On Instagram, a badge on the Reels page shows when content appears because of a user's choices, and a "why you're seeing this" option links back to algorithmic preferences. Facebook offers the explanation feature too, but without granular topic controls users cannot iterate on it.

Even Instagram's model, the CDT argues, remains opaque and potentially superficial. Because most recommended videos go untagged, the badges "could simply placate users' desire for additional control". Without actionable transparency, it warns, such tools may offer "incomplete" or "illusory" control.

Filters and engagement signals

TikTok and Instagram let users remove specific words and phrases from their feeds, and Instagram also allows emojis. The CDT says these filters have notable gaps, especially in short-form video, where content appears in captions, comments, audio, stickers and remixed music. It adds that users have anecdotally documented failures on both platforms.

YouTube, TikTok, Instagram and Facebook allow users to filter out "sensitive content", and Instagram and Facebook allow filtering of "political" content. The CDT says it is ambiguous how these labels are applied. It notes that LGBTQ+ content has sometimes been labeled sensitive and that political content is hard to define. Poorly designed filters, it says, can lead to unfair censorship of marginalized groups.

The post also covers implicit levers, such as likes, comments, reposts, shares, watch times, reporting content and blocking users. These are not meant to alter feeds directly but still influence them. TikTok says it uses "user interactions" to inform recommendations. The CDT says feedback on these actions is limited and variable, and that only Instagram and Facebook provide it, through the "why you're seeing this content" feature, for example by telling a user she sees content because "you interacted with content about travel and more". The source text notes this insight is not available for all posts.

What it means for users

The CDT's central message is that control without visibility has limited value. Users on all five platforms can nudge their feeds, but the analysis suggests they often cannot tell whether their choices worked. The group suggests companies that have not implemented feedback measures consider telling users how labels and settings alter personalization. For parents and teens, the post cautions that even sweeping parental controls may not always be effective.