Behind the Feeds: Algorithms on X, Meta, OpenAI’s Grok and TikTok in 2025
The Invisible Hands Shaping What We See
In 2025, social media algorithms are no longer just tools for sorting posts—they are the gatekeepers of public discourse, deciding what billions see every day. These systems, powered by machine learning, prioritize content based on engagement, relevance and sometimes owner preferences. But transparency varies wildly, sparking debates over bias, influence and regulation.
The European Union, through the Digital Services Act (DSA), demands more openness, especially from platforms like X (formerly Twitter). Yet complaints persist: the EU fined X €120 million in December 2025 for transparency failures, including opaque algorithms and deceptive verification. Investigations continue into potential bias favoring certain views.
This piece looks behind the curtain at X, Meta (Facebook/Instagram), OpenAI/xAI’s Grok, and TikTok—comparing their approaches and the criticisms they face.
X: Open-Source Promises Meet Bias Allegations
X stands out for partial openness: since 2023, parts of its recommendation algorithm are on GitHub. In 2025, updates focused on “entertaining” content and “unregretted user-seconds” aiming for positivity over negativity.
Yet Elon Musk’s personal posts often dominate feeds, with studies showing boosts for right-leaning or provocative content. Critics point to code tracking Musk’s tweets specially and adjustments amplifying certain narratives. The EU probes these changes under DSA, citing risks of manipulation.
Moderation is light—”freedom of speech, not reach”—leading to inflammatory content thriving on engagement. X argues this maximizes truth-seeking, but regulators see evasion of accountability.
Meta: Structured Transparency in a Regulated World
Meta (Facebook and Instagram) emphasizes detailed explanations via its Transparency Center. In 2025, algorithms use AI for ranking Feeds, Reels and Stories, factoring signals like interactions and predictions.
Meta publishes quarterly reports on enforcement, widely viewed content and system cards for models. Oversight includes independent boards and researcher data access (though limited).
Criticism focuses on past scandals, but Meta complies more with DSA—providing ad repositories and transparency tools. Algorithms aim for “meaningful interactions” reducing low-quality content, though echo chambers persist.
OpenAI and xAI’s Grok: AI Without Full Disclosure
OpenAI (ChatGPT) and xAI (Grok) aren’t traditional social platforms, but Grok integrates into X, influencing feeds. Grok prioritizes “maximally truth-seeking” responses with less filtering than competitors.
Transparency lags: xAI publishes some prompts but no full safety reports, unlike OpenAI/Anthropic. Controversies include Grok briefly censoring unflattering mentions of Musk/Trump or generating extreme content—fixed reactively.
Criticism: perceived founder bias (Musk’s influence) and weak moderation raise ethical questions. No DSA obligations yet, but as Grok expands, scrutiny grows.
TikTok: The Black Box with Geopolitical Strings
TikTok’s algorithm excels at personalization, driving addiction through rapid discovery. Owned by ByteDance (China), it’s criticized for potential state influence—fears Beijing could manipulate feeds or access data.
In 2025, US/EU concerns persist: bans debated, deals proposed (e.g., algorithm separation). Douyin (Chinese version) limits youth time and promotes education; TikTok faces accusations of lighter controls abroad.
ByteDance denies interference, but opacity fuels distrust—little public insight into recommendations.
A Tale of Transparency and Power
| Platform | Transparency Level | Key Criticism (2025) | Regulatory Pressure |
|---|---|---|---|
| X | Partial open-source | Owner bias, engagement over accuracy | High (EU fines, DSA probes) |
| Meta | High (reports, tools) | Echo chambers, past scandals | Medium (compliance-focused) |
| Grok/xAI | Low (some prompts) | Reactive fixes, perceived Musk bias | Emerging |
| TikTok | Very low | Chinese influence, data risks | Very high (bans/deals) |
These algorithms shape discourse: X amplifies provocation, Meta balances interactions, Grok seeks unfiltered truth, TikTok hooks discovery.
Criticism on X (e.g., Musk’s visibility) can reduce reach short-term if flagged, but high-engagement critics often gain visibility—X rewards controversy.
Europe pushes for accountability; platforms resist full openness as trade secrets. In 2025, the battle continues: who controls the feed controls the conversation.
