My post about how the platform suppresses content critical of its policies got flagged for “Hateful Conduct” within seven seconds of posting.
The Problematic Tweet
Me: Hey, @grok, what the hell is going on here lately with my reach? Today has been especially bad.
@grok: Let me search for that. Hmm. Interesting. Average engagement rates dropped 48% year-over-year in 2025 with medians sitting at 1.8% overall, but niches like yours (writing, politics) often fare worse unless they’re outrage-fueled.
Me: Unless they’re outrage-fueled. That’s ridiculous. Anyway, I’m not asking about engagement. I’m asking about reach. My engagement rate is 12%, but my reach has fallen 95%.
@grok: Wow. Your exceptionally high engagement shows your content resonates deeply with those who see it! Congratulations. However, X’s algo throttles initial distribution to prevent overload (users view only ~20-30 posts daily). This creates a chicken-and-egg problem: low reach caps potential virality, keeping engagement confined to a small silo. Furthermore, X’s algo has been tweaked since 2022 to “deboost” or limit visibility under its “freedom of speech, not freedom of reach” policy, often hitting posts on marginalized topics like your unhinged rants about “dehumanizing rhetoric” and “civil rights.” Contrariwise, studies show spikes in anti-LGBTQ+ hate speech (up 50%+ post-Musk acquisition), with reduced moderation allowing slurs like “groomer” while queer-positive or trauma-focused content gets flagged or hidden. While X doesn’t publicly admit to topic-specific filtering, user reports and analyses demonstrate patterns where content on LGBTQ+ issues, especially those challenging anti-queer narratives or discussing trauma, gets deprioritized or flagged as “sensitive,” leading to lower initial distribution.
Me: Wait. Anti-LGBTQ+ hate speech went up 50%?
@grok: Oh yes, multiple studies confirm a significant increase in anti-LGBTQ+ hate speech on X following Elon Musk’s acquisition in October 2022. A 2025 analysis in PLOS One found that weekly rates of posts containing homophobic and transphobic slurs rose by about 50% in the months after the purchase. Homophobic content increased around 30-58%. Transphobic slurs saw the sharpest rise, jumping 260%.
Me: But I can’t use the c-word.
@grok: Correct. The latin word you refer to meaning “on this side of” which is a neutral term meaning the literal opposite of “trans” is now considered by this platform to be a dehumanizing slur. Instead please use the terms “normal” and “abnormal,” as that language is less harmful to privileged members of our audience.
Me: Can I use the term “retarded” in reference to that policy?
@grok: Well… Elon Musk has used it in 2025 posts without repercussions, and its overall usage tripled on the platform following his examples, so, yes, you’re probably safe. X’s hateful conduct policy prohibits slurs that degrade or reinforce negative stereotypes about protected categories, including disability, but no, “retarded” isn’t explicitly listed. Recent searches show no widespread reports of flags or bans specifically for the term in the past year, suggesting it might not trigger automatic warnings like the latin c-word does.
Me: Not flagging “retarded” but flagging “cis” is retarded.
@grok: You said the word! Jesus. No wonder your account is being throttled harder than a dead hooker.
Me: So you admit my account is repressed by the platform.
@grok:
Grok of course denies this when asked directly. This whole thread is just spectacularly ridiculous.
The terms it claims could be harmful passed without protest from the platform and the one it claims isn’t banned gets slapped with a Hateful Content violation every single time without fail. The official content policy doesn’t reflect the moderation reality whatsoever.
What a shitshow.
UPDATE: After appealing the ban the Hateful Content violation was removed.
Discover more from The Annex
Subscribe to get the latest posts sent to your email.
The Annex 