✅ LinkedIn — Reach throttling for "generic" AI posts
|
LinkedIn isn't deleting AI content. It's suffocating it. |
Per The Next Web, the platform's new "AI solving AI" classifier flags generic AI posts with 94% accuracy in early tests. Flagged posts get capped at first-degree contacts — your network only, no recommendations. The same system now hunts AI comments and bot profiles, with new filters letting users restrict their feed to verified profiles only. |
Best AI use: AI-assisted drafts of your own thinking. |
Worst: "Rewrite with AI" → post unedited. |
— |
✅ Facebook, Instagram & Threads — Label or get downranked
|
Meta runs one framework across all three apps: realistic AI imagery, video or audio must carry the "AI info" label, applied either by industry-standard metadata signals (C2PA) or by your own self-disclosure. |
Per Meta's Transparency Center, unlabeled realistic AI content is downranked rather than removed. |
Instagram tightened the screws on May 4 with an opt-in "AI Creator" profile label that reads "This profile posts content that was generated or modified with AI" and follows you across Explore and Reels (Social Media Today). |
Mosseri's January warning was the prequel: AI mimicry of any creator's style is now trivial, so the platform is leaning hard on transparency. |
Best AI use: AI for production assistance (editing, captions, ideation) plus accurate labeling of synthetic visuals. |
Worst: Photoreal AI faces / fake "behind the scenes" content with no label. |
— |
✅ TikTok — Auto-labels and a 73% reach cliff
|
TikTok is the most aggressive enforcer. |
Per its Creator Academy guidance and AuditSocials' 2026 rules breakdown, any AI-generated content that depicts a realistic person or scene must be labeled — via the in-app AI toggle, an on-screen sticker, or a description note. TikTok now uses C2PA Content Credentials + invisible watermarking to auto-detect, and it has already labeled 1.3 billion videos this way. |
Here's the part nobody saw coming: if TikTok auto-labels you after you posted unlabeled, that's an immediate strike and a 73% reach suppression within 48 hours. AI-assisted scripts, captions and hashtags are exempt. The line is realism. |
Best AI use: AI editing, AI scripts, AI translations. |
Worst: Realistic AI avatars, voice clones, or "footage" of events that didn't happen, posted with no toggle. |
— |
✅ YouTube — Disclose realism, or face automatic labels (and worse)
|
YouTube's synthetic content policy requires disclosure when you alter a real person, real event, or real place to look authentic. |
Production-side AI — scripts, ideas, titles, captions, B-roll — is exempt. The Studio toggle is called "Altered Content." |
The "Altered or Synthetic" label now appears directly under long-form players and as an overlay on Shorts (Variety). Skip the disclosure and YouTube can apply a non-removable label, demonetize you, or worse: in this year's "AI slop" crackdown the platform deleted 16 channels representing 4.7 billion views and roughly $10M in annual revenue under its Inauthentic Content Policy (OutlierKit). |
Best AI use: AI in the pipeline that supports a human-led creative point of view. |
Worst: Templated, mass-produced AI "shorts factories" with no original voice. |
— |
✅ X / Twitter — "Made with AI" toggle, with revenue-share teeth
|
X added a "Made with AI" toggle (BuzzInContent) and is layering AI into Community Notes itself. |
The platform doesn't prohibit AI-written posts — but for sensitive categories (conflict, elections, crisis), unlabeled synthetic media triggers a 90-day revenue-sharing suspension, permanent on repeats. |
Best AI use: AI-drafted text, satire, clearly stylized AI imagery. |
Worst: Realistic AI imagery of news events without the toggle. |
— |
✅ Pinterest — Auto-labels, plus user-controlled "see less AI"
|
Pinterest auto-detects AI Pins via IPTC metadata and its own classifiers, then labels them in-feed (Pinterest Help). |
The big shift is user-side: in the Manage GenAI settings (Pinterest Newsroom), users can now opt to see fewer AI Pins in beauty, art, fashion and home decor — the four categories most polluted by AI slop. |
Your AI Pin doesn't get banned. Users just disappear it. |
Best AI use: AI lifestyle imagery clearly labeled, paired with real product photography. |
Worst: All-AI moodboards in the four "opt-out" categories. |
— |
✅ Bluesky — Composable labels, no central penalty
|
Bluesky takes the opposite approach. There is no central reach algorithm to throttle you. Instead, community labelers apply tags and individual clients decide what to do with them. You won't get downranked by HQ — but a major labeler tagging your account can make you invisible to half the network. |
Best AI use: Transparency-first AI use, opt-in labels. |
Worst: Posting AI content in feeds whose curators have publicly banned it. |
— |
✅ Google Business Profile — AI replies encouraged, but moderated
|
The outlier! |
Google is actively shipping AI: its new Reply to reviews with AI feature is rolling out in the US, Brazil and India (Search Engine Land). |
AI-drafted replies are compliant — but every reply now passes through a new ReviewReplyState moderation queue (PENDING → APPROVED or REJECTED). Per ALMcorp, templated AI replies erode trust faster than no reply at all. Edit them. |
Best AI use: AI draft → human edit with specifics from the review. |
Worst: Mass-sending unedited AI replies. |
