Meta's flagship platform has 3.07 billion monthly active users and generated $164.5 billion in revenue in 2024. By almost every financial metric, it's thriving. Spend 10 minutes on Facebook today, though, and something feels fundamentally broken. The feed is choked with AI-generated images of wounded veterans asking for prayers, fake historical photos, and fictional plants with fabricated care instructions. Real friends? Nearly invisible.
That tension — between financial health and social decay — is what makes the "Facebook is cooked" argument worth examining carefully. The money is real. The rot is also real. Understanding why both can be true simultaneously tells you something important about where platforms go when engagement becomes the only metric that matters.
This piece covers how AI-generated spam is structurally destroying Facebook's core social function, why the platform's algorithm actively rewards the problem, what the demographic data says about long-term viability, and where functional lock-in keeps the platform alive despite the decay.
Key Takeaways
- Facebook reported 3.07 billion MAUs in Q4 2024, but Gen Z is largely absent, with Gen X now the platform's largest demographic — a structural aging problem no engagement metric can mask.
- The American Dialect Society named "AI slop" its 2025 Word of the Year; Facebook is its primary distribution channel, with operators exploiting the $68.44 vs. $5.52 ARPU gap between North American and Asia-Pacific users.
- Meta's 2018 "meaningful social interactions" algorithm shift inadvertently rewarded spam operations — the system built to increase connection ended up accelerating content degradation.
- Facebook Marketplace and Groups retain no viable competitor alternatives, creating functional lock-in that keeps MAU numbers elevated even as the social layer collapses.
- Meta generated $164.5 billion in revenue in 2024 (up from $134 billion in 2023), removing almost all financial pressure to reform the user experience.
How Facebook Got Here
Facebook wasn't always an AI slop factory. For roughly a decade — call it 2008 to 2018 — it functioned as advertised: a social graph where you saw content from people you actually knew. The shift started in 2018.
Under pressure following the Cambridge Analytica scandal and congressional hearings about misinformation, Meta announced an algorithm change prioritizing "meaningful social interactions." The intent was to surface content that generated comments and discussion, deprioritizing passive video consumption and publisher posts. Structurally sound reasoning. Terrible outcome in practice.
Comments and reactions are easy to engineer. Outrage generates both. Spam operators — according to Groundy's analysis of the platform's decay — quickly identified that emotionally manipulative AI-generated content targeting US audiences could generate massive engagement while costing almost nothing to produce. The economics are stark: North American users generate $68.44 ARPU versus $5.52 in Asia-Pacific. Operators based in lower-cost regions flood feeds targeting high-value audiences. The margin is enormous.
By January 2026, academic researchers had formally defined "AI slop" with three properties: superficial competence, asymmetric effort, and mass producibility. All three are actively rewarded by Facebook's current ranking system. The American Dialect Society named it their 2025 Word of the Year. That's not a niche technical complaint — it's a mainstream cultural diagnosis.
The platform didn't arrive at this state through negligence alone. Financial incentives, algorithmic architecture, and the deliberate absence of behavior-regulating tools all contributed to a system where "Facebook is cooked" isn't hyperbole. It's an accurate description of what the social layer has become.
The AI Slop Pipeline Is Structurally Embedded
The content problem on Facebook isn't a moderation failure that stricter policies could fix. It's architecturally embedded.
According to Groundy's reporting, the typical operation works like this: operators use ChatGPT-generated prompts fed into image generators like Midjourney to produce emotionally resonant imagery — wounded soldiers, grieving children, miraculous plant cures — then deploy it into interest-based Facebook communities. Fake Holocaust victim imagery and fabricated historical photographs regularly penetrate history and memorial groups. Fictional plant species with false care instructions spread through gardening communities.
The content targets emotion because emotion drives engagement, and engagement drives reach. Facebook's algorithm doesn't distinguish between a genuine post from a veteran's family and a synthetic image engineered to look identical. Both generate reactions. Both get distributed.
What makes this structurally difficult to solve: the cost of production approaches zero, the financial return is positive, and the platform's ranking signals can't tell the difference. Moderation at scale against near-infinite content generation is a losing race. Meta knows this. The quarterly revenue numbers suggest they've accepted it.
This approach can fail to self-correct precisely because the economics work so cleanly. When a behavior is profitable and technically difficult to detect, hoping a platform polices it away is wishful thinking — especially when that platform faces no meaningful financial penalty for the problem continuing.
The Missing Feedback Loop
There's a deeper structural problem that predates AI slop entirely. A Hacker News discussion analyzing Facebook's decline surfaced a compelling explanation: online communities deteriorate specifically because non-verbal social signals that regulate behavior in physical settings don't exist online.
In person, a disengaged look or turned back provides immediate behavioral feedback. Online, those signals are entirely absent. People don't calibrate their behavior the same way. And the platform design ensures this absence remains.
Facebook's reaction system responds to a post's emotional content — not to the poster's behavior. There's no equivalent of "this person is being disruptive." No detach rate visible to community members. Contrast this with YouTube's creator analytics, which exposes immediate click-away data and demonstrably helps creators calibrate content quality. That behavioral signal exists on YouTube because it helps creators improve. It's absent on Facebook because behavior-regulating signals reduce aggregate engagement — and platforms with engagement-linked monetization have direct financial incentive to exclude any tool that lowers those numbers.
This is the mechanism that makes the platform's decline feel inevitable rather than fixable. The business model and a healthy social environment are in direct conflict. One of them wins. The business model always wins.
The Demographic Cliff
The revenue numbers look strong. The user composition tells a different story.
Gen X is now Facebook's largest demographic. Gen Z is largely absent. According to Groundy's platform analysis, this demographic shift represents a slow-motion structural problem that financial metrics obscure. Advertisers targeting 18–34 audiences are already moving budget elsewhere. The users who remain skew older, and the content those users engage with — AI-generated emotional bait — accelerates the platform's reputation among younger demographics as somewhere they simply wouldn't go.
This is the pattern of a platform in terminal demographic decline. Not fast collapse. Slow aging. MySpace didn't disappear overnight — it became irrelevant over several years while its surviving users didn't notice because they were still there. That's the trajectory Facebook is on. The lights stay on. The culture empties out.
Comparison: Platform Decay Patterns
| Dimension | Facebook (2026) | Twitter/X (2023–2026) | Reddit (2023–2026) |
|---|---|---|---|
| Content Quality Signal | Engagement-linked, no behavior regulation | Algorithmic + manual, chaotic | Community moderation, variable |
| Revenue Model | Advertising, $164.5B in 2024 | Advertising + subscriptions, struggling | Advertising, IPO'd 2024 |
| Core User Retention | Marketplace + Groups lock-in | News/real-time events | Niche community depth |
| AI Spam Exposure | Extremely high | High | Moderate (moderation helps) |
| Demographic Trajectory | Aging toward Gen X | Mixed, volatile | Retaining younger users |
| Functional Replacement | No Marketplace alternative | Mastodon, Bluesky (partial) | Partial (Discord, forums) |
| Verdict | Zombie platform | Unstable transformation | Cautious growth |
The comparison clarifies something important: "Facebook is cooked" doesn't mean it disappears. Twitter/X demonstrates that a platform can lose its core social function while remaining financially operational for years. Reddit shows that community moderation provides structural protection against content decay — a design choice Facebook explicitly moved away from in favor of algorithmic control.
Facebook's position is arguably worse than Twitter/X in one dimension: the functional lock-in through Marketplace and local Groups keeps users engaged despite hating the experience. That's a zombie platform dynamic. Users aren't there because they want to be. They're there because they need to sell a couch.
Practical Implications
Developers and engineers building social features or content platforms should treat Facebook's trajectory as a case study in what engagement-maximizing architecture produces at scale. The 2018 "meaningful social interactions" algorithm change is a documented example of Goodhart's Law in production: when engagement becomes the metric, content optimizes for engagement rather than value. Build accordingly — or inherit the same problem.
Companies running Facebook ad campaigns need to audit audience quality, not just reach. The 3.07 billion MAU figure includes a growing proportion of low-intent, algorithmically-captured users. Cost-per-acquisition data from 2025 Facebook campaigns should be scrutinized against actual conversion quality, not click-through rates.
End users keeping Facebook accounts for Marketplace or local Groups are making a rational choice. But the social layer — news feed, friend content, organic discovery — is functionally broken and isn't getting fixed. Manage expectations accordingly.
Short-term actions (next 1–3 months):
- Audit Facebook ad spend against CPA data from actual converted customers, not click metrics
- If running community groups on Facebook, establish a parallel presence on Discord or a self-hosted alternative — Groups aren't going away immediately, but the content environment degrades your brand association over time
- Developers: review your own platform's engagement metrics for Goodhart's Law exposure before it becomes your problem at scale
Long-term strategy (next 6–12 months):
- Watch EU Digital Services Act enforcement actions against Meta — Brussels has moved faster on platform content accountability than Washington, and regulatory pressure is the most likely external forcing function for behavior change
- Monitor whether Marketplace gains a credible competitor; that's the actual lock-in mechanism keeping engagement numbers elevated, and it's also the clearest product opportunity in this space
Opportunity worth noting: Marketplace has no viable competitor — but that gap is a genuine product opening. A focused, moderation-heavy local commerce platform could capture displaced users as content quality continues declining. The core problem for Marketplace users isn't finding more listings. It's distinguishing real sellers from scammers. Build for trust signals first, not scale.
What Happens Next
Facebook is cooked as a social platform. It may persist for years as a commerce and local utility platform. Those are different things, and conflating them is what makes the financial metrics misleading.
The AI slop pipeline is structurally embedded — not a moderation problem. Missing behavioral feedback loops make social degradation a design feature, not a bug. Gen X is the platform's largest demographic, and that slow demographic cliff is real. Marketplace and Groups lock-in creates zombie platform dynamics, not genuine engagement.
Watch two things over the next 6–12 months. First, EU Digital Services Act enforcement actions against Meta — that's the most likely external pressure that could force meaningful change. Second, any credible Marketplace competitor emerging. That's the actual structural threat to whatever functional utility Facebook still provides.
The clearest takeaway: stop reading Facebook's MAU numbers as evidence of platform health. Reach and value have decoupled. The platform that showed 3.07 billion people their friends' posts in 2018 is not the same product serving 3.07 billion accounts in 2026.
That distinction matters — whether you're allocating ad budget, building community tools, or deciding where to put engineering resources. The social network is cooked. The commerce utility is on life support. Plan for both realities, not the revenue figure.
Top comments (0)