← Cross-cutsCross-cut · Social & creator

Attention, and who pays for it

Six Teardown studies cover social, content and creator platforms — Reddit, Substack, ByteDance, RedNote, Tencent and Discord. Read together, they're all answering one question in different currencies: how do you convert a community's attention into money without spending the trust that made the community worth anything?

Three ways to monetize attention

The cluster splits by how it turns engagement into revenue. The advertising platforms — Reddit (~94% ads), Tencent (WeChat's 1.42bn-user graph), ByteDance (Douyin/TikTok), RedNote (~80% ads) — sell attention to marketers. The subscription/community players — Substack (a 10% cut of paid newsletters) and Discord (Nitro plus a new opt-in ad model) — charge the users or creators directly. Different currencies, same underlying asset: a community whose attention or trust is the thing being sold. And every one of these studies turns on the tension between monetizing that asset harder and degrading it.

The shared risk: monetization spends trust

The cross-section's sharpest insight is that for these platforms, the balance sheet asset is community trust — and every monetization lever spends some of it. RedNote makes it measurable: the share of users calling its posts "truly trustworthy" fell from 76% to 41% as it leaned into ads and commerce. Discord is reversing a years-long "no ads" stance with opt-in Quests right before an IPO, betting it can charge rent without breaking the culture. Substack's 10% take and recurring trust controversies strike the writer-reader relationship directly. The platforms that monetize too hard risk the authenticity that gave them anything to monetize — a risk no amount of scale removes.

AI is both the funder and the threat

AI cuts two ways across the cluster, and Reddit is the clearest case. Its human-written archive made it the most-cited domain in AI answers and a premium data-licensing seller (Google ~$60M/yr, OpenAI ~$70M/yr) — AI as a new revenue line. But AI answer engines also let users get Reddit's content without visiting Reddit: ChatGPT's Reddit-citation rate fell from ~60% to ~10% in six weeks. ByteDance shows the other edge — deliberately spending its margin down on an AI capex super-cycle (~RMB 160B in 2026) to defend its recommendation engine. For attention platforms, AI is simultaneously a buyer of their data, a bypass of their traffic, and a cost center — and the studies disagree on which dominates.

The China–West split is a regulatory fault line

Half the cluster is Chinese (Tencent, ByteDance, RedNote) and half Western (Reddit, Substack, Discord), and the dividing line is regulation as a wild card. ByteDance carries the most extreme version — a forced TikTok US divestiture into a majority-American JV that still licenses its algorithm, a deferral rather than a resolution. RedNote drew a one-year Taiwan ban and US national-security scrutiny after its brief "TikTok refugee" surge. Tencent sits on the US 1260H list. The Western names face their own regulatory front — Discord's child-safety suits, Substack's content controversies. Geopolitics and platform-safety law are a layer of risk this cluster carries that pure-software clusters don't.

Where they agree — and where they split

All six agree the asset is community and the constraint is trust, and all carry an AI question (data-licensing upside vs traffic-bypass downside) and a regulatory overhang. They split on monetization maturity and on price: Tencent and ByteDance are scaled, profitable attention machines deciding how to deploy AI; Reddit just hit its first profitable year at a rich ~15× sales; RedNote is ramping ads into a ~$50B mark; Substack and Discord are pre-IPO bets monetizing trust they can't yet prove they'll keep. The demand for community and attention isn't the question. Whether each platform can charge for it without spending the trust underneath is.

The cluster at a glance

CompanyPlatform & scaleMonetizationGrowthBiggest risk
RedditNYSE:RDDT126.8M daily uniques (Q1'26)~94% ads · $5.23 global ARPU · 91% GM+69% rev · 7 straight 60%+ qtrsAd+Google concentration; AI-summary bypass
SubstackPrivate5M paid subs · 50k+ publications10% take rate · ~$45M est. rev2M→5M paid subs in 2yrs10% take vs 0%-fee rivals + churn
ByteDancePrivateDouyin 936M MAU + TikTok global~60% ads, 26% live, 14% e-com~$186B rev 2025 (+~20%)TikTok divestiture + AI capex squeeze
XiaohongshuPrivate~400M MAU · ~72% femaleAds ~76-80% rev; e-comm sidecarProfit $1B (2024)→~$3B guided 2025Trust erosion vs commercialization
TencentHKEX:0700WeChat/Weixin 1.42bn MAU super-appGames + ads + fintech · 56% GMFY25 rev +14% to ¥751.8bnDouyin attention drain + AI fast-follower
DiscordPrivate200M+ MAU · ~2B gaming hrs/moNitro + commerce + opt-in Quests ads; ARPU ~$3~$725M ARR 2024, ~21% YoYChild-safety AG suits + ad-pivot backlash

Figures as of each study’s stated date (2026-06); on mixed bases and not strictly comparable. See each teardown for sourcing and the full weighing.

The six studies — and the question each turns on

Reddit, Inc.NYSE:RDDTAre the things that make Reddit special — AI-citation crown, ARPU runway, elite margins — durable, or does AI both fund it and replace the need to visit it?Read the full weighing →Substack Inc.PrivateCan a 10% cut of other people's newsletters ever justify a $1.1B valuation on ~$45M of revenue?Read the full weighing →ByteDance (字节跳动)PrivateCan ByteDance convert Meta-rivaling scale into durable, profitable advantage while it spends down its margin on AI and absorbs a forced TikTok US divestiture?Read the full weighing →Xiaohongshu (RedNote)PrivateCan Xiaohongshu keep growing revenue without spending the community trust that is its only real asset — and does an ~80%-ad business justify a ~$50B valuation?Read the full weighing →Tencent HoldingsHKEX:0700Is Tencent a still-compounding platform, or a defensive incumbent harvesting a maturing Chinese internet?Read the full weighing →Discord Inc.PrivateCan Discord monetize a 200M-person community without breaking the culture that made it loved — and police it well enough for regulators and a jittery IPO market?Read the full weighing →

This is the kind of reading the Desk does for you

A cross-cut takes six teardowns and one thesis and asks what they say together. The Desk does the same for the companies you actually own — your thesis, the rivals that move it, and the tripwires that would change your mind.

See the Desk →