Vol. 57 The Content Business—Half Emotion, Half Algorithms—A Conversation with Zheng Xun of Kuaishou's Magnet Engine
Summary
Short-video user penetration and time-spent growth are largely at their ceiling; over the next 3-5 years, competition will shift from expanding traffic to productizing content, improving conversion efficiency and deepening monetization. 庄明浩 cites industry reports showing that short-video platforms were the first in the sector to surpass 1B users, while time spent continued to grow even as most audiovisual categories declined in 2023 and 2024. But 2022 had already delivered an extreme stress test. Under the framework of “users × user time × money = value,” the first two variables offer limited room, leaving platforms to keep extracting more ROI and more “money.”
Native closed loops are rewriting the unit economics of content user acquisition by reducing cross-app leakage, unifying attribution and turning advertisers from one-off buyers of traffic into operators of durable businesses. 郑迅 says jumping from Kuaishou to an external app is one of the biggest points of leakage in the ad funnel; after moving native, activation costs can differ by “nearly 10x.” ROI bidding for mini-games launched around late Q1 to early Q2 2024, with penetration quickly reaching 70%-80% and now exceeding 90%. Platforms and clients are no longer trapped in “guess what I’m guessing”; with unified data, they can manage revenue, re-engagement, repurchase and private-domain operations.
Short dramas and mini-games have shown that IAA and IAP are not mutually exclusive, but two “payment methods” for reaching different users. IAA makes users pay with 30 seconds of attention under forced viewing, while IAP exchanges cash for content. For Kuaishou short-drama clients, the window between starting with IAP and adding IAA has narrowed from roughly 7 days in 2023 to 1-3 days; the two models are now broadly equal in scale on the platform, with IAA occasionally larger for mini-games. Kuaishou’s platform-wide ad-free pass has lifted purchasing users’ time spent by more than 50%, showing that hybrid monetization can extend engagement and create more downstream monetization.
Dynamic comics may be the most important new content vertical to watch in 2025, because existing demand, AI-enabled supply and sharply lower costs are converging. 郑迅 estimates that short dramas cost several thousand yuan per minute on average, while rough dynamic comics can cost several hundred yuan and polished versions around RMB1,000-2,000 per minute. Genres such as fantasy and mythology that are difficult to carry with live-action short dramas may find supply through generation technologies such as Kling. Potential creators span animation, Chinese animation, novels, short dramas and light-comedy teams, though 郑迅 says demand has not yet been fully released.
AI’s clearest commercial value is not replacing top creators, but systematically raising the floor of content supply. Kuaishou already has tens of millions of yuan in daily ad spend backed by AI-generated creative; this is booked as advertising revenue rather than Kling revenue. Digital humans fill time slots when real hosts cannot go live, AIGC reduces repetitive, mismatched and non-compliant creative, and large-model dialogue addresses the lack of customer-service staff at small merchants. 庄明浩 calls this the “drudgery of the present” in existing businesses, but it has clear ROI; video generation through Kling is closer to the “poetry and distant horizons” of opening new categories.
The core opportunity in using large models to overhaul recommendation is to use reasoning to compensate for sparse ad signals and create exploration room for cold-start products. Feeding behavior directly into a model only produces answers such as “Taobao, WeChat and Pinduoduo”—correct but useless. Kuaishou instead uses 3 steps: describe the user, infer latent needs, then map them to the closest products on the platform. 郑迅 says this makes recall results differ from traditional models, increases exposure for new games and products, and partially turns the black box of “that is just how the algorithm works” into a glass box.
The AI application layer has not settled; the decisive competition between platforms and third parties is not simply who owns a general-purpose model, but who packages specific use cases and complete workflows better. After DeepSeek emerged, companies rapidly integrated open-source models and the focus shifted to “what problem are you actually solving?” Kuaishou has also avoided making AI a standalone OKR for a small group, preferring every OKR to account for AI. The next phase is to break tasks such as script rewriting, video generation and editing into measurable Agents, moving from “language” toward tool-calling, collaboration and permission-constrained “behavior.”
Deep dive
1. Content Is Becoming a Consumable Product That Can Be Operated and Delivered
庄明浩 sets a deliberately “snobbish” lens for the conversation: rather than looking only at DAU, MAU and time spent, he examines how content becomes a product—“product” in quotation marks—through ad spend, GMV, consumption and conversion. The term is not derogatory; it means content has entered a complete system of consumption decisions and delivery.
郑迅’s role in strategy products at Kuaishou’s Magnet Engine sits precisely at the intersection of content, advertisers and platform mechanics. For investors, this vantage point makes it easier to see the supply structure, unit economics and extractable platform value behind traffic metrics.
Their shared premise is that flourishing content consumption does not mean reducing users to numbers. If productization cuts switching, unifies interaction and improves supply quality, commercial efficiency and user experience may simply be two ends of the same causal chain.
2. The Structural Changes of 2024 Will Deepen Further in 2025
郑迅 attributes the most significant business changes of the past year to the rapid growth of short dramas, mini-games and the novel ecosystem. Kuaishou has chosen native in-app distribution, closed-loop mini-games, short dramas and novels, seeking to control user experience while monetizing; investment will continue in 2025.
庄明浩’s 2024 snapshot is more aggressive: short-drama scale exceeded box-office revenue, domestic live-stream tipping revenue “may” have exceeded gaming, and mini-program game ad spend may already have surpassed native gaming-app spend. Regardless of the precise definition used for each metric, the changes collectively point to a shift in the revenue structure of content products.
AI is the new variable. Content production, ad creative, copywriting, customer service and recommendation systems are all beginning to be rewritten, so 2025 is not merely a continuation of short-drama and mini-game growth; it also layers a new efficiency “buff” onto the existing industrial chain.
3. Short Video Has Become Content-Consumption Infrastructure
庄明浩 cites an annual audiovisual report showing that short-video platforms were the first in the sector to surpass 1B users; the broader audiovisual category crossed 1B around December 2022, while short video remained more resilient afterward.
The unusual environment of 2022 produced exceptionally high time spent. Many categories declined in 2023 and continued falling in 2024, while short-video figures still moved higher. This means short video is no longer just a content format, but infrastructure for internet consumption in China and beyond.
Once infrastructure matures, growth is no longer simply another user or another few minutes of scrolling. Standardized deliverables such as short dramas, novels, dynamic comics and mini-games begin growing out of the short-video “skin.” Platform competition consequently shifts from a single feed toward vertical products and industrial-chain operations.
4. Once Users and Time Peak, Value Growth Is Reduced to Conversion Efficiency
庄明浩’s blunt formula still works: “users × user time × money equals value.” China’s user base is approaching its ceiling, while time spent has already undergone the extreme test of 2022; the most exploitable variables ahead are payments, advertising and ROI.
His view is that for at least the next 3-5 years, platforms will continue directing R&D and operating resources toward monetization efficiency. It sounds less romantic than content creation, but it is the unavoidable medium-term theme of a mature market.
At the end of 2023, he felt that platforms had almost no projectable directions left beyond combinations of overseas expansion, mini-games, short dramas and UGC. AI has reopened the option set through efficiency gains in existing businesses and exploration of new formats.
5. AI Offers Both “Drudgery in the Present” and “Poetry and Distant Horizons”
庄明浩 calls AI enhancement of existing businesses “the drudgery in the present”: once a path is found to improve ad efficiency, retention, time spent or activity, ROI is usually visible quickly, and the project does not need repeated explanations of why it is worth doing.
“Poetry and distant horizons” refers to new AI-native businesses. For Kuaishou, Kling’s video generation has reached interim milestones in revenue, reputation and overseas reception, and may expand from auxiliary creative into short dramas, dynamic comics and new subject matter.
His self-disclosure is also worth preserving: in the social and companionship business he works on, there is a clear direction for AI to improve core metrics, but as of Q1 2025, apart from AI companionship, there is still no clear answer to what “the farther thing” is.
6. The Deeper the Ad Evaluation, the Sharper the Traditional Bargaining
郑迅 traces the evolution of conversion objectives: early advertisers wanted only activation, then next-day and seventh-day retention, then payment, and eventually the ratio of LTV15 or LTV30 to CAC—ROI.
The problem with deep objectives is that high-quality users are scarce. Optimizing for deep targets may produce insufficient volume, while shallow targets fail to meet operating benchmarks. Ad platforms need profit, while clients need cheaper growth; their interests are not naturally aligned.
In the era of fragmented external-link data, platforms did not know what clients were truly optimizing for, while clients guessed at the rules for cold starts, volume allocation and “watering down.” 郑迅 summarizes the relationship with a joke: “Between us and the client, it’s like a never-ending guessing game.”
7. Native Closed Loops First Eliminate App Switching and Data Fragmentation
Kuaishou launched its first mini-game around 2021, but did not formally connect native in-app mini-game user acquisition until Q4 2023. The key change was not another entry point; advertising, consumption, payment and attribution data began aligning within one system.
Short-drama operators have profile pages and series collections. A clip or full episode can acquire feed traffic, after which users watch continuously and pay through IAA or IAP on the same platform. Organic and paid traffic can also work on the same content asset.
郑迅 calls the switch from Kuaishou to another app “the biggest pain point” in the ad system. After moving to a native funnel, activation costs alone can differ by nearly 10x, while complex event tracking, data backhauls and cross-app attribution are eliminated.
庄明浩 asks whether this reduces the need for companies to operate independent apps or WeChat mini-programs. 郑迅 broadly agrees. Card-binding rates across multi-platform mini-programs are no longer low, and payment is not the main obstacle; users’ willingness to install new apps continues to decline.
8. Unified Interaction Converts Experience Gains into a Second Layer of Cost Savings
郑迅 observes that some younger users keep 王者荣耀 permanently installed, while retaining only one other game: “finish it, delete it; delete it, reinstall it.” Installation friction makes app operations increasingly difficult and strengthens the value of instant in-app access.
Native short dramas can standardize payment buttons, episode switching and viewing history. In the past, different mini-programs placed swipe gestures and buttons differently, hurting viewing time and payment conversion; standardized interaction lowers payment costs again after activation costs.
Mini-game visuals and gameplay cannot be forcibly standardized. But when payment buttons and next-episode switching in short dramas are unified through a native funnel, user experience improves. Native integration therefore does not flatten content; it standardizes the transaction infrastructure around it.
9. ROI Bidding Prices High- and Low-Value Users Rationally
Traditional payment bidding may be willing to spend RMB100 or RMB300 to acquire one paying user, forcing the system to keep searching for high-value players. High-value users are scarce and have already been fought over repeatedly for years.
ROI bidding can pay more for high-value users and less for users who will spend only RMB1, as long as the latter’s traffic is cheap enough. Clients no longer optimize for a single “payment event,” but allow users of different values to meet the target return.
郑迅 says Kuaishou mini-games introduced ROI bidding around late Q1 to early Q2 2024. Penetration quickly reached 70%-80% and now exceeds 90%; app-based games, by contrast, still mix in payment bidding and other methods.
Short-drama IAA buying launched around Q3 2024, and clients shifted rapidly to ROI bidding within 1-2 months. The completeness of native-funnel data made the transition significantly faster than in external-link businesses.
10. The End State of Native Operations Is Not Traffic Buying but Multichannel Operations
郑迅 divides the path into several layers: once content is uploaded, it first receives organic distribution, then paid promotion, followed by distribution through influencers and agencies, and finally IP licensing, cross-category adaptation and promotion.
Influencer distribution extends the logic of film and television editing and secondary creation, compressing plots through “Little Handsome” and “Little Pretty” personas to attract users from new angles. It also creates new revenue sharing rather than merely directing traffic to the original production.
Agency distribution addresses the capability mismatch between content owners and traffic buyers. Companies strong in content can keep producing, while agencies with capital and buying capabilities but no shows of their own can promote other people’s content.
A further layer includes adapting novels into short dramas, combining short dramas with games, and coordinating mini-game launches, premium short-drama releases and holiday promotion. The platform wants to operate the full content lifecycle, not a single burst of ad spend.
11. Industry Deceleration Is Forcing Participants to Build Operating Capabilities
庄明浩 believes short dramas have grown into a market worth tens of billions of yuan and cannot continue multiplying several times every year. Companies that entered through an early traffic window must now pay the tuition for refined operations, product, marketing and industrial specialization.
郑迅 acknowledges that short-drama demand remains unmet. The issue may lie in the distribution system itself, or the corresponding supply may simply not have been produced. But incremental growth is more likely to come from refined content marketing than another nationwide traffic bonanza.
The same drama can acquire users through full episodes, edited clips, influencer creative or actor persona accounts. Once short-drama actors become popular, their personas, livestreams and continuing content become adjacent assets, a model familiar to a mature film and television industry.
12. Northeast Content Agencies Show the Real Path to Multiform Operations
庄明浩 uses agencies in Liaoning, Shenyang and Dalian as examples. Many teams began with live-show streaming, then moved into plot-based accounts before transferring their characters and livestreaming experience into serialized short dramas.
These agencies no longer rely on one content format or revenue source. They continue updating along a chain of plot, persona, livestream and short drama. Short-video infrastructure gives them a set of modular components for content consumption.
郑迅 adds that many former KOLs have started making short dramas, while variety-show formats are also entering the space. The trend is not platforms inventing new formats from scratch, but existing content capabilities naturally converging into a more complete operating chain.
13. IAA and IAP Represent Two User Groups, Not Just Two Payment Buttons
郑迅 calls both IAA and IAP “payment methods”: IAP pays with money, while IAA pays with time; watching an ad is equivalent to being “held hostage” for 30 seconds.
The two payment methods have clear breakout potential. Kuaishou short dramas typically start by buying IAP traffic and then add IAA; the gap narrowed from roughly 7 days in 2023 to 1-3 days, showing that producers increasingly accept the need to operate both user groups at once.
Users willing to pay contribute higher ARPU, while users willing to pay with time are more numerous. Kuaishou’s average time spent per user is about 120 minutes; adding another dozen minutes barely changes the overall ratio, but can unlock a sizable pool of ad monetization.
IAA and IAP are currently broadly equal in Kuaishou’s short-drama ecosystem. Mini-games are lighter and more casual, and IAA spend has at times exceeded IAP. A single payment model is no longer enough to describe these products.
14. The Real Challenge of Hybrid Monetization Is Product Design
The gaming industry has already blended IAA and IAP into a single experience, but 郑迅 notes that the two product types are led by different capabilities: IAP teams excel at numerical design, while IAA teams are stronger in gameplay creativity.
The more common model is IAP-led with a small amount of IAA, using ads to help some users continue playing and extend their time. Moving in the opposite direction—adding IAP to an IAA product—is harder.
Kuaishou has observed that IAA mini-game users are not categorically unwilling to pay when they cannot clear a level. Asking them to pay a small amount to remove ads may better match their immediate need than continuing to demand more ad views.
15. A Platform-Wide Ad-Free Pass Validated a New Payment Combination
Kuaishou first encouraged game clients to launch their own ad-free passes, then introduced a platform-wide pass. After paying, users can avoid ads across all Kuaishou mini-games for a period of time.
The platform still knows how many ads users skipped and can therefore split revenue with game clients based on their monetization value. Clients do not lose revenue because of the ad-free experience, while users play for longer.
郑迅 says users who bought the platform-wide ad-free pass increased their time spent by more than 50%. They ended up playing more and watching more ads as well.
On the short-drama side, Kuaishou has tested two designs: pay for an ad-free pass to watch the remaining episodes, or watch 5 ads consecutively in advance and then watch the rest of the story uninterrupted. Both create new exchanges between IAA and IAP.
16. Client Penetration Is Closer to the Truth Than a Single A/B Test
郑迅 believes the content-consumption industry is young and carries little historical baggage. If a new feature helps a drama or game scale, participants are willing to adopt it quickly. Once early hybrid-monetization games began scaling, peers immediately asked about penetration and demanded access.
The team evaluates products not only through experimental metrics, but also by whether clients “vote with their feet.” If an A/B test improves but penetration stalls, client experience is often divided: the feature may work only for certain products and still requires iteration.
By 郑迅’s estimate, the content-consumption sector may have more than a dozen major iterations each month that pass experiments and enter Launch Review, with many more minor features. Monetization is not a single model solving every problem; it is continuous repair of the entire chain.
17. The First Large-Model Recommendation Experiment Was Correct but Completely Useless
Kuaishou initially converted all user behavior into text and asked a large model what products the user needed. The answer was “Taobao, WeChat and Pinduoduo.” 郑迅 admits the model was not wrong, but the answer had no recommendation value and was nearly impossible to tune.
The redesigned first-layer Agent only describes what kind of person the user is. Kuaishou has roughly 120 minutes of platform behavior per user, not an omniscient view, but the model can infer interests beyond the logs like someone who has just had a brief conversation with the user.
The second-layer Agent reasons from those interests to what the user might like, allowing results outside Kuaishou’s supply and even permitting hallucinations. The third layer then selects the closest product from the platform catalog, sharply improving hit rates.
The failure reinforced a product lesson for both speakers: directly calling a powerful model is not the same as solving a problem. The task must be decomposed into steps that can be evaluated and optimized.
18. Reasoning Models Make Part of the Recommendation Black Box Visible Again
Traditional operators use explicit rules: someone who plays match-three games may play another match-three game, but not necessarily an SLG. A model may instead discover that users of “comeback” content are also willing to play hidden-object games—an association that is not obvious.
The same applies to short dramas, which cannot be matched only through surface tags such as urban or period settings. The interest distance behind mother-in-law conflicts, ancient warfare and light contemporary themes may differ from human category assumptions.
郑迅 says that after large models were added, “the black box became a little more white-box.” The system no longer mechanically converts “watched urban drama” into “recommend more urban drama”; it adds an interpretable layer of interest inference when behavior is sparse.
19. Dynamic Comics Meet the Conditions for an Independent Category
One of the verticals 郑迅 most wants to explore with clients in 2025 is dynamic comics. Even without dedicated operations, the platform already has large volumes of anime clips and secondary creation; Bilibili is a center of anime culture, and demand clearly exists, but short dramas have not yet covered this consumption.
Traditional anime episodes run 10-20 minutes and remain close to the long-video model. Just as viewing time migrated from television dramas to short dramas, it may migrate from anime series to shorter, lighter dynamic comics or comic dramas.
The first type is built mainly from comic panels, with only limited movement such as hair animation, plus TTS voiceover; it is light in style and suited to younger users. The second carries fantasy, mythology and large-scale scenes that are difficult to shoot in live-action short dramas.
Supply could come from anime and Chinese-animation teams, short-drama companies, novel writers and light-comedy short-video teams. 郑迅 is looking for “a flourishing supply,” while acknowledging that platform-side demand has not yet been fully released.
20. AI Can Push Dynamic-Comic Costs Below Short-Drama Costs
Based on industry discussions, 郑迅 estimates average short-drama production costs at several thousand yuan per minute. Rough dynamic comics may cost several hundred yuan per minute, while polished versions cost around RMB1,000-2,000.
Fantasy productions require studio shooting, special effects, universes and explosions—burdens traditional short dramas cannot afford. Generation technologies such as Kling may allow these difficult subjects to enter video consumption as dynamic comics.
庄明浩 places the shift within China’s distinctive industrial path. Japan has built a mature chain from light novels to manga, dynamic comics, animation, film and games; China may not copy the same subdivisions, but short-video infrastructure, AI and IP could jointly generate its own intermediate forms.
21. Platform Operations Will Keep “Splitting Categories” Like E-Commerce
庄明浩 cites a Taobao category operator who began with the anime category, which later split into animation and collectibles, before “谷子” emerged in recent years. His core OKR was essentially to identify one new category worth operating independently each year.
Content platforms will follow the same process. Once a vertical’s supply, monetization, producer specialization and user habits cross a threshold, it becomes worth separating from the broad feed and assigning dedicated products, tools and operations.
Dynamic comics therefore mean more than another content format. If they develop an independent supply-demand and transaction structure, the platform can build a new category node around them. Categories are usually more precise than broad sectors and more operationally efficient than individual SKUs.
22. The Next Mini-Game Question Is 30 Days and 3 Months, Not Day One
郑迅 admits that Kuaishou has not yet solved long-term mini-game ROI. A user paying on day 1 does not mean the user will play for 30 days or 3 months, nor that a “whale” will keep spending.
His sequence is to solve day-one ROI first, then move toward seven-day and longer cycles. IAA games are also becoming more sophisticated: design capacity has grown from roughly 10 hours to 30 hours, making it impossible for users to finish in 1-2 days and turning retention into the main bottleneck.
The weakness of a short-video platform is that users return each day to watch videos first and may not remember yesterday’s game. Kuaishou has saved game records in the sidebar and encouraged developers to use coins, items and other rewards to steer users toward a fixed entry point, gradually building a return habit.
23. AIGC’s First Commercial Role Is to Raise the Industry Floor
Kuaishou began experimenting with AIGC around 2023. Its most mature applications today are digital-human livestreaming and ad-creative production. 郑迅 says the platform already has tens of millions of yuan in daily ad spend backed by AI-generated creative.
This value is booked as advertising revenue rather than Kling revenue. It shows that generative AI’s contribution to a platform cannot be measured only through standalone product revenue; indirect gains to the existing advertising base also matter.
郑迅 repeatedly emphasizes that digital humans do not outperform top hosts, and AIGC may not beat the best editors. Their role is to solve the “floor problem”: fill slots when real people cannot go live, reduce repetitive creative and prevent mismatches between content and products.
Rules can also be written directly into the generation process, helping reduce common compliance violations. For small clients that are weak at creative production or media buying, consistent adequacy is more commercially valuable than occasional brilliance.
24. Conversational AI Is Turning Inefficient Lead Capture into Ongoing Operations
Traditional lead ads ask users to leave a phone number on a webpage and wait for a salesperson to call, with low efficiency. Kuaishou first promoted private-message marketing, allowing users and advertisers to talk a few more times before lead capture so both sides could assess fit.
Early scripted bots could answer only by rule. After large language models were fine-tuned for advertising scenarios, conversations became more nuanced and better suited to vehicle-model introductions, service coverage and basic inquiries.
The platform has accumulated substantial private-message marketing data, so it does not currently require every client to participate in training. Many small merchants have no customer-service staff; AI can first assess intent, answer common questions and support human follow-up.
Kuaishou’s externally offered products include 开创, an ad-creative platform; digital humans; the dialogue product 派; and UAX, an intelligent media-buying product. UAX can decide when to switch an ad campaign on or off and whether to use existing creative or generate new creative.
25. Large-Model Recommendation Helps New Products Cold-Start and Builds New Commercial Infrastructure
Ad behavior is sparser than content behavior: payment rates may be only a few per thousand or even a few per ten thousand, while the share of ads in the feed is constrained by the platform’s marketing-content ratio. Yet the system must protect client costs with limited signals.
The longer an old game runs, the more data it accumulates and the more stable its performance becomes, but delivery also keeps concentrating on the same users. New games face a difficult cold start. Ads cannot explore as freely as ordinary content because every experiment spends the client’s money.
郑迅 says large models produce different products from traditional models at the recall layer, increasing exposure for new games and products. With several thousand short dramas, roughly 1,000 mini-games and even more novels arriving each year, this exploration capability can shorten the time it takes new products to find their target users.
Kuaishou is also using large models internally to understand videos, landing pages, apps and real-world knowledge, then organizing products and users into layered categories such as legal consultation, property, inheritance and marriage. The goal is to consolidate reasoning currently scattered across algorithms and product managers into infrastructure, because “that’s just how the algorithm works” is ultimately meaningless.
26. The AI Application Battle Ultimately Returns to Scenarios, Workflows and Good Content
庄明浩 worries that general capabilities such as digital humans and creative generation will be absorbed by platforms as the water level rises, squeezing third-party service providers. 郑迅 preserves the uncertainty: “AI is far from settled,” and it is too early to say whether the market will become oligopolistic. DeepSeek changed the choice of base models, but application-level division of labor remains far from complete.
Platform AI often serves clients who are not good at media buying but want to advertise on Kuaishou, with the system moving from assistance toward taking over. Kling targets more creative video production. The two can share underlying technology but must be packaged around different problems, and Kuaishou will not give traffic preference merely because creative was generated by Kling.
The next challenge for digital humans is interaction. E-commerce is relatively advanced, and 郑迅 sees lead-generation livestreams as similar to e-commerce; show-style livestreams depend on dense, real-time exchange. For now, comment-section assistants and private messages can handle questions, while “Human in the Loop” fills the points that cannot yet be automated.
郑迅 ultimately abandoned a separate AI OKR because it would imply that only a small group was responsible for AI. His choice was to make every OKR consider AI, while insisting that product managers personally try tools such as GPT and Kling before returning to the question, “What problem are you actually solving?” 庄明浩 summarizes the next phase as a move from language to behavior: MCP lets Agents use tools, A2A connects multiple Agents, and script rewriting, video generation and editing gradually form measurable workflows.
The final calibration for content consumption is that monetization cannot degenerate into “pay-to-win.” 郑迅 argues that payment points can be deferred; good content may even make users ask to pay for an early screening of the finale. 庄明浩 cites repeat viewings of Ne Zha and the fact that the first achievement completion rate in Black Myth was not 100% as examples of emotional attachment. Good content can make users willing to pay; the more watchable a short drama, the more likely it is to generate payment, while good games can earn users’ time and money. That is the foundation distinguishing content consumption from pure content advertising.