2026 AI Marketing Casebook: How Brands in China Are Turning AI from a Tool into a Storytelling Engine

What marketers need to know about the new AI playbooks, the case lessons, and the trust risks to manage

A marketing reset: from “how to use AI” to “how to create with AI”

Source: endata

As generative AI moves into more mature application, the marketing logic is changing with it. The report frames 2025 as the year AI moved beyond assistance and became the creative backbone of campaigns, capable of building complete narratives and driving emotional response. In other words, AI is increasingly treated as content itself, not just content support.

Source: endata

This shift is happening alongside fast market growth. The report estimates China’s AIGC market reached RMB 25.7 billion in 2025, and highlights a 2026 global AIGC compound annual growth rate expectation of 60%. At the same time, usage is normalising in everyday behaviour: AI native app users exceeded 120 million, with monthly per capita usage time above 133 minutes.

On the enterprise side, brands are deploying AIGC for cost reduction and efficiency gains, with the report noting that some companies have shortened new product development cycles by more than 20% through AIGC enabled workflows.

Source: endata

AIGC is shifting from tool enablement to value co creation, with advertiser adoption surpassing 50% in 2025

AIGC is no longer an edge capability. It is rapidly becoming normal marketing infrastructure. In 2025, AIGC penetration among advertisers surpassed 50%, signalling that “testing” is giving way to routine deployment in real campaigns. More specifically, 53.1% of advertisers have already used AIGC in creative content, which supports the report’s core argument that AI is moving from a backstage assistant to a visible driver of brand narrative.

Source: endata

The operational shift is clearest in video. Nearly 20% of advertisers now rely on AI for more than half of the steps in video creation, showing that AI is starting to reshape workflows, not just outputs. The slide also indicates that AI marketing content has exceeded 10%, reinforcing the direction of travel: more AI content, faster cycles, and higher expectations for quality control.

For brands, the implication is straightforward. When AI becomes mainstream, differentiation moves from “using AI” to “using AI well”, with stronger standards for realism, accuracy, and brand safe storytelling.

The three-layer AIGC ecosystem: why AI marketing in China is now a “stack”, not a single tool

Source: endata

China’s AIGC is ecosystem consisting a clear three-layer structure: upstream foundational technology that provides technical support, midstream vertical applications focused on specific industries and functional scenarios, and downstream end-user applications that directly face consumers.

In practical terms, this means brand performance with AI increasingly depends on how well you connect capability → workflow → distribution across the chain.

1) Upstream: foundational technology (sets the creative ceiling)

Upstream is the “infrastructure and core capability” layer: pre-trained models, foundation models, and compute/hardware resources. For marketers, upstream choices shape:

• Quality limits (realism, character consistency, motion stability, multi-modal output)

• Unit economics (how expensive it is to generate at scale)

• Risk baseline (how controllable outputs are; what safety/guardrails exist)

  
  

If your ambition is cinematic storytelling or high-realism video, upstream becomes the quiet determinant of whether teams can hit a “looks real enough” bar without endless manual correction.

2) Midstream: vertical applications (turn capability into repeatable marketing production)

Midstream is where AI becomes usable marketing machinery: AIGC tools and application layers that operationalise “generation” into production. Endata positions this layer as the vertical application layer, focused on industry needs and functional scenarios, with AIGC at the centre.

This is the layer that determines whether AI stays a one-off experiment or becomes a scalable system, because it’s where brands build:

• Workflows (brief → scripts → storyboards → assets → video → localisation)

• Brand consistency (style rules, reusable templates, asset libraries)

• Governance-by-design (QC gates, fact-check steps, approval accountability)

  
  

In other words: the midstream layer is where “AI content” becomes “AI operations”.

3) Downstream: end-user applications (where content becomes participation and reach)

Downstream is the terminal application layer that faces C-end users. This layer determines whether AI is simply a production shortcut,or a distribution engine, ßßbecause it’s where AI content is:

• Consumed (feeds, communities, short video environments)

• Shared (social mechanics and cultural meme loops)

• Co-created (templates/effects that turn users into creators)

  
  

For many categories, the biggest unlock isn’t just “making more content”; it’s designing AI content so it can be adopted, remixed, and spread inside platform-native behaviours.

How brands should use this framework

A quick diagnostic that often clarifies strategy:

• If you’re struggling with realism and quality, your bottleneck is usually upstream.

• If you can generate assets but can’t scale consistently, your bottleneck is midstream workflow + governance.

• If you can produce at scale but engagement is flat, your bottleneck is downstream distribution + participation mechanics.

  
  

That’s the shift this chart signals: in China, AI marketing outcomes are increasingly a stack decision, not a single creative tool decision.

The productivity engine: AI expands the scale and ambition of storytelling

The first mainstream paradigm is AI as a productivity engine: using AI to unlock visuals and narratives that would be too expensive, too slow, or too complex to produce through traditional methods.

Source: endata

A signature case here is By Health’s 30th anniversary brand film, released in December 2025 and produced entirely with AI. The campaign used AI to build large scale cinematic scenes and demanding environments, positioning AI as the solution to high cost production barriers while aiming for a “real shoot” logic in look and feel. A key lesson is that the audience benchmark is not “is it AI” but “does it feel real enough”. The report shows social discussion praising craft and spectacle, while also flagging that visible AI artefacts remain a trigger for negative reactions.

Another productivity case is Yili Jinlingguan’s AIGC animation short series, launched in July 2025. The report notes that the series reached over 100 million plays within 10 days, with total plays exceeding 126 million. It also highlights the operational benefit: an AI plus human workflow compressed production from around six months to under two months for a six episode series, demonstrating how AI can turn longer form storytelling into a faster, more repeatable content format.

What this means for brands: if you use AI to go bigger, the craft standard rises. Quality control becomes the differentiator, especially for human faces, motion, and physical realism.

The narrative protagonist: AI becomes the content, not the backstage tool

The second paradigm is AI as the narrative protagonist. Here, AI is not hidden. It is the point. Brands use AI native visual language, AI characters, or AI logic to make the campaign itself feel like a statement about modernity, creativity, and participation.

Source: endata

Lenovo’s back to school campaign is one example. The report describes how AI enabled creative shorts used the terracotta warrior motif and modern student life to create a cultural remix story world. The innovation is not only in production, but in how the campaign uses AI to make the brand’s product story feel like entertainment rather than an advert.

Beijing Tong Ren Tang’s AI short film series offers a different angle: AI is used to reconstruct traditional narratives into a modern, playful style, combining myth and cultural icons with a younger content rhythm. The report also notes that the wider campaign extended into platform challenges, showing how AI content can be designed as both a hero asset and a participation trigger.

What this means for brands: if AI is the story, then the creative idea must be strong enough to stand on its own. AI cannot replace concept. It can only amplify it.

The co creation tool: AI lowers the barrier and turns audiences into producers

The third paradigm is AI as a co creation tool, where the marketing system is designed to generate massive volumes of user content, not just impressions.

Source: endata

Mixue Bingcheng’s collaboration with Kuaishou is a clear example. The report highlights an AI effect that lets users upload photos and receive personalised story continuations and AI generated videos. This is not “one template for everyone”. It is designed for “one person, one version”, which is why it scales in participation and repeat usage.

Douyin Mall’s AI themed activity also fits this model, using AI tools to visualise user ideas and turn creativity into an interactive loop that benefits both consumers and merchants. The report positions this as a practical method to collect real demand signals, while also making users feel like creators rather than targets.

Source: endata

What this means for brands: co creation works when the output feels personal, shareable, and low friction. The AI interface is part of the product experience, not just a campaign gimmick.

The risk chapter: realism, accuracy, and content safety are now brand liabilities

As AI use expands, the report is explicit about the downside risk. It highlights three recurring problem areas that can quickly damage trust.

Visual distortion and “uncanny” aesthetics

The report notes intense user discussion around distorted AI visuals, using examples where obvious physical errors triggered public discomfort and ridicule. The core lesson is simple: poor AI craft does not look innovative, it looks careless.

Content inaccuracy and factual mistakes

The report describes cases where AI generated claims were challenged and disproven, creating reputational blowback. When AI produces false information, the blame does not sit with the model. It sits with the brand approval process.

Value and safety red lines

The report also flags controversies where AI generated content created ethical concerns or crossed sensitive boundaries. As AI speeds up production, it also increases the risk of teams publishing content that has not been properly checked for social context.

What this means for brands: AI content needs a stronger governance layer than traditional content, not a weaker one.

What brands should do next

To win with AI marketing while protecting trust, brands should treat AI as a system, not a shortcut.

Five practical shifts:

• Build a quality standard for realism, especially for people, hands, motion, and physical logic

• Put factual validation into the workflow, with clear responsibility and final human sign off

• Design AI experiences around emotional payoff, not just novelty

• Use co creation when personalisation is real and sharing is effortless

• Define content safety rules in advance, including sensitive topics, minors, and cultural boundaries

Closing thought

AI marketing in China is moving fast, and the advantage now comes from using AI well, not simply using it. If you would like support building a scalable system that balances speed, craft, and governance, reach out to our team. 

You can also explore our companion guide, AI Native Marketing in China: How Xiaohongshu, Baidu, WeChat and Douyin Are Rebuilding the Funnel, for a practical breakdown of how China’s major platforms are rebuilding the funnel.