Providing customer service in China involves much more than translating an existing help centre into Chinese. Users need to be able to reach support through channels that work reliably in Mainland China, receive assistance in Mandarin and resolve account, payment, order, technical and after-sales issues.
This is particularly important for both foreign digital products and international brands selling online in China. An app may be fully localised and available through the Chinese Apple App Store or local Android app stores, but users will still encounter questions involving login methods, Chinese payments, device compatibility, subscriptions and refunds. A store on Taobao, Tmall, JD, Douyin or RedNote must also handle pre-sales questions, order enquiries, returns, complaints and platform-specific service targets. In both cases, customer service needs to be designed as part of the China market entry strategy rather than added after launch.
China has also introduced a new national standard for its customer service. GB/T 47746-2026, Customer Contact Service—Requirements for Collaboration Between Human and Intelligent Customer Services, was published on 25 May 2026 and takes effect on 1 September 2026. This guide will explain everything you need to know about customer service in China.
Customer service in China is not limited to telephone support. Depending on the product, it may include:
The correct service structure will therefore differ between a consumer app, a mobile game, an enterprise SaaS platform and a store on Taobao or Douyin. An app operator controls its own support interface, whereas an e-commerce merchant will normally work inside the platform’s messaging, order and dispute systems. What they share is the need for a visible service entrance, trained personnel, clear escalation rules and a complete record of the interaction.
Chinese users are accustomed to communicating with companies through instant messaging and in-platform service interfaces, only in Chinese. On an e-commerce platform, a user can normally contact the seller directly from a product or order page. For an app, directing the user only to an overseas email address often creates unnecessary friction.
Foreign businesses also need to consider whether their existing support tools work reliably in Mainland China. A help centre may load slowly, an embedded chat tool may depend on blocked or unstable third-party resources, and an overseas support team may not have access to the app’s Chinese account, payment or distribution systems. Different time zones should also be take into account if the team is based outside of China.
Customer service is governed by a combination of laws, regulations, national standards, sector rules and platform agreements.
GB/T 47746-2026 sits within this wider framework. It provides an authoritative operational benchmark and may be incorporated into procurement requirements, contracts, internal policies or industry assessments. Separate platform rules can also impose response targets and other requirements directly on merchants.
GB/T 47746-2026 applies to organisations that provide both human and intelligent customer service. It covers inbound and outbound interactions conducted through voice, text and video, as well as service completion, ticket handling, information security and continuous improvement.
The standard states that AI should handle high-frequency, standardised questions, while human agents remain necessary for complex requirements, emotional communication and special situations. In its official interpretation of the standard, the State Administration for Market Regulation identified arbitrary task allocation, broken handovers, poor data synchronisation and inconsistent service quality as the problems the standard is intended to address.
Organisations should establish appropriate human-service capacity, internal procedures, an emergency mechanism for disasters or system failures, a customer-information and ticketing system, a knowledge base, training for human agents and a process for regularly improving the intelligent service.
The standard does not contain a separate chapter for mobile apps, but its requirements are channel-neutral. They apply to native apps, games, SaaS products and Mini Programs whenever human and intelligent customer service operate together. In practice, app operators should:
These measures are especially important for apps involving payments, healthcare, transport, security or other time-sensitive services.
The appropriate allocation will depend on the product and the quality of its knowledge base. The following table provides a practical starting point.
| Generally suitable for intelligent service | Should normally be handled by a human |
| Basic instructions for using the app | Repeated technical problems that remain unresolved |
| Supported devices and operating systems | Unusual device-specific compatibility problems |
| Known error messages with verified solutions | Unknown or serious errors |
| Standard password-reset instructions | Account access remains blocked after standard recovery |
| Order, delivery or service status | Contested charges, refunds or compensation |
| Standard subscription and cancellation information | Lost data, content or virtual items |
| Store opening hours and standard return policies | Suspected account compromise or information leakage |
| Routine stock and promotion questions | Complaints involving product or personal safety |
| Standard invoice instructions | Highly personalised or ambiguous requests |
| Service notifications and reminders | A customer explicitly rejects AI or requests a person |
High enquiry volume can justify routing more standardised work to AI, but it does not justify preventing access to a human agent. The aim is to use AI to increase capacity while preserving an effective route for situations that require judgement, empathy or authority.
GB/T 47746-2026 requires intelligent customer service to provide a human-service entrance. This can be triggered through customer keywords, a menu option, a dedicated button or key, a voice instruction or an automatic system rule.
Expressions such as “转人工”, “人工客服” or “联系人工” (all meaning “transfer to human customer support”) should initiate the transfer process. Voice systems should provide a consistent instruction and move into the transfer process once the relevant phrase is recognised.
During the handover, the system should provide queue information or an estimated waiting time. It must also synchronise the relevant identity details, interaction history, problem description and ticket status. Once the transfer is complete, the customer should not be required to repeat information already provided.
The reverse transfer is more restricted. A human agent may send a routine enquiry to intelligent service, but the customer should first be clearly informed and must consent. Workload alone should not be used to move a customer back to AI without their knowledge.
The appropriate channels depend on whether the business controls its own customer interface or operates through an e-commerce platform.
All channels should connect to a consistent ticketing, human-transfer and escalation process.
Operating within WeChat or another third-party platform does not remove the product operator’s responsibility for customer-service quality, data protection and effective human escalation.
Customer service on an e-commerce platform normally involves several parties:
Contracts and internal procedures should define who handles each type of enquiry, who can approve refunds or compensation and who communicates with the platform during disputes.
Where both AI and human customer service are used, GB/T 47746-2026 should be applied throughout the customer journey:
The platform may provide the chatbot and messaging system, but the merchant remains responsible for configuring its knowledge base, staffing, operating hours and escalation procedures. Platform tools support compliance; they do not establish compliance by themselves.
Alibaba provides AI customer service for merchants through Dianxiaomi (店小蜜). In March 2026, Alibaba reported that the tool was being tested by 200,000 merchants and was being upgraded from scripted chatbot responses towards agents capable of handling conversations and completing tasks. Alibaba Group
Merchants should use AI for standard questions about stock, logistics, promotions and return policies while retaining human capacity for unusual promotion combinations, refund disputes, product problems and safety issues. Knowledge bases and staffing plans should be updated before 618, Singles’ Day and other major campaigns. To learn more, read our complete guide to Tmall e-commerce.
JD merchants use Jingmai (京麦) and Jingdong Dongdong (京东咚咚) as part of the merchant-service environment. JD also provides intelligent customer-service, work-order, knowledge-base and agent-assistance technology through JoyAI.
Responsibility will depend partly on whether the product is sold directly by JD or by a marketplace merchant. Standard order and delivery enquiries may be automated, while warranty exceptions, compensation, product-quality complaints and complex after-sales cases should be directed to the responsible human team.
To learn more about selling on JD.com, read our full guide.
Douyin merchants use Feige (飞鸽), which includes chatbot and human-transfer configuration. Douyin’s learning centre provides specific guidance on configuring transfer from the Feige robot to a human agent and monitors merchant service indicators such as the three-minute reply rate.
AI can help absorb repetitive questions generated during livestreams, such as product links, coupons, stock and delivery. It should not become a barrier when a livestream enquiry develops into a complex order, return, product-claim or safety complaint.
To learn more about how to sell on Douyin, read our full guide.
RedNote merchants have access to the Qianfan Customer Service Workspace and platform intelligent-service functions. Customer journeys can move between content, private messages, product recommendations and store transactions, making consistent context particularly important.
Many RedNote enquiries are also open-ended. Questions about skincare suitability, fashion, appearance, lifestyle or product performance may not have a single standard answer. These interactions should not be fully automated unless the knowledge base can support a reliable answer, particularly where health, safety or regulated product claims are involved.
To learn more about how to sell on RedNote, read our full guide.
Common mistakes include relying only on an overseas email address, providing support only in English, hiding human service behind a chatbot and failing to retain context when users move between an app, WeChat, an e-commerce platform and a human agent.
Other problems include using machine-translated scripts without product localisation, leaving the AI knowledge base unchanged after a product release or shopping campaign, providing insufficient Beijing-time coverage, allowing AI to handle sensitive or high-risk issues, and giving overseas teams broader access to customer records than they require. E-commerce merchants may also focus on reply speed while ignoring whether the issue was actually resolved, or assume the platform’s default chatbot and return settings require no further configuration.
Businesses should also avoid assuming either that all customer service requires a B24 licence or that no licensing questions apply. The correct answer depends on whether the company is serving its own users, acting for third parties and operating telecommunications numbers and call-centre infrastructure.
Finally, outsourcing does not mean the service can be ignored. The foreign company still needs to train the provider, supply current documentation, assign escalation contacts and review performance.
A practical implementation process is:
This process, particularly identifiying keyword triggers for certain automated responses or escalation to human services, will help optimise the user experience. The process should also be repeated whenever the product, pricing, payments, platform configuration or applicable rules change.
As part of AppInChina’s customer service solution, we can help by:
Contact us to book your free consultation call and establish a customer-service solution for your app, website or e-commerce business in China.
