Classification follows intended use. Marketing that sells harder can drag a wellness prototype into a device pathway, or create false-advertising exposure without a registration home. This article explains how digital health and software as a medical device teams in China should draft, lock and govern intended use statements, and how the NMPA classification system converts those statements into regulatory obligations.
Why intended use is the hinge of the entire regulatory framework
The Legal Rule
Why digital health teams in China must freeze SaMD intended-use language before marketing launch—classification, advertising, and liability notes for product counsel.
The Business Impact
Map the product, claims, distribution model and responsible entity to the relevant approval, quality and promotional controls before launch. Small commercial changes can shift which compliance obligations apply. Apply that to the facts of Lock SaMD Intended Use Before the Growth Team Ships WeChat Claims.
In China, as in most major markets, whether a software product is a medical device depends primarily on its intended use. The same algorithmic engine can be a wellness calculator, a non-device clinical decision support tool, or a Class II or III medical device depending on how it is described, how it is marketed and what the end user is told it does. The NMPA does not classify software by its code structure or data inputs alone; it classifies software by the medical purpose the manufacturer claims. This makes the intended use statement the single most important document in the regulatory file, and it explains why marketing teams must never write their own claims without legal and regulatory review.
The risk for digital health companies is that marketing language drifts. A product launched as a health education tool gains features over time, and the growth team begins describing it in ways that imply diagnosis or treatment. Each new claim moves the product closer to the medical device boundary, and at some point the claim itself becomes the basis for enforcement even if no NMPA registration was ever filed. The enforcement authority does not need to prove that the software actually performs a medical function; it needs to show that the product was marketed as performing one.
- Intended use is the hinge of NMPA classification, labeling and promotion
- Indications, population, environment, users
- NMPA five-step classification tree
Statutory framework
Medical device regulations
The Regulations on the Supervision and Administration of Medical Devices, issued as PRC State Council Order No. 739, define a medical device as an instrument, apparatus, implement, in vitro diagnostic reagent or calibrator that is intended for use in the diagnosis, prevention, monitoring, treatment or alleviation of disease, or for other medical purposes. Article 6 of the regulations requires medical devices to be classified and registered before sale. Article 81 establishes penalties for the unregistered manufacture or operation of medical devices, including fines calculated as a multiple of the illegal gains. Article 101 defines the scope of the regulations and confirms that software intended for medical purposes falls within the device definition.
Advertising and e-commerce law
The PRC Advertising Law Article 17 prohibits advertisements for non-medical products from claiming medical functions or using medical terms that could confuse consumers. The PRC E-Commerce Law Article 17 requires e-commerce operators to disclose truthful and complete information about their products and services. Together, these provisions mean that a health app that publishes diagnostic-sounding claims on its product page, in its WeChat articles or in paid advertising creates exposure under three regimes: advertising law, e-commerce law and the medical device regulations. The company that avoids device registration by claiming it is not a device must therefore also avoid every claim that sounds like a device claim, which is a much harder discipline than most teams expect.
AI software guidance
The NMPA Guidelines for the Classification and Definition of AI Medical Software, issued in 2021, apply the general classification logic to artificial intelligence products. The guidance confirms that AI software intended to support diagnosis, treatment or risk prediction for a specified medical condition falls within the medical device classification system, and that the classification is based on the intended use and the level of risk. The guidance also addresses the boundary between decision support and autonomous functions, with products that provide diagnostic or treatment recommendations generally classified at a higher level than products that only organise information for a clinician.
Enforcement context: the Haidian mini-program case
A 2022 administrative penalty issued by the Beijing Haidian Administration for Market Regulation illustrates the practical stakes. The operator of a WeChat mini-program that generated diagnostic risk scores was fined for operating an unregistered Class II software as a medical device. The enforcement authority examined the marketing claims of the mini-program, determined that the risk scoring function constituted a medical diagnostic function within the intended use analysis, and imposed the penalty even though the product had no physical hardware and was accessed through a consumer channel. The case is significant for digital health companies because it confirms that the NMPA and market regulation authorities treat software claims, not hardware, as the basis for classification, and that consumer-facing distribution through WeChat does not exempt a product from the device regime.
The intended use statement template
The practical deliverable is a bilingual intended use statement that every digital health product should maintain in a controlled document. The statement has two scopes, and the drafting rule is simple: the product cannot describe itself as doing more than its registration or exemption status permits.
Wellness and educational scope
For a product that is not registered as a medical device, the intended use statement should use language such as: the product provides general health information, educational content and lifestyle tracking; it does not diagnose, treat, cure or prevent any disease; it is not a medical device and should not be used as a substitute for professional medical advice. Every marketing asset, product description and in-app copy should be traceable to this statement. The company should also maintain a claim library that lists every approved claim, the evidence supporting it, and the version of the statement from which it derives.
Medical diagnostic scope
For a product that is registered as a medical device, the intended use statement must match the NMPA-approved registration document precisely. The statement should identify the intended medical purpose, the target population, the conditions or parameters the product evaluates, and any contraindications. Chinese and English versions must be consistent, because the Chinese version governs for domestic regulatory purposes while the English version is used for international coordination. Any change to the intended use requires a regulatory assessment, and a material change may require a new registration or a registration amendment.
- Draft intended use
- RA + clinical + product
- Run classification tree
- Confirm device class path
- Freeze statement version
The NMPA five-step classification decision tree
Companies that are unsure whether their product is a device should run a five-step classification assessment, documented in writing, before any marketing launch.
Step 1: State the intended use
Write the intended use in one or two sentences, in Chinese and English, using the templates above. The statement should describe what the product does for the end user, not what the technology does in technical terms.
Step 2: Identify the medical purpose
Determine whether the intended use involves diagnosis, prevention, monitoring, treatment or alleviation of disease, or the evaluation of physiological parameters for a medical purpose. If the answer is no, the product is likely outside the device definition, but the advertising analysis still applies.
Step 3: Apply the classification rules
If the product has a medical purpose, apply the NMPA classification catalogue and the AI software guidance. Software that provides diagnostic or treatment recommendations is generally Class II or III depending on the risk level. Software that only stores, transmits or displays clinical data without analysis may fall outside the device definition, but the distinction is fact-specific.
Step 4: Assess the enforcement boundary
Even if the technical classification is uncertain, assess how a market regulation authority would read the marketing materials. The authority reads claims, not code. If the product page says risk assessment or diagnosis, the authority will treat that as the intended use regardless of the technical architecture.
Step 5: Document the decision
Record the assessment, the intended use statement, the marketing claims review and the decision whether to file a classification consultation with the NMPA. The documentation protects the company in an enforcement scenario by showing that the classification decision was made deliberately rather than ignored.
Operating governance for product and marketing teams
The intended use statement is only useful if it governs behaviour. Digital health companies should require medical-legal approval for any claim that diagnoses, triages or treats, and should implement version control on all marketing materials. If the software's AI outputs influence clinical decisions, the company should plan model versioning, hospital override logs and clinician training from day one. The growth team should be trained that the phrase intended for general wellness and the phrase intended for the diagnosis of diabetes are not interchangeable marketing options; they are different regulatory products with different timelines and different liabilities. Where the product evolves from wellness to medical function, the company must decide deliberately whether to pursue registration or to constrain the product, and that decision must be documented in the same file as the intended use statement.
Finally, companies should monitor the competitive and enforcement landscape. When a competitor is penalized for unregistered software claims, the enforcement theory used in that case should be re-run against the company's own product. The Haidian case and the AI software guidance will continue to develop, and the intended use statement should be reviewed at least annually and whenever a new feature or new marketing channel is added.
Read next: SaMD & AI · Internet hospital ads · Device advertising
Discussion
Share experience or questions about this topic. This is a public discussion — not legal advice. Do not post confidential case details.
Have a question after reading? Leave it here, or Ask a Lawyer for a free initial intake.
Comments are moderated. China Legal Portal is a directory and information resource; no attorney–client relationship is formed by posting here.
