With the rapid development of the digital economy, AI and platform-based business models, the traditional system of examining trademark similarity based on the classification of goods and services is facing structural challenges.
This article examines the issue from a legal perspective, explores practical dilemmas and emerging trends, and offers observations on how the definition of similarity for trademark-designated goods and services should evolve amid rapid change.
In China, the trademark examination process determines similarity between goods and services based on several core factors, including:
These criteria originate from the fundamental legislative goals of trademark examination:
Overall, the traditional criteria for assessing similarity have three features.
First, trademark examiners rely on a static classification system. In China, the assessment of similarity usually relies on the Chinese Classification Book as a fundamental reference point. China adopts a subclass system in which goods or services falling into the same subclass are generally considered identical and/or similar. This means that even if the goods or services belong to the same class, they may be deemed dissimilar due to differences in subclass.
Second, trademark examiners assume relatively clear boundaries between goods and services, which are generally viewed as two separate types of commercial activity.
Third, examiners use the ordinary consumer’s cognition as the benchmark. Whether confusion may occur is still grounded in a relatively stable consumer structure and market perception.
However, with the rapid rise of AI, platform-based business models and virtual industries, these traditional standards for assessing similarity seem increasingly insufficient.
Social and technological change is affecting the stability of China’s traditional classification system.
Increasingly, the boundary between goods and services is being blurred due to the shift from simply “selling products” to “providing integrated solutions”. This has caused a strong overlap between items that traditionally belonged to different classes.
For example, home furnishing companies now sell physical products while also offering full-house design and installation services. Elsewhere, the automotive industry covers a wide range of activities, including vehicle manufacturing, autonomous‑driving software and mobility service platforms.
As a result, the traditional distinction between 'goods' and 'services' no longer reflects the actual connections in today’s market.
In addition, rapid AI development is accelerating cross‑industry integration, given that AI technologies are highly transferable and naturally cut across industry boundaries.
For example, image‑recognition technology can be used in many fields, including healthcare, security, quality inspection and e‑commerce. Large language models can support education and training, advertising creativity, software development assistance and administrative services.
This means that similarity assessments can no longer rely on rigid 'goods versus services' classifications. Instead, they need to consider the logic of the underlying technology and the user experience that it creates.
Elsewhere, China is seeing a rise in 'ecosystem-based competition' and platform economics. Major internet platforms often operate across multiple sectors (eg, entertainment, communication, payment, retail, logistics and cloud services), forming complex business ecosystems.
Consumers generally view these companies as 'naturally' capable of operating across different fields. Brand recognition is also no longer tied to a single product but takes on the nature of an 'ecosystem brand'.
Consequently, classifications of goods and services that were traditionally considered unrelated may, in a platform context, lead to confusion or mistaken association due to expectations of brand extension.
Trademark examination in China currently faces several key structural challenges.
The Chinese Classification Book often lags behind rapid changes in modern business. Although the book and Nice classifications are updated periodically, business innovation happens continuously and in real time. As a result, there is an inevitable time gap and a growing mismatch between trademark classification and the characteristics of businesses.
For example, should virtual dialogue characters used in online English‑learning platforms (featuring live people or cartoons) belong to education services (Subclass 4101) or entertainment services (Subclass 4105)?
In the absence of established precedents and clear classification guidance, trademark examiners inevitably reach inconsistent conclusions when assessing similarity. In practice, they must often rely on supplementary explanations, relevant precedents or even cross‑industry analysis to make decisions that align with today’s market reality.
Furthermore, different examiners or authorities may have varying understandings of 'relevance'. For example, are 'payment software' and 'financial services' similar? Should an autonomous‑driving system be seen as part of 'automobile manufacturing' or as an independent software‑related service?
Currently, none of these questions have universally accepted answers. They require dynamic assessment based on market perception and industry practices.
Changing consumer perceptions also influence the assessment of likelihood of confusion. Young consumers today are familiar with cross‑industry operations and brand collaborations, and they tend to accept them easily. Examples include:
These cross‑industry offerings and co‑branded products are no longer surprising to the public, which means that consumer perception is constantly evolving. This evolution affects how consumers interpret brand associations and, in turn, impacts the likelihood‑of‑confusion analysis.
Although current trademark examination standards have not fully departed from the traditional legal framework for assessing similarity, it is encouraging to see a growing number of cases that break through cross-class and/or cross-subclass classifications.
In (2023) Yi Zi No 0000098660, the China Trademark Office issued an opposition decision on the opposed mark JIZHONG (in Chinese) and JIZHONG and device (Application No 54716321), which designated insecticidal veterinary washes and chemical preparations for veterinary purposes in Class 5.
The cited marks were:
The trademark office found that the goods designated by these marks are similar in terms of their:
Since the designated goods were considered similar, and the distinctive element “JI ZHONG” in the opposed mark and cited marks was essentially the same, the concurrent use of both marks was likely to cause consumer confusion or misunderstanding. Therefore, the office found that the marks constituted similar trademarks used on similar goods.
In (2016) Jing 73 Xing Chu No 4951, the Beijing IP Court issued an administrative judgment, which found that the disputed mark and the cited mark constituted similar trademarks used on identical or similar goods.
The cited mark AVAILA (Registration No 1099573) designated Class 5 and was approved for use on medical animal feed additives in Class 5.
The disputed mark AN WEI LE in Chinese and AVAILA (Registration No 6947874) designated Class 31 and was approved for use on animal feed, byproducts of processed cereals for animals and pet food.
Although there is no cross-check in the classification book, the court again found that the two classes of goods were closely related in terms of:
If the marks were used on these goods, the products were likely to mislead the relevant public into believing that the products were associated, thereby causing confusion regarding the source of the goods.
In (2020) Yi Zi No 0000143294, the China Trademark Office issued an opposition decision, having found that two marks had designated goods that shared certain similarities in terms of function and purpose, and were therefore considered related goods.
The opposed mark CJSJ (Application No 34683146) designated electric locks and biometric fingerprint door locks in Class 9.
The cited mark CJSJ (Registration No 3645765) designated nozzles of metal and metal building components in Class 6.
Since the opposed mark was identical to the cited mark, it was likely to cause confusion among consumers, the office found.
In Shang Ping Zi [2026] No 0000015589, the Trademark Review and Adjudication Board issued an invalidation decision regarding:
The board found that the goods designated by the disputed mark were similar to those designated by the cited mark.
In addition, it found that the marks were similar in terms of textual composition and pronunciation.
The coexistence of the two marks on similar goods was likely to cause confusion and misunderstanding among the relevant public regarding the origin of the goods, the board stated.
Therefore, the marks constituted similar marks used on similar goods.
In Shang Ping Zi [2025] No 0000367889, the Trademark Review and Adjudication Board issued an invalidation decision regarding:
The board found that the goods designated by both marks were closely related in terms of function, purpose and distribution channels. The coexistence of the two marks on these closely related goods was likely to cause confusion among the relevant public regarding the source of the goods.
These decisions offer a glimpse into several emerging trends in trademark examination and judicial practice.
It seems that the Chinese Classification Book will no longer serve as the core basis for assessment, but merely as a reference tool. Examiners will place greater emphasis on:
In addition, assessing similarity should return to the fundamental purpose of trademark law – that consumers should not be misled.
Consumer perception will become the key factor, and the criteria for assessing likelihood of confusion should be more dynamic and behaviour-based, rather than limited to classification judgments.
In the age of AI, technical attributes and data correlations will be increasingly used to evaluate similarity. Examiners may ask the following questions:
For large entities with significant brand extension capabilities in the market (eg, Alibaba, ByteDance, Meituan and Tencent), examiners may apply a higher standard to assess likelihood of confusion to reflect their actual influence.
Examiners should also explore approaches that distinguish between so-called 'ordinary' and 'ecosystem' brands.
Against the backdrop of today’s digital, intelligent and platform-based industrial landscape, traditional mechanisms for assessing the similarity of goods and services are facing systemic challenges.
Future trends can be explored from the perspectives of consumers, technological advancement, industrial convergence and corporate business ecosystems.
Assessing similarity is no longer a simple 'systematics' issue, but rather a complex problem that integrates legal, economic, technological and social perspectives. Only by continuously updating legal thinking and understanding the business logic of the new era can we make more accurate and fair determinations of trademark similarity.
This article was first published on World Trademark Review in April 2026.
Author: Jojo Lu