September 8, 2026 · 14 min read · By Logo China Pro

China Trademark Subclasses: Why One Nice Class Isn't Enough

Here is the mistake Logo China Pro sees roughly once a week from foreign applicants.

A company sells phone cases, so it files in Class 9 and assumes every gadget, app, and charger is now locked down. Six months later a rival starts selling a copycat-branded smartwatch app, and the original brand files a complaint only to discover -- too late -- that "Class 9" in China does not mean what it meant back home.

The trademark is valid. The class number is correct. And the protection still does not reach the product being sold.

This is the single most misunderstood part of Chinese trademark practice, and it costs foreign brands more than almost any other filing error. China does not just use the 45 Nice classes. It splits every class into a second, hidden layer of subclasses -- officially called "similar groups" -- and a registration only covers the specific groups and items a filer actually ticks. Filing a class without understanding its subclasses is like buying a house on paper and later discovering you only own one bedroom.

The scale of the misunderstanding matters because the playing field is so crowded. By the end of 2024, China reported more than 47.6 million valid registered trademarks, with nearly 6.8 million new applications filed in that year alone. In an environment that dense, the difference between a correct filing and a nearly-correct filing is the difference between a usable right and a paper decoration.

One Class Is Not a Box. It Is a Grid.

Every country in the Nice system shares the same 45 top-level classes. Goods sit in Classes 1 through 34; services sit in Classes 35 through 45. That part is identical from Washington to Munich to Beijing -- which is exactly why a US applicant and a Chinese applicant can spend an entire project thinking they are speaking the same language when they are not.

What China adds on top is a document foreign filers rarely read before their first application: the Classification of Similar Goods and Services. It is built from the international Nice Classification but rebuilt around how Chinese examiners actually decide cases. The Chinese version runs on a four-level structure:

1. Class -- the familiar 1 to 45.
2. Similar group -- a four-digit code where the first two digits repeat the class and the last two name the group. Code 2507, for example, means Class 25, group 07: footwear.
3. Section -- some groups are further split into parts (the official text numbers them one, two, three, and so on). Goods inside one section are presumed similar to each other; goods in different sections are not.
4. Item -- the specific good or service actually applied for, usually written as a six-digit code. A leading letter "C" marks items that exist in the Chinese list but not in the international Nice list.

Read that again, because that second level does not exist in the United States or in the European Union.

A US applicant works off the USPTO's ID Manual and picks a goods description. An EU applicant picks directly from EUIPO's Nice-based list. Neither system has a "similar group" tier sitting between the class and the item. In China, that middle tier is where protection is actually decided. Miss it, and a filer can own a registration in the right class that protects nothing it sells.

What a Similar Group Actually Does

The similar group is not bureaucratic decoration. It is the unit Chinese authorities use to answer the one question every trademark system ultimately has to answer: are these two goods similar enough that consumers would be confused?

That question decides three things in China:

1. Whether your application is refused. An examiner searching for a conflict does not search an abstract "Class 9." The search runs group by group, with cross-references on top. A prior mark in the same group is a near-automatic citation; a prior mark in a distant group often is not.
2. Whether you can stop an infringer. Under China's Trademark Law, infringement turns on whether the accused goods are identical or similar to the goods actually covered by the registration -- and that comparison is read off this same Classification. If a copycat sells in a group your registration does not cover, your infringement claim is weak to begin with.
3. Whether you can file against a squatter. In an opposition or invalidation, the first thing examined is whether the squatter's goods and the senior brand's goods are similar. That, too, is a group-level judgment.

In other words, the similar group is not a filing nicety. It is the exact boundary of the right you are buying. Everything a brand wants to do with a trademark -- keep a rival out, pull a fake listing, fight a squatter -- works or fails at the level of the group, not the level of the class.

The Ten-Item Cap: A Forcing Function, Not a Bug

Here is the second thing that surprises foreign filers: in China, an applicant does not list a whole class heading. The applicant lists specific items, and a single class application covers ten items by default. Every item beyond ten costs extra.

The official numbers are public. Per CNIPA's fee schedule, an e-filed application costs RMB 270 per class, covering the first ten goods or services, with each additional item adding RMB 27. A paper filing is RMB 300 per class, with RMB 30 per extra item. At current rates, the base fee works out to roughly USD 40 for a class -- one reason a defensive, multi-class strategy in China is genuinely affordable, and a reason there is little excuse for leaving a core group uncovered.

But the ten-item cap does something useful if read correctly: it forces a filer to decide what it actually sells.

• Pick precisely. CNIPA wants the standard item name, not a broad category title. "Cosmetics" is not a valid Chinese filing entry; specific items like facial cleansers, face masks, or make-up preparations are. Vague language gets a correction notice and costs time.
• Pick defensively. The ten slots should mix core products sold today with adjacent items likely to sell tomorrow, because whatever is not covered, someone else can register.

The cap also cuts the other way. Under the three-year rule, any registered mark can be challenged for non-use on specific goods if it has not been genuinely used on them for three consecutive years. Load a filing with ten items the brand never plans to sell, and a competitor can later have those items carved out of the registration. Under-fill it, and a squatter fills the gap. There is a reason experienced Chinese filers treat the ten-item list as strategy rather than paperwork -- a "5 + 3 + 2" shape of five core items, three likely items, and two defensive slots is the common working template.

How a Class Breaks Apart, in Practice

The abstract structure only lands with examples, so here are the ones foreign brands keep running into.

Class 25: Footwear Is Not "the Whole Clothes Class"

Class 25 covers clothing, footwear, and headwear. But inside it, group 2507 is footwear specifically -- sneakers, slippers, sandals, and the like -- while clothing proper lives in groups like 2501, and infant textiles sit in 2502. A brand that files "shoes" in Class 25 has not automatically protected "clothing" or "hats"; those are different groups and need their own item slots.

The reverse also happens. The broad entry for "clothing" (item 250045) spans groups 2501 through 2505 at once -- five groups under a single item. That is why a careless item list can quietly over-cover one product and under-cover another, depending on which precise goods the filer actually means.

Class 9: Technology Splits the Way Engineers Think

Class 9 is the classic trap, and it is worth a close look precisely because it is the class most tech and e-commerce brands reach for first.

Group 0901 holds computers and data-processing equipment. Group 0908 holds audio and video apparatus. Both are unambiguously "Class 9," and both are entirely different markets. A company that files Class 9 for computer software has zero automatic cover for a phone app, a speaker, or a pair of headphones -- and the gap between 0901 and 0908 is exactly where a smart copycat plants a near-identical mark. The 2026 reclassification discussed below only sharpened this: items have been moving between electronic groups, which means a group map built from memory is increasingly dangerous.

Class 35: The Service Class That Looks Obvious

Class 35 is the service class foreign brands file almost reflexively, usually on the advice that "everyone needs Class 35." What many miss is that Class 35 is itself carved into separate service groups -- advertising, business management, marketing for others, and retail-related services each sit in their own group. Filing "Class 35" without specifying which services you perform is not a filing at all; it is a guess. A brand whose only activity is selling on Amazon and TikTok, for example, does not automatically need the marketing-services group just because a template said "Class 35."

The theme across all three examples is the same: the class number is a direction sign, not a description of what is protected. The protection is always, and only, the group-and-item combination actually claimed.

Cross-Reference: The Rule That Hides Mines

If the subclass system were a tidy grid, filers could learn it once and move on. It is not.

The Classification is full of cross-reference notes that declare certain goods "similar" even when they sit in different groups or entirely different classes. This is where a lot of foreign filing strategy quietly falls apart.

A frequently cited example: "tea-based beverages" sits in Class 30, group 3002, but the Classification cross-references it against soymilk in Class 29 and fruit juice in Class 32. In examination, those are treated as similar goods for confusion purposes -- even though the class numbers look unrelated on paper.

What follows is a hard lesson in search discipline. A brand runs a search in its own class, finds nothing, files, and then gets refused because a prior mark exists in a class it never searched. The cross-reference notes mean the searches a filer runs, and the classes a filer defends, cannot be read off the class number alone; they have to be read off the Classification itself.

The same logic governs enforcement and disputes. For an office action, a squatter case, or a non-use cancellation, the test of whether two goods are "similar" is driven by this document -- and supplemented by how real consumers understand the goods' function, purpose, sales channels, and users. The class number is the starting point, never the endpoint.

China vs. the US vs. the EU: Where the Model Diverges

It helps to place China's system explicitly against what foreign brands already know, because the comparison is the fastest way to see what is different.

• United States (USPTO). The USPTO uses the Nice classes but maintains its own ID Manual of acceptable goods descriptions. There is no "similar group" tier. A US filing describes specific goods, and similarity is judged on the goods as a matter of fact and consumer perception, not on a published group matrix.
• European Union (EUIPO). The EUIPO applies the Nice classes almost directly, operating from class headings and an alphabetical list. Again, no subclass or similar-group level sits between the class and the item.
• China (CNIPA). China keeps the Nice classes, then re-works them into the Classification of Similar Goods and Services, layering in similar groups, sections, and cross-reference notes.

The practical consequence is subtle but important: a framing a brand can safely ignore in the US or EU is one it cannot afford to ignore in China. The phrase "a Class 9 filing" means three different things in three jurisdictions. In China, it means the specific group-and-item combination claimed -- and nothing beyond it.

This is also why copying a US or EU specification straight into a Chinese application is a recurring error. The item names, the group structure, and the similarity links are all slightly, decisively different.

The 2026 Reclassification: Why Yesterday's Filing Can Slip

There is a live, time-sensitive reason to care about all of this right now. In January 2026, China put the 13th edition of the Nice Classification into force, and the changes were not cosmetic. Several everyday goods moved class, and a filing made under the old scheme did not automatically migrate with them.

The headline example is eyewear. Spectacles, lenses, frames, cases, and other eyewear moved out of Class 9 (group 0919) into Class 10 -- reclassified around their corrective, medical nature, with a new group 1011 created for them. Heated clothing and heated socks moved from Class 11 into Class 25. Emergency-rescue vehicles moved from Class 9 into Class 12.

What that means is brutal for a specific kind of brand: one that filed eyewear before January 2026 and assumed the matter was closed. The old registration remains valid, but its scope now sits in a class examiners no longer treat the same way, and the new scheme is not retroactive. A brand selling eyewear -- or any of the other migrated goods -- should review the old filing and, where needed, file the new class, rather than assuming the earlier registration travels with the reclassification.

The same edition also added genuinely new items to the list: AI-as-a-service in Class 42, NFC tags and "virtual mirrors" in Class 9, wearable health-monitoring devices in Class 9, and AI-driven marketing services in Class 35, among others. It is a reminder that the Classification tracks the market -- and that coverage has to track it too. The map is re-drawn whenever the goods themselves evolve.

A Worked Example: Filing a Consumer-Electronics Brand

To make the strategy concrete, consider a hypothetical brand -- call it "NovaLoop" -- selling smart fitness bands, a companion phone app, and branded apparel. The example is illustrative, not a real case, but it shows the discipline.

A naive filing might grab "Class 9" for everything electronic and stop there. An experienced Chinese filer instead opens the Classification and maps the actual offering:

• Smart bands land in the wearable-electronics group inside Class 9, not in the computer or audio groups -- so the exact sixteen-digit and group codes matter, not just "Class 9."
• The companion app sits in the software group, the same family that downloadable apps fall into. It is not automatically covered by the wearable-hardware item.
• Branded apparel belongs in Class 25, in the clothing group -- a wholly separate class that "Class 9" will not touch.

The filing becomes something like: Class 9 for the device group, Class 9 (same class, different group) for the software, and Class 25 for the apparel -- with each class's ten-item budget allocated across core and defensive slots. That is three groups across two classes, and each one has to be claimed. None of it is implied by the word "electronics."

The point of the example is not the specific classes. It is the habit: map the product line onto groups, then file groups. Anyone who files the class instead of the groups is buying coverage by category instead of by boundary.

How Experienced Filers Actually Choose Classes

Strip away the taxonomy and the strategy comes down to a few durable rules that veteran Chinese practitioners use, and that first-time foreign filers almost always learn too late.

1. Cover the group, not the class. Before filing anything, open the Similar Goods Classification and find every group the product genuinely touches. A yoga-mat brand is not filing "Class 27"; it is filing the specific group inside Class 27 that holds exercise and gym mats, and checking the cross-references to see whether a rival already sits in an adjacent group.

2. Run the search in the right groups, including the cross-references. A CNIPA search that only looks at the brand's own class number will miss the marks the examiner is required to cite against it. The search has to follow the same cross-reference logic the examiner follows, or the filer is paying for false confidence.

3. Build "core plus defense" into the ten items. The standard shape is a handful of items sold today, a couple sold next, and a couple held as defensive slots against exactly the copycat scenario that opens this article. Because the cost of a China filing is low by global standards -- roughly USD 40 a class at current rates -- a second related class or a fuller item list is usually the cheapest insurance a brand will ever buy.

4. Use multi-class filing for breadth, not depth. China allows a single application to cover multiple classes, a modernization that replaced the old one-class-at-a-time rule. Multi-class cleans up administration, but it does not change the subclass rule one bit. A two-class filing with the wrong groups is still two classes' worth of gaps.

5. Treat the subclass map as living. The 2026 reclassification is not the last one. Every new Nice edition shifts something, which is why a brand should re-check its class map whenever it adds a product, enters a platform, or renews.

The common thread is that class selection in China is a judgment call, not a checkbox. The tools are public -- the Classification, the fee schedule, the official item list -- but the judgment comes from having done it enough times to know which groups bite.

The Defensive Layer: Filing Two Steps Ahead of the Copycat

The best class strategy in China is not only about what a brand sells today. It is about what a copycat will sell against it tomorrow.

Squatters and copycats do not attack a brand's strongest category. They attack the gap next to it -- the related class the brand skipped, the adjacent group it never listed, the product line it plans to add next year. The defensive layer is the answer, and in China it is unusually cheap to build.

A defensive filing strategy usually stacks three tiers:

• The core class -- where the brand actually sells, filed with the precise groups and items mapped earlier.
• The adjacent class -- where the brand is likely to expand, or where its customers reasonably expect to find it. For a cosmetics brand, that might mean Class 3 (the products) alongside Class 35 (retail and marketing) and Class 44 (beauty services). For a tech gadget, Class 9 alongside Class 42 (software development) and Class 35 (online retail).
• The defensive class -- the class a squatter would naturally pick if the brand were the target, often Class 35 for e-commerce sellers or the class a close competitor already occupies. The logic is simple: a mark is harder to squat in a class it already occupies.

Because the base fee is roughly USD 40 per class at current rates, carrying a second or third related class is rarely the expensive line item in a China filing budget. What is expensive is the alternative -- a one-class filing that leaves the adjacent class open for someone else, discovered only when a fake store or a copycat product appears.

The defensive layer is not about filing everything. It is about filing the two or three classes that actually matter to the brand's business and its exposure, then keeping the item lists inside each one honest and defensible. Enough to close the obvious doors; not so much that the registration invites a non-use challenge.

When an Examiner Cites Your Group: The Three Ways Out

When a prior mark in the same similar group does appear, foreign filers often assume the fight is binary -- surrender or appeal. In practice there are three standard moves, and all of them operate at the group level.

Narrow the list. If the collision is on one item and the brand does not actually need that item, the cleanest resolution is to delete it and let the rest proceed. This is not a defeat; it is the partial-refusal system working exactly as designed, and it is often the cheapest and fastest exit.

Argue the goods are not similar. A citation only sticks if the goods are genuinely similar under the Classification. When the prior mark sits in a different section of the group, or the goods differ in function, sales channel, and user enough to overcome the presumption of similarity, a written argument can win. This is where the three-level structure -- group, section, item -- stops being theory and becomes the actual battleground of the filing.

Take the mark itself on. If the goods are truly similar and the brand genuinely needs them, the fight shifts to the mark: distinctiveness, coexistence, the citation's own vulnerability, or negotiating with the other side. That is a mark-level fight, but it only begins after the group-level question is settled.

The point for a foreign brand is that the similar group is not just where protection is defined; it is where refusals are decided and where they are resolved. Understanding the map is not a one-time formality. It is the language in which the next several months of a filing will be argued.

The Mistakes That Come Back Most Often

Certain errors recur because they feel safe. Here are the ones Logo China Pro keeps seeing.

Filing the class heading instead of items. A class heading is not a valid Chinese filing. Applications that attempt it get a correction notice, and the delay eats into a priority window.

Assuming one class covers the whole category. The entire premise of this article -- and the single most expensive assumption a foreign brand makes.

Skipping the cross-reference search. A clean search in Class 9 means nothing if a cross-reference ties the goods to a prior mark in Class 30 or Class 32.

Copying a US or EU specification. The item names and group structure differ, and the copy often produces a specification that is either vague or mis-scoped.

Filing ten items to "use up" the budget. Ten items with no real use plan is a roadmap to a three-year non-use cancellation, not to broader protection.

Ignoring the January 2026 reclassification. Any brand in a migrated category -- eyewear especially -- should have re-checked its coverage the moment the 13th edition took effect.

The through-line is the same in every case: the error is invisible at filing time and only surfaces later, usually in a refusal, a failed takedown, or an invalidation.

What Happens When It Goes Wrong

The cost of a subclass mistake is rarely just the filing fee.

Get the class wrong and CNIPA issues a correction notice before examination even begins -- delay, and sometimes a lost priority window. Get the group wrong and a near-identical prior mark sails through as a citation, and the application is refused on grounds that were avoidable. Under-cover the items and a rival registers the same name in the group left open, then the platform takedown fails because the mark held does not actually cover the goods being sold. Over-cover with junk items and a competitor files a three-year non-use challenge and carves the dead weight out of the registration.

None of these are edge cases. They are the recurring failure modes seen when a mark was filed abroad-style -- as one tidy class -- and dropped into a system that expects a group-by-group decision.

And because China runs a first-to-file system, the gap a brand leaves is precisely the asset a trademark squatter is watching for. Squatters do not file random classes. They file the goods the brand forgot to list, in the groups the brand never checked -- then wait for the brand to come looking.

The better habit, learned by every brand that eventually gets this right, is to treat the class map as a standing asset: built carefully once, re-checked on every product launch and every renewal, and treated with the same seriousness as the mark itself.

The Checklist Before Filing (or Re-Checking) Any Class

If a brand is about to file in China, or re-checking an existing mark, this is the short discipline that separates a real filing from a regret:

• Open the Similar Goods Classification and list every group the product touches -- not just the home market's guess.
• Verify each item against CNIPA's accepted names; do not invent descriptions or paste a class heading.
• Fill the ten items with a core-plus-defense mix, not ten random goods.
• Search all relevant groups plus every cross-reference note, not just the class number.
• Check whether the goods appear in the January 2026 reclassification list and, if so, whether a supplementary filing is needed.
• If in doubt about the map, walk through how registration works before committing the fee.
• Keep the protection honest: only file items the brand will actually use, and re-check the map whenever the product line changes.

Frequently Asked Questions

Sources

• WIPO Nice Classification
• CNIPA Similar Goods and Services Classification