A Chinese patentability search is the one most often skipped, and Chinese prior art counts everywhere.
Prior art has no nationality. A Chinese application published before your priority date can anticipate a US claim exactly as a US document would.
The volume is enormous and it is systematically under-searched — not because anyone decided to skip it, but because searchers work in English, in familiar tools, and Chinese documents require different queries.
Which makes it where undiscovered art disproportionately sits.
Why it gets missed
| Cause | Effect |
|---|---|
| Searching in English only | Chinese technical terms do not map word for word |
| Unfamiliar databases | Different interfaces and behaviour |
| Utility models overlooked entirely | An entire category missed |
| Reliance on translated abstracts | Abstracts lose claim precision |
| Habit | The search stops at familiar sources |
Utility models are the largest single blind spot. They are numerous, they publish, and they are prior art — and searchers looking for "patents" frequently do not include them.
Where to search
| Source | Coverage | Cost |
|---|---|---|
| CNIPA public search | Authoritative Chinese records | Free |
| Espacenet | Chinese documents, families, translations | Free |
| Google Patents | Full text, machine translation | Free |
| WIPO Patentscope | PCT and national collections | Free |
| Commercial databases | Better Chinese-language searching | Paid |
Use more than one. Coverage, translation quality and search behaviour differ, and a document found in one may not surface in another.
Free tools handle screening well. The gap opens between screening and a reliable conclusion.
Invention patents and utility models
| Invention patent | Utility model | |
|---|---|---|
| Substantive examination | Yes | Formality only |
| Term | 20 years from filing | 10 years from filing |
| Subject matter | Broad | Shape or structure |
| Time to grant | Longer | Much shorter |
| Volume | High | Very high |
| Prior art value | High | High, and frequently missed |
| Design patents | Separate category | — |
Not being substantively examined does not reduce prior art value. A published utility model discloses what it discloses, and disclosure is what matters under §102.
They grant fast and publish, which means they can appear as prior art earlier than a corresponding invention patent would.
Searching without the language
| Method | Why it works |
|---|---|
| IPC and CPC classification | China uses both — transfers directly |
| Citation networks | Follow references from close art |
| Drawings | Carry meaning without language |
| Applicant and assignee names | Companies are identifiable |
| Family members | An English-language counterpart may exist |
| Chemical structures | Language-independent |
Classification is the most reliable entry point. It groups documents by what they do, which is precisely what a language barrier obscures.
Drawings do more work than people expect. A mechanical disclosure is frequently assessable from figures alone, which is enough to decide whether translation is worth paying for.
Check for family members. A Chinese application with a US or EP counterpart gives you the disclosure in a language you read. See prior art.
Machine translation limits
| Use | Adequate? |
|---|---|
| Screening for relevance | Yes |
| Understanding the general disclosure | Usually |
| Reading claim scope precisely | No |
| Relying on it for a conclusion | No |
| Citing it in a proceeding | Human translation needed |
Technical Chinese loses precision in translation exactly where it matters — the qualifiers, relationships and functional language that decide claim scope.
Screen with machine translation, conclude with human translation. The cost applies only to the small number of documents that survive screening.
Building the search
| Step | Action |
|---|---|
| 1 | Identify IPC and CPC classes from close known art |
| 2 | Browse those classes in Espacenet, restricted to CN documents |
| 3 | Include utility models explicitly |
| 4 | Follow citations from the closest references |
| 5 | Check drawings on anything structurally relevant |
| 6 | Look for English-language family members |
| 7 | Human-translate the few that survive |
Step three is the one to write down. Utility models get excluded by default in some interfaces and by habit in all of them.
Step six can eliminate the translation cost entirely where a counterpart exists.
Worked example: a search that changed the answer
A US application on a mechanical coupling. English-language search found nothing close.
| Step | Action | Finding |
|---|---|---|
| 1 | English search, US and EP | Nothing within range |
| 2 | Identified IPC subclass from the closest art | — |
| 3 | Browsed that subclass, CN documents only | 180 documents |
| 4 | Included utility models | +340 documents |
| 5 | Screened on drawings | 6 structurally close |
| 6 | Machine translation | 2 potentially anticipatory |
| 7 | Human translation of 2 | 1 anticipates claim 1 |
What this cost and saved
| Steps 1–6 | Free |
| Step 7, two documents | Modest translation cost |
| Anticipatory art found | Before filing the non-provisional |
| Cost avoided | $12,000+ and two years to a §102 rejection |
The anticipating document was a utility model. It would have been missed by any search excluding that category — which is most of them.
Step four more than doubled the result set. One checkbox.
Reading the reference then produced the reformulation. Understanding exactly what it disclosed showed where the claim could be narrowed to something still commercially useful.
If you are filing in China
| Point | Detail |
|---|---|
| CNIPA will search Chinese art thoroughly | Better than you will |
| Utility models are searched too | Both categories |
| Searching before filing is cheaper | Than a rejection |
| A utility model may be a faster route | Where the invention is structural |
| Annuities are annual | Including during pendency |
Chinese annuities are payable annually, which is a different cost structure from the US three-fee schedule. See patent annuity fees by country.
Rights are national, disclosure is not
| Chinese patent | Chinese publication | |
|---|---|---|
| Restrains activity | In China only | Nowhere |
| Counts as prior art | Everywhere | Everywhere |
| Affects your US validity | Through its disclosure | Yes |
| Affects your US freedom to operate | No | No |
That distinction is the whole point. A Chinese patent cannot stop you in the US; its published disclosure can invalidate your US claim.
Manufacturing coverage is separate and strategic. China is frequently where products are made, so patent coverage there can reach a supply chain rather than a distribution channel. See freedom to operate.
Search before spending
| Ipiry Patent Survival Curve v1.0 | Rate |
|---|---|
| Survive the 3.5-year fee (2022 cohort) | 85.8% |
| Survive the 7.5-year fee (2018 cohort) | 64.6% |
| Reach full term (2014 cohort) | 41.4% |
| Abandoned before full term | 58.6% |
Computed from 27,273,654 USPTO maintenance fee records covering 8,262,336 US utility patents — see the patent survival curve.
Most patents do not survive to term. Finding anticipating art before filing costs nothing and avoids joining that statistic expensively.
When to use a professional
| Situation | Approach |
|---|---|
| Initial screening | Free tools |
| Pre-filing assessment | Free tools, then targeted translation |
| Invalidity search for a proceeding | Professional, Chinese-language capable |
| High-value filing decision | Professional |
| Litigation support | Professional, with certified translation |
Ask specifically about Chinese-language capability. A searcher working through machine translation has the same blind spots you do. See patentability search services.
Chinese patentability search: the checklist
- Treat Chinese prior art as prior art. Language and jurisdiction are irrelevant to §102.
- Search classification, not keywords, when working across languages.
- Include utility models explicitly. They are the biggest blind spot.
- Use Espacenet and Google Patents alongside CNIPA. Coverage differs.
- Screen on drawings for structural inventions.
- Look for English-language family members before paying for translation.
- Use machine translation to screen, never to conclude.
- Human-translate only what survives screening.
- Search before filing in China. CNIPA will find it otherwise.
- Distinguish rights from disclosure. A Chinese patent restrains only in China; its disclosure counts everywhere.