China supplier names can change. The code stays with the entity.

The English name is an alias until its relationship is recorded

Copy the English name exactly. Save the URL, marketplace listing, quotation or contact channel where you found it. The name may be a brand or sales label. A similar-looking Chinese name does not establish which entity is behind it.

Ask the supplier to state the exact Chinese legal entity and Unified Social Credit Code that relate to the proposed transaction. Keep that statement, the channel and the date as a separate evidence line.

Use the Chinese legal entity and USCC as the official-record anchor

Search the National Enterprise Credit Information Publicity System by the exact Chinese name or Unified Social Credit Code.

SAMR’s Unified Social Credit Code measures have been in force since 1 February 2026. Order No. 112 defines the code as a national, unique and life-long identifier. It must appear on the business licence under a “one licence, one code” rule.

Save the displayed legal name, code, status, source URL and query date. An undated screenshot leaves you without the time of the check.

A supplier’s licence image gives you a code to look up; it does not replace a current record. Follow the company-registration search steps →

In the 8 August 2026 study of 264 selected company names, 53.4% had at least one former legal name on record and 22.7% had two or more. The sample-list source remains unconfirmed. These figures do not describe all manufacturers or tell you whether your supplier’s invoice uses an old name. Check the code and current name together.

Read the code’s five parts

The State Council scheme defines 18 characters in five parts.

• 1: registration management department.
• 2: institution category.
• 3–8: the registering authority’s administrative division.
• 9–17: subject identifier.
• 18: check character.

The same scheme lists the category values used by the market-regulation authorities: 1 for an enterprise, 2 for a sole trader and 3 for a farmers’ specialised cooperative.

Characters 3–8 identify the authority that registered the entity, not where it operates. Check a claimed factory address separately →

The final character is a check character, included in the scheme to detect errors. A passing result shows that the 18-character string is internally consistent with the coding rule. It does not prove error-free transcription, that the organization exists, that it remains registered or that it has any relationship to the person quoting it.

Forging, fraudulently using, transferring or renting out a code can attract a fine of up to RMB 30,000, or RMB 30,000 to 100,000 in serious cases. Check a quoted code against the record.

USCI and USCC refer to the same 18-character code. The mandatory coding standard is GB 32100-2015. Its official English title is The coding rule of the unified social credit identifier for legal entities and other organizations.

A freight forwarder, bank or customs broker may ask for a USCI number or USCI code. They mean that same code. There is no separate number to look up.

The code follows the entity; the names do not

The State Council scheme keeps the code with the entity throughout its life. Moving, changing the legal representative or changing the business scope leaves the code unchanged. Order No. 112 also preserves it in two conversions: a sole trader becomes an enterprise, or a sole proprietorship becomes a company or partnership.

A new legal entity receives a new code. The old code is never reissued; the measures require code data to remain permanently available for traceable queries.

Use the code to compare records across dates. A changed sales name does not tell you whether the legal entity changed. Ask for the current Chinese name and code, record when you received them, then check the record instead of assuming continuity.

Ask for an explanation of a name difference

A difference is an unresolved evidence question, not proof of fraud. Re-query the code, preserve the original document and ask for a written relationship explanation through an independently verified contact channel. Do not force a close-looking English label, Chinese name or beneficiary into a match because the commercial story sounds plausible.

What an English-name search returns

Searching 106 selected English names from the NHTSA vehicle-manufacturer list through one licensed data source produced 84 candidate records in August 2026.

A public-data check on the same question, run on 6 September 2026. The table above uses a licensed data source. This check uses a public roster. Every registrant in the US vehicle regulator's own roster carries a self-reported English name, and 2,633 of them give China as their country. After normalising case, punctuation and the usual company suffixes, 8 of the 2,633 (0.3%) share a name with another registrant, in four pairs. Exact English-name duplicates inside a curated roster are rare.

One roster cannot say whether that rate describes Chinese company names or the way a single regulator collects them. The second aggregation uses the same normalisation on a US device register. Its filename is dated 7 September 2026, but its generation timestamp is 6 September; the query date remains unconfirmed. The saved file names the openFDA device registration listing as its source. We read that file on 10 September 2026. It reports 5,012 distinct registration numbers and 4,993 normalised names. The file reports a difference of 19 (0.4%) as a collision lower bound. But the names and numbers were not paired. These totals alone do not establish which registrations share a name or a company-level collision rate.

The FDA aggregate needs that limit. The public aggregation returns names and registration numbers separately, never paired, so it does not independently establish a one-to-one name-to-registration map. This route reached 5,012 registration numbers. The bulk-export census uses a separate collection path and version; matching a past total is not independent identity validation. The full file is here, with the method and its limits.

Read that result the right way. It does not say English names are safe to rely on. For your order, check the link between the supplied English name and the Chinese legal entity. The table above shows provider match labels; it does not measure the size of that gap or the rate of wrong matches. The vPIC collision count depends on the stated normalisation. A one-letter difference counts as two names here. The roster also holds self-reported English names, which are not a substitute for the Chinese legal name. Data: the full name-collision file.

The public file dates the search 18 August 2026; the original run and retest use 19 August. The study explains this unresolved date difference.

These match labels do not reveal the provider’s algorithm or explain why 45 inputs returned no candidate. A missing result does not prove a company is absent from the register.

Treat the result as a lead to check. An English name on file may be an old name or brand. Compare the Chinese legal name and code with your deal documents.

Method and limitations

The official sources below cover search inputs, code structure, stability and misuse. Each carries its check date.

This guide does not assess your supplier or validate a code. A structurally well-formed code is not a verified organization. For your order, use current records and seek legal advice where needed.