OCR and Data Extraction in Customs: What Technology Providers Must Know

CBP rules that OCR tools extracting entry data from shipping documents constitute customs business. Key precedents and compliance guidance for tech providers.

Chen Cui
Chen Cui9 min read

Co-Founder of GingerControl, Building scalable AI and automated workflows for trade compliance teams.

Connect with me on LinkedIn! I want to help you :)

Is Using OCR to Extract Entry Data from Shipping Documents Customs Business?

Yes. CBP has consistently ruled since 2009 that any tool or person identifying what data from shipping documents should appear on a customs entry is conducting customs business under 19 U.S.C. 1641. This applies whether the extraction is automated (OCR, AI) or manual, and regardless of whether a licensed broker reviews the output before filing with CBP.

Does It Matter If a Licensed Broker Reviews the OCR Output Before Filing?

No. The preparation of entry documents is customs business "whether or not signed or filed by the preparer." The act of identifying and extracting entry-relevant data is itself preparation, not merely transmission. CBP made this clear in HQ H068278 (2009), HQ H326926 (2023), and HQ H350722 (2026).


Document processing technology is transforming customs operations. OCR tools can scan a commercial invoice, bill of lading, or packing list and extract shipper names, consignee details, product descriptions, quantities, values, and country of origin in seconds. This data can then be used to prepare entry documents. The efficiency gains are obvious. But CBP's regulatory framework draws a hard line: deciding what data should appear on an entry is customs business, regardless of whether a human or machine makes that decision.

Last updated: March 2026

GingerControl is a trade compliance AI platform that helps importers, exporters, and customs brokers classify products, simulate tariff costs, and track policy changes.

The Precedent Chain: Three Rulings That Define the Rule

HQ H068278 (2009): The Original OCR Ruling

In HQ H068278 (September 28, 2009), a service bureau in Buffalo provided software to Canadian customs brokers and freight forwarders. In one scenario, an unlicensed Canadian contractor would use OCR technology to scan shipping documents and extract data necessary for entry. The bureau would maintain a web-based database to fill in missing fields. A licensed U.S. customs broker would ultimately transmit the entry to CBP.

CBP found this constituted customs business. The contractor and bureau were "determining valuation, classification and other entry-related information," which goes beyond "mere electronic transmission of data." The key principle: an unlicensed entity making decisions about what constitutes relevant entry information, and then having software automatically apply that decision matrix, is conducting customs business through its tool.

CBP also found that the bureau was conducting customs business "through its contractor" by directing the contractor to create the decision matrix for what data would appear on entries.

HQ H326926 (2023): Manual Extraction Is Also Customs Business

In HQ H326926 (December 19, 2023), the broker Heizwerthy sought to hire an offshore, unlicensed data entry company to review shipment documents (bills of lading, commercial invoices) and manually key the relevant data into Heizwerthy's ABI software. Heizwerthy would then audit the data and transmit the entry to CBP.

CBP held that "identifying entry-related data and keying it into an ABI system falls squarely within the scope of preparing parts of an entry intended to be filed with CBP." This was customs business in violation of 19 U.S.C. 1641(b)(1) and 19 C.F.R. 111.37.

CBP also raised three additional concerns: the unlicensed company was not eligible to access ABI under 19 C.F.R. 143.1(a); the company had not been approved by CBP to access ABI; and permitting third-party access to Heizwerthy's account would breach client confidentiality under 19 C.F.R. 111.24.

HQ H350722 (2026): OCR on an Online Platform

In HQ H350722 (January 16, 2026), CBP examined an online platform that offered an OCR tool to importers. The tool scanned uploaded shipping documents and identified information for entry, pre-filling data into entry documents. This tool was available regardless of whether the importer engaged a third-party broker through the platform.

CBP found this constituted customs business. An OCR tool that "identifies precise pieces of information for a shipment intended to be entered thus entails customs business because it prepares parts of the data appearing on an entry in an electronic format." CBP explicitly cited H326926 and H068278 as supporting precedents.

CBP also noted that "the filing of an entry is not a prerequisite" to the activity constituting customs business. The statute covers preparation of documents "intended to be filed with CBP," and an OCR tool designed specifically to cull entry data from shipping documents meets that standard even if the importer has not yet engaged a broker to file.

What This Means for Technology Providers

The rule is clear: any technology that identifies which pieces of information from a shipping document belong on a customs entry is conducting customs business. This applies to:

Technology Customs Business? Rationale
OCR scanning commercial invoices to extract entry data Yes Identifies entry-relevant data (H068278, H350722)
AI parsing bills of lading to populate entry fields Yes Prepares parts of entry documents (H326926)
Manual offshore data entry from shipping docs to ABI Yes Identifies and keys entry data (H326926)
Software with decision matrix determining entry fields Yes Makes decisions about entry content (H068278)
Generic document scanning with no entry-specific logic Depends If not identifying entry-specific data, may be "mere transmission"
Importer scanning own documents for own entries No Corporate compliance activity, acting on own behalf

The "mere electronic transmission" exception is narrow. The statutory exclusion covers only the transmission of data "received for transmission to Customs" without any decision-making about the data's content. Once a tool identifies which data elements are relevant for entry, applies a decision matrix, or pre-fills entry fields, it crosses the line.

Broker review does not cure the issue. CBP has rejected the argument that having a broker review OCR output before filing makes the extraction permissible. The preparation itself is customs business under 19 U.S.C. 1641(a)(2), which explicitly covers preparation "whether or not signed or filed by the preparer."

The territorial requirement applies. Under 19 C.F.R. 111.3(a), all customs business must be conducted within U.S. customs territory. Technology providers operating OCR tools from outside the United States face both the customs business issue and the territorial requirement.

How to Structure Document Processing Compliantly

Technology providers seeking to offer document processing for customs purposes have several options:

Option 1: Operate under a broker's license. If the technology provider holds a customs broker's license, its document processing tools can constitute customs business, provided the broker exercises responsible supervision and control under 19 C.F.R. 111.28 and the work is conducted within U.S. territory.

Option 2: Provide the tool to licensed brokers as a "corporate compliance activity." If a broker uses OCR technology internally as part of its own entry preparation process, the broker is conducting customs business under its own license. The technology provider is supplying a tool, not conducting customs business, provided the provider does not make decisions about entry content. The distinction in H068278 was that the bureau directed the contractor to create the decision matrix; if the broker creates and controls the decision logic, the provider may be supplying technology rather than conducting customs business.

Option 3: Limit the tool to non-entry purposes. Document processing for purposes other than entry preparation (trade analytics, supply chain visibility, compliance monitoring, cost estimation) does not constitute customs business. The key is ensuring the tool is not designed to identify and extract data "intended to be filed with CBP in furtherance of" entry activities.

For companies that want to build compliant document processing workflows, GingerControl offers AI Agentic System Build services, helping teams structure their technology and processes to align with CBP's regulatory framework.

Frequently Asked Questions

Can an importer use OCR to process its own shipping documents?

Yes. The customs business restriction applies to acting "on behalf of another." An importer processing its own documents for its own entries is acting on its own behalf. Corporate compliance activities are also excluded from the customs business definition by regulation.

What about OCR tools that extract data for ISF filing?

ISF filing is not customs business (HQ H045695, 2010), so extracting data for ISF purposes (limited to six-digit classification) is permissible. However, the same data cannot be used for entry purposes unless processed by the importer or a licensed broker.

Does this ruling affect customs brokers using OCR internally?

Licensed customs brokers may use OCR tools as part of their entry preparation process under their broker's license. The rulings address unlicensed entities using OCR technology, not licensed brokers. However, a broker must exercise responsible supervision and control over any automated tools used in its customs business.

Can a technology company provide generic document scanning without triggering customs business?

Potentially. If the tool provides general document scanning (text recognition, data extraction for generic purposes) without entry-specific decision logic, it may fall under "mere electronic transmission." The critical question is whether the tool identifies which data elements are relevant for customs entry. A generic OCR tool that extracts all text without categorizing it for entry purposes is different from one designed to identify consignee, value, classification, and country of origin for entry filing.

How does GingerControl's approach differ?

GingerControl is a pre-classification research tool, not an entry preparation tool. It does not scan shipping documents to extract entry data or pre-fill entry forms. The Classifier accepts product information and surfaces candidate HTS codes with supporting analysis for broker review. This research function is distinct from the document-to-entry data extraction that CBP has consistently found to be customs business.


Building compliant customs technology workflows requires understanding where data extraction ends and customs business begins. GingerControl offers AI Agentic System Build services to help companies structure their automation compliantly. Talk to our team


References

[REF 1] HQ H068278 (Sept. 28, 2009): OCR extraction as customs business Source: CROSS

[REF 2] HQ H326926 (Dec. 19, 2023): Manual data entry as customs business Source: CROSS

[REF 3] HQ H350722 (Jan. 16, 2026): Platform OCR tool as customs business Source: CROSS

[REF 4] 19 U.S.C. 1641: Customs brokers statute Source: LII

[REF 5] 19 C.F.R. Part 111: Customs broker regulations Source: eCFR

[REF 6] HQ H045695 (Oct. 15, 2010): ISF filing not customs business Source: CROSS

[REF 7] Delgado v. United States, 581 F. Supp. 2d 1326 (Ct. Int'l Trade 2008)

Chen Cui

Written by

Chen Cui

Co-Founder of GingerControl

Building scalable AI and automated workflows for trade compliance teams.

LinkedIn Profile

You may also like these

Related Post