Patent Ownership Due Diligence: USPTO Assignment Worksheet

Executive Summary

Patent ownership due diligence at the screening stage asks one narrow question: does the public USPTO (United States Patent and Trademark Office) record connect a specified patent or application number to the legal entity a target company says owns it, on a stated date? It does not ask whether the patent is valid, enforceable, free to practice or valuable. This report treats the answer as a chain-of-record worksheet keyed by number rather than company name, ending in one of four screen outcomes: supported recorded chain, mismatch, incomplete, or counsel review.

The USPTO search database holds recorded patent assignment information from 1980 to the present ^[1]. The USPTO's own rule text, reproduced in MPEP (Manual of Patent Examining Procedure) section 301, puts the National Archives boundary at May 1, 1957 ^[2], so the boundary should be confirmed rather than assumed. The tool is also new: the current Assignment Search went live on October 20, 2025 ^[3].

Three limits drive every conclusion. First, recording is ministerial: MPEP 301 calls it merely a ministerial act ^[4], and 37 CFR 3.54 says it is not a determination of validity ^{[5] [6]}. Second, recording is voluntary, so a missing record is not a negative ^{[7] [8]}. Third, document labels mean different things: a name change is a link in the chain of title but not a change in legal entity, a security interest leaves ownership unchanged unless and until a lender forecloses, and a license is not an assignment.

The quantitative context is modest. A 2018 study reports that 38 percent of US patenting firms had pledged patents as collateral by 2013 ^[9]. In the 1970 to 2014 dataset, recording lagged execution by roughly 182 days on average with a 21 day median ^[10]. The worksheet therefore computes elapsed days from execution to recordation as a freshness flag only, never as a statement about legal priority.

An analyst can complete the screen in a few steps: fix the claim unit, search the record by number and by party, preserve each reel and frame, build the chronological chain, reconcile names against registry records, run lien searches, and assign an outcome. Anything that turns on whether a transaction was effective belongs with intellectual property counsel. This report does not give a title opinion or legal advice.

Introduction and Background

Private equity origination teams routinely meet companies that describe "our patent portfolio" in a deck or on a website. A screening analyst must decide before spending more review time: is the public record strong enough to link specified patent or application numbers to the matched legal entity, or should the claim stay open for IP counsel? The World Intellectual Property Organization (WIPO) describes IP due diligence as a comprehensive audit of a company's intangible asset portfolio (Source: www.wipo.int) that assesses the financial, commercial and legal benefits and risks of the portfolio (Source: www.wipo.int) and seeks to determine whether the relevant IP is free of encumbrances (Source: www.wipo.int), including whether any security interests or liens have been granted (Source: www.wipo.int). The full audit is counsel's work. The screening step here is a small, repeatable slice that can be done from public records.

Docket, a deal-origination and evidence-backed acquisition target research platform and service for private equity firms, publishes this report as a research reference. It is not a law firm, and nothing here is legal advice or a title opinion. The Docket research library carries related guides on evidence standards for acquisition research.

The legal backdrop is short. Under 35 U.S.C. 261, patents have the attributes of personal property ^[11], and applications, patents and interests in them are assignable by an instrument in writing ^[11]. The same section directs the USPTO to maintain a register of interests in patents and applications ^[11]. The USPTO's own description is that an owner may transfer ownership through an assignment or retain ownership but change its name ^[1]. The two events look similar in a search result and mean very different things for a screen.

This report proceeds as follows. The first section fixes the claim unit and separates the roles around a patent. The second walks through the current Assignment Search. The third maps recorded document types to permissible inferences (Table 1). The fourth presents the worksheet with a fictional example and elapsed-day calculations (Table 2). The fifth covers name reconciliation and security interests. The sixth sets out coverage limits, followed by the evidence base, the screen outcomes (Table 3), common questions and implications.

Defining the Claim Unit and the Roles Around a Patent

The claim unit: number, entity, date

A reliable screen starts with a claim stated as three elements: a specific patent or application number, a specific legal entity (not a brand), and an as-of date. "The company owns its patents" is not testable. "Harbor Ridge Robotics, LLC owns US 99,000,001 B2 as of 2026-10-20" is. USPTO rules identify recorded documents by number: an assignment relating to a patent must identify it by patent number ^[5], while an assignment signed before an application is filed identifies the application by inventor names and title ^[5]. A search by company name alone can miss documents indexed under another name.

The worksheet should record which number form the claim uses (patent number, application number or publication number), because a published application and its granted patent carry different numbers. A continuation, division or continuation-in-part is a separate application, so each family member gets its own row rather than inheriting the parent's chain.

Inventor, applicant, assignee, owner, licensee, operating brand

Six roles recur in diligence materials, and the record treats them differently.

• Inventor: the individual who conceived the invention. MPEP 301 states that, for applications filed before September 16, 2012, ownership of a patent or application initially vests in the named inventors ^[4], and an inventor can assign the interest held, so an inventor is not necessarily the current owner.
• Applicant: the party that filed. For applications filed on or after September 16, 2012, the original applicant is presumed to be the owner absent an assignment ^[5]. Section 118 lets an assignee or obligated assignee file ^[12], and then the patent is granted to the real party in interest ^[12].
• Assignee: the party named as receiving an interest in a recorded document, which the cover sheet must identify by name and address ^[5]. A recorded assignee is a claim made in a document, not a verified owner.
• Owner: the party that holds title now. A partial assignee is any assignee with less than the entire right, title and interest ^[13], and each joint owner may exploit the invention without the others' consent absent agreement ^[14].
• Licensee: a user of rights. A license is not an assignment of the patent ^[4], so a licensee never appears in the ownership chain.
• Operating brand: the trade name a customer sees. It has no standing in the assignment record unless it is also a recorded legal name.

Front-page data need particular care. MPEP 307 says the assignment data printed on a patent are based solely on the information the filer supplied ^[15] and that this information is not updated after issuance ^[15]. A blank or old front-page assignee is a prompt to search the assignment record, not an answer.

Ancillary filings are weaker. Naming an assignee in an application data sheet does not substitute for recording the assignment ^[16], and a maintenance fee payer's name proves nothing about title.

Searching the USPTO Assignment Record by Number and Party

What the current tool is

The USPTO page for assignment searching describes a database of all recorded Patent Assignment information from 1980 to the present ^[1]. The agency announced that the current Assignment Search went live on October 20, 2025, replacing two legacy search applications ^[3]. A later system-status notice, marked completed, said that effective July 24, 2026 the Patent Export Assignment Data Search moved from Assignment Center to the USPTO Open Data Portal for bulk data searches, and that Assignment Search was enhanced to provide updated assignment information ^[17]. Saved bookmarks or scripts for older interfaces should be re-tested, and every worksheet should carry a "last checked" date.

Screen layouts and field labels in the live tool can change, so the steps below describe a general order of work rather than exact screens. A reviewer should open the live Assignment Search tool and confirm current field labels before relying on them.

Search by number first, name second

A practical order for a screen:

• Start with the number. Enter the patent or application number from the claim and record every result.
• Search the claimed owner as assignee. Use the claimed name and obvious variants to find properties the number search did not link.
• Search earlier names as assignor. Name-change and merger records often sit under the prior name.
• Open the recorded document image. A result is an index entry derived from a cover sheet; the document shows signatories, granting language and any property schedule.
• Preserve identifiers. Save the reel and frame numbers so each recorded document can be located again.

The index fields come from what filers type. The cover sheet must describe the interest conveyed or the transaction ^[18], so a conveyance label is the filer's description. The Chief Economist's working paper reports that recorded data come with minimal to no verification, validation, or standardization ^[10]. A 2024 preprint adds that multinational companies file patents under many names, including alternate spellings of the same entity ^[19].

Why a record may not appear

A missing record is not a clean negative. Government-interest instruments may be held on a secret register with no information given ^[5]. Assignment records for applications not open to the public are obtainable only on written authority, though a bona fide prospective or actual purchaser, mortgagee, or licensee of such an application can be among those who may obtain them ^[2]. The USPTO states that recorded assignment documents are available for public inspection, with limited exceptions ^[20], and the assignment-records rule makes patent assignment records open to public inspection ^[2]. A very recent transaction may also not be visible yet. Each is a reason to log the search date and to read "no result" as "no result found on that date".

Reading Document Types: What Each Record Does and Does Not Support

Not every recorded document changes ownership. The USPTO describes its dataset as including non-ownership items such as licensing agreements, security interests, mortgages, and liens ^[21]. Table 1 converts the document types an analyst is likely to meet into a permissible inference. The columns are conservative: they state what the record supports for screening, not a legal conclusion about effect.

Table 1 shows that only two of its nine rows can support a "transfer" step, and both carry caveats. The others describe identity, encumbrance, correction or absence. The most common mislabel in screening notes is treating a merger or name-change row as proof of a transfer to the claimed owner. The record supports a link, and the analyst still has to show that the entity at the end of the link is the entity in the claim. Another shortcut is reading a security agreement as proof that the grantor owns the patent; it is a pointer, not proof of title.

Building the Chain-of-Record Worksheet

Fields

The worksheet has one row per recorded link, grouped under a header that fixes the claim, with one block per patent or application number. The fields:

• Publication, application and patent number: the number the claim relies on and any linked numbers.
• Claimed owner: the exact legal entity name asserted, as of the stated date.
• Assignor and assignee: as typed on the record, preserved verbatim.
• Conveyance label: the filer's description plus the analyst's classification from Table 1.
• Execution date and recordation date: both, because they answer different questions. The recording date is the date a document meeting the requirements is filed in the Office, not the date it was signed ^[5].
• Reel and frame or document identifier: the key to retrieve the image.
• Days from execution to recordation: a computed freshness flag.
• Entity-name crosswalk: old name to new name, parent to subsidiary, registry identifiers.
• Missing link: the step between the last recorded assignee and the claimed owner.
• Counsel question: written so it can be answered from the document.
• Last checked: the search date.

A fictional worked example (Hypothetical Example)

The example is fictional. The patent number, company names, reel and frame numbers and dates are invented and match no real record. The claim is that "Harbor Ridge Robotics, LLC owns US 99,000,001 B2 as of 2026-10-20", with every search dated 2026-10-20. Table 2 lays out the chain; elapsed days run from execution to recordation.

The outcome for this example is incomplete. Rows 1 through 3 form a recorded path from the inventor to the Holdings entity, and row 2 shows why a name change must not be read as a transfer. Row 4 is the gap: the claimed owner appears nowhere as an assignee, and the screen cannot close it by assumption. Row 5 adds an encumbrance flag. The elapsed days are plain date arithmetic: 16 days for row 1, 227 for row 2, 117 for row 3 and 28 for row 5. Rows 2 and 3 exceed the three-month window that 35 U.S.C. 261 uses ^[11], so they are flagged "recorded more than 90 days after execution". The flag prompts a question and nothing more.

Why elapsed days are a freshness flag only

The statute provides that an unrecorded assignment, grant or conveyance is void against a later purchaser or mortgagee for valuable consideration, without notice, unless recorded within three months from its date or before the later purchase ^[11]. That is a priority rule between competing parties, and the worksheet does not apply it. A long gap tells the analyst where to look harder, because other documents may have been recorded in the interval. Row 2 at 227 days is therefore a prompt for the registry check on the name change, not a finding.

Reconciling Names, Parents and Security Interests

Entity-name crosswalk

An assignee name in a USPTO record is a string. Matching it to a legal entity is a separate step, and each public source has limits.

• State registries: Delaware tests a corporate name for distinguishability only against Division of Corporations records ^[24]. Texas applies a test within its secretary of state records (Source: www.sos.state.tx.us), and Colorado against names on record in that state (Source: www.sos.state.co.us). Identically named entities can therefore exist in different states, and a bare name match is weak evidence.
• Legal Entity Identifier data: the Global Legal Entity Identifier Foundation (GLEIF) says its LEI pool can be searched free of charge and without the need to register ^[28]. Level 2 data reports parents as self-reported by legal entities that have or acquire an LEI ^[29], and a parent's LEI appears on a child's record only when the parent has one ^[29]. Collection started in May 2017 ^[29], so coverage is uneven.
• Subsidiary lists: Regulation S-K requires registrants to list subsidiaries and their jurisdictions ^[30], but subsidiaries that are insignificant in the aggregate may be omitted ^[30], so absence from a list does not exclude group membership.

For each assignee string the crosswalk should record the registry checked, the identifier found and the date. The discipline to retain: each name-change or merger row needs registry corroboration, and each similar name needs an explicit disposition (same entity, related entity or unrelated) before it enters the chain.

Security interests and the lien search

For a screen, a recorded security agreement is an encumbrance flag, and WIPO's diligence material tells reviewers to determine whether any security interests or liens have been granted in the IP assets (Source: www.wipo.int), and lender diligence requires searching the chain of title back to the original creator (Source: www.wipo.int). A careful lien screen therefore runs two searches. Under Article 9 of the Uniform Commercial Code as enacted in Delaware, a financing statement must be filed to perfect all security interests, subject to listed exceptions ^[31]. The law of the jurisdiction where the debtor is located governs perfection and priority ^[31], and a registered organization organized under a state's law is located in that state ^[31]. A financing statement for most collateral is filed with the Secretary of State ^[32], and Delaware accepts UCC filings electronically ^[33]. A filed financing statement is effective for five years after the date of filing ^[34] and lapses unless a continuation statement is filed before the lapse ^[34], so an old filing may no longer be effective while a USPTO security agreement record remains. Delaware defines a "general intangible" as any personal property other than listed categories ^[35]; whether a patent falls there is a legal question this report does not answer.

The practical rule: search the USPTO record for security agreements and releases by number, then search the debtor's state of organization for UCC filings by exact legal name. The sources can disagree, and neither alone resolves whether a lien is outstanding. A recorded conditional assignment is regarded as an absolute assignment for Office purposes, and the Office does not decide whether conditions were met ^[5]; a security agreement is a different document type. Flag either form for counsel.

Coverage Boundaries: Ministerial Recordation, 1980 and What Silence Means

Recordation is ministerial

MPEP 301 states that recording an assignment is merely a ministerial act ^[4]. The USPTO's 2024 Federal Register notice describes recording as an administrative action and not a determination of validity ^[6], while noting that the public uses the records for ownership and chain-of-title searches ^[6]. Both are true at once: the record is the standard place to look, and it is not a verification. The regulation, MPEP and Federal Register text carry the point.

The 1980 boundary and where earlier records sit

The USPTO page says the search covers recorded information from 1980 and that earlier assignments are maintained at the National Archives and Records Administration, whose Record Group 241 holds Patent Office records ^[36]. The regulation on assignment records, also reproduced in MPEP 301, says records recorded before May 1, 1957 are maintained there ^[2]. The two statements give different dates and may describe different things: the coverage of the online database versus where older records are held. The safe practice is to treat 1957 to 1980 as an open question for older chains and to ask the Assignment Recordation Branch or counsel where the earliest links sit.

Recording is voluntary and incomplete by design

An NBER working paper and an OECD report both state that recording patent transfers is not mandatory ^{[7] [8]}. The rule is not unique to the United States: a PATSTAT Register paper notes that Europe has no centrally managed register for patent transfers ^[37], so jurisdiction must be fixed in the claim. A 2012 public comment to the USPTO tested a 100-patent sample and found the assignment to the front-page non-inventor assignee unrecorded for thirty of them, while cautioning that the sample may overstate the rate ^[38]. That is a single small sample, not a population estimate, but it shows an unrecorded link is not exotic. Corrections add another layer: a corrected cover sheet is filed for recordation, so two entries can exist for one event. Silence supports only "no document was found on this date". It supports neither "the company owns the patent" nor "the company does not".

Data Analysis and Evidence

The public numbers below frame how much weight the record can carry. They come from federal notices, regulations and academic studies, and most predate the 2025 tool change, so they describe the record's character rather than today's counts.

• Volume. The USPTO's June 2024 notice estimated 724,442 annual responses for the recordation collection ^[6]; the September 2024 notice raised estimated burden to 485,376 hours after comment ^[20]. Estimates combine patents and trademarks.
• Dataset size. The academic description of the USPTO Patent Assignment Dataset reports a relational database of roughly 6 million assignments, licenses, securitizations, and other conveyances ^[39], which is the scale of the record an analyst is sampling.
• Fees. Under 37 CFR 1.21(h), electronic recordation carries a fee of $0.00 ^[40], while a paper submission is $54.00 per property ^[40].
• Timing. Recording lagged execution by roughly 182 days on average with a 21 day median in the 1970 to 2014 data ^[10]. The gap between mean and median indicates a long tail of late recordings.
• Pledged patents. A Journal of Financial Economics study reports that 38 percent of US patenting firms had previously pledged patents as collateral by 2013 ^[9], so encumbrance checks are not an edge case.
• Other jurisdictions. A Scientometrics study of European data found that more than 30 percent of EPO patents change ownership ^[41], a reminder that ownership change is common outside the US as well.

Read together, the numbers support three working assumptions. The record is large and filer-typed, so name matching needs a registry check. Recording is optional and often late, so an absent or delayed entry is common enough to plan for. Pledges of patents as collateral are widespread among patenting firms, so a lien check belongs in every screen. Every figure is an official or published-study number from a defined period. None measures how often a diligence screen reaches a wrong conclusion, and no public benchmark of that error rate was found. The limits are directional: assignment data are voluntary, filer-typed and unstandardized, and large enough that matching on names alone will produce false links.

Screen Outcomes and Implementation Guidance

Four outcomes

A screen should end in one of four labelled outcomes, each describing the record, not the title. Table 3 defines them with triggers and next actions.

Apply Table 3 with a stated precedence. The primary label is the highest-ranked row whose trigger is met among Mismatch, Incomplete and Supported, in that order. A counsel-review trigger is carried as a secondary flag on that label, and it becomes the primary label only when no other row's trigger is met. A chain with an unreleased security agreement therefore cannot be labelled supported. In the fictional example the primary label is incomplete, because the missing link in row 4 is unresolved, and a counsel-review flag is attached for the unreleased security agreement. The value of the table is reproducibility: two analysts using the same worksheet should reach the same label, which serves a screening archive better than an unsupported yes or no.

A checklist for private equity screening

• Fix the claim. Record the number, the entity and the date; one row per family member.
• Pull the front page. Note the printed assignee and date, remembering it is not updated after issuance.
• Search by number, then names. Log result counts and the search date.
• Open every image. Capture signatories, granting language and property schedule.
• Compute days. Flag long lags; draw no priority conclusions.
• Run the crosswalk. Registry, LEI and subsidiary corroboration for every name change and merger.
• Run the lien searches. USPTO security agreements and releases, then the debtor's state UCC office.
• Assign an outcome. Use Table 3 and write counsel questions in plain language.
• Keep the scope. The screen is separate from validity, enforceability, freedom to operate, licensing terms or value.

Implications and Future Directions

The record is changing underneath the workflow. The 2025 replacement of the legacy search applications and the July 2026 move of the Patent Export Assignment Data Search from Assignment Center to the Open Data Portal mean automated pulls built for earlier interfaces may break or return different fields. Teams that depend on bulk data should re-baseline scripts against current outputs and keep the dated worksheet as the system of record.

Identity resolution, not search, is the bottleneck. The preprint on name harmonization worked with a large set of assignee names and concluded that variants distort company-level counts ^[19]. Any automated "company owns N patents" figure should be treated as a lead list, and a non-US patent needs its own jurisdiction check because Europe has no centrally managed transfer register ^[37]. Each number should be re-checked through the worksheet before it enters a memo.

For research reviewers the main discipline is wording. "The USPTO record shows an assignment to X" is a statement about a record; "X owns the patent" is a legal conclusion. Making the difference visible in a table keeps a memo honest and lets counsel pick up where the screen stops. Uncertainty remains in three places: unrecorded and confidential arrangements will never appear in the public record, the 1957 versus 1980 archive boundary needs direct confirmation, and the live tool may keep changing. Each is a reason to keep screens dated and to prefer a clear "incomplete" over a guess.

Frequently Asked Questions (FAQs)

How can an analyst check who owns a patent?

Search the USPTO Assignment Search by patent or application number, open each recorded document and build the chain from the first recorded owner forward. Cross-check the front-page assignee, which reflects what the filer supplied. MPEP 301 says recordation of the assignment provides legal notice to the public of the assignment ^[4]. The result is a statement about the record; ownership is a legal question.

What is a USPTO patent assignment search?

It is the USPTO's public search of recorded patent assignment information from 1980 to the present ^[1].

How can a patent assignment be verified in M&A diligence?

Treat the recorded assignment as an index entry, then read the document. Confirm the property schedule names the patent, the assignor matches the prior link, and the assignee reconciles to the entity in the deal. WIPO's guidance says a venture in diligence must provide evidence that it owns the IP and the chain of ownership (Source: www.wipo.int), and lender diligence requires searching the chain of title back to the original creator (Source: www.wipo.int). Anything beyond the content of documents is a counsel question.

What are recordation gaps, and does the company own its patents?

A gap is a missing step between the last recorded assignee and the claimed owner. It can mean an unrecorded document, a different owner or a restricted record. The public record can support, contradict or fail to settle a specific claim, but it cannot confirm title, because recording is not a determination of the validity of the document or its effect on title ^[5]. Use the four outcomes in Table 3.

How can a patent lien or security interest be searched?

Search the USPTO record for security agreements and releases by number, then search the debtor's state UCC records. Perfection and priority questions belong with counsel.

Conclusion

Patent ownership due diligence at the screening stage is a record-reading exercise with a strict vocabulary. The claim unit is a number, a legal entity and a date. The USPTO's Assignment Search is the standard public source, covering recorded information from 1980, with an earlier archive boundary that should be confirmed. Recording is ministerial and voluntary, and the label on a recorded document says what the filer called it, not what it achieved.

The worksheet turns that vocabulary into a repeatable process: one row per recorded link, both dates, the reel and frame, a name crosswalk, a missing-link field and a counsel question. Elapsed days help decide where to look harder but say nothing about priority. The four outcomes describe the record. They never replace a title opinion, and they never speak to validity, enforceability, freedom to operate, licensing or value.

A team that adopts the worksheet will not always reach a clean answer, and that is the point. A dated, documented "incomplete" with a precise counsel question is more useful to a deal team than an unsupported assertion that the company owns its patents. Anything that turns on legal effect, including validity, enforceability, freedom to operate, licensing terms and value, stays with IP counsel.