USPTO Trademark Data: The Federal Brand Registry Behind 3 Million Active Marks and the TESS Search System

The United States Patent and Trademark Office maintains the definitive federal register of American brand identity — roughly three million active registered trademarks covering everything from household consumer goods to financial services to software platforms. Under the Lanham Act of 1946, registration confers nationwide constructive notice, federal court jurisdiction, and the right to record the mark with US Customs to block infringing imports at the border. Unlike patents, which expire after 20 years, a trademark can survive indefinitely as long as the owner continues using it in commerce and renews on schedule. The result is a federal dataset that functions simultaneously as a business formation indicator, a brand valuation index, and a map of commercial competition across every industry sector.

What a Trademark Is — and Is Not

A trademark is a word, phrase, symbol, design, sound, color, or combination thereof that identifies the source of goods or services and distinguishes them from those of competitors. The legal basis is the Lanham Act (15 USC §1051 et seq.), which created the Principal Register and the Supplemental Register and established the USPTO as the administering agency. Trademark law coexists with common law rights: use of a mark in commerce creates enforceable rights even without registration, signified by the “TM” symbol. Federal registration adds the ® symbol and the statutory benefits that come with it.

Trademarks are frequently confused with patents and copyrights, but they protect distinct things. A patent protects a novel invention or design — the mechanism behind a product — and lasts a fixed term (20 years for utility patents, 15 years for design patents). A copyright protects original creative expression — a book, a photograph, a software codebase — and endures for the life of the author plus 70 years. A trademark protects brand identity and source identification, with no fixed expiration. Apple's iPhone design patents have expired; the iPhone trademark has not, and will not as long as Apple continues using it.

Federal registration on the Principal Register provides five concrete benefits beyond common law protection: (1) nationwide constructive notice to all subsequent users, eliminating good-faith defenses; (2) legal presumption of the registrant's exclusive right to use the mark in commerce for the identified goods and services; (3) the right to use the ® symbol; (4) the ability to record the registration with US Customs and Border Protection to stop importation of counterfeit goods; and (5) eligibility for incontestability status after five years of continuous post-registration use, which limits the grounds on which a third party can challenge the mark.

The Application and Registration Process

Applications are filed electronically through the Trademark Electronic Application System (TEAS) at www.uspto.gov. Two filing bases are available. A use-in-commerce basis applies when the applicant is already using the mark in commerce on the date of filing — it must submit a specimen showing the mark as actually used (a product label, a screenshot of a website offering services, a brochure). An intent-to-use (ITU) basis applies when the applicant has a bona fide intention to use the mark in commerce but has not yet done so; it allows applicants to secure a priority filing date before launch, with the obligation to prove actual use later.

Every application must identify the goods or services by reference to the Nice Classification system, an international system administered by the World Intellectual Property Organization (WIPO) with 45 classes: classes 1 through 34 cover goods (class 1 chemicals, class 25 clothing, class 33 wines and spirits) and classes 35 through 45 cover services (class 35 advertising and business management, class 36 financial services, class 42 scientific and technology services). A single application may cover multiple classes for an additional fee per class. The identification of goods and services must be sufficiently definite — “computer software” alone is unacceptable; “computer software for project management in the construction industry” is acceptable.

After filing, a USPTO examining attorney reviews the application, typically within three to six months. The examiner checks for procedural compliance, descriptiveness of the mark (merely descriptive marks are refused on the Principal Register absent acquired distinctiveness), and likelihood of confusion with existing registrations. If no objections exist, or after the applicant successfully argues against any office actions, the mark is approved for publication in the Official Gazette, a weekly USPTO journal. Third parties then have 30 days to file an opposition before the Trademark Trial and Appeal Board (TTAB). If no opposition is filed or any opposition is resolved in the applicant's favor, a use-in-commerce applicant receives a registration certificate. An ITU applicant receives a Notice of Allowance and then has six months — extendable by request up to 36 months from the NOA date — to file a Statement of Use (SOU) with a specimen demonstrating actual use in commerce.

Post-registration obligations keep the register current. Between the fifth and sixth years after registration, the owner must file a Section 8 Declaration of continued use (or excusable non-use) to avoid cancellation. Simultaneously, the owner may file a Section 15 Declaration to claim incontestability — available after five years of continuous post-registration use without a final adverse decision. After the initial ten-year registration period, renewal is required every ten years thereafter via a combined Section 8 and Section 9 renewal. Failure to file timely results in cancellation. The USPTO estimates that approximately 20 to 25 percent of registrations are cancelled for failure to file maintenance documents.

Mark types have expanded beyond the traditional word mark. A standard character mark protects the text regardless of font, size, or color — the broadest form of word protection. A design mark protects a specific stylized logo or graphic. Sound marks protect distinctive sounds (NBC's three-note chime, MGM's lion roar). Color marks protect a single color used as source identifier in a specific context (Tiffany blue for jewelry boxes, UPS brown for delivery vehicles) — these require proof of acquired distinctiveness and are rarely granted. Trade dress covers the overall commercial image of a product or its packaging — the shape of a Coca-Cola bottle, the layout of an Apple Store.

TESS: The Trademark Electronic Search System

Before filing, applicants and attorneys search for conflicting marks using the Trademark Electronic Search System (TESS), accessible at tmsearch.uspto.gov. TESS provides free public access to the full register. Searches can be free-form word searches, field-restricted Boolean queries (mark text, owner name, attorney, international class, filing date range), or design code searches for logo marks.

Logo marks are indexed using design search codes developed from the Leavitt design search code system. These codes categorize visual elements hierarchically: code 01.01 covers circles and ovals; code 05.01 covers natural plants and flowers; code 26.01 covers geometric squares and rectangles. An examiner comparing a stylized star logo would search the relevant design codes to identify all registered marks incorporating similar star elements, regardless of word content.

Each mark in TESS carries one of twelve status codes. Analytically, the critical distinction is Live (the mark is active — pending, registered, or published for opposition) versus Dead (the application was abandoned, the registration was cancelled, or the mark expired). Bulk data users typically filter to Live status for current competitive intelligence and retain Dead marks for historical trend analysis.

The legal standard for refusing a mark on confusion grounds is the likelihood of confusion test, governed by the thirteen DuPont factors articulated in In re E.I. du Pont de Nemours & Co. (1973). The two most heavily weighted factors are the similarity of the marks in appearance, sound, connotation, and commercial impression; and the similarity of the goods or services. Additional factors include the similarity of trade channels, the sophistication of the relevant purchasing public, the number and nature of similar marks in use on similar goods, the length of time during which there has been concurrent use without evidence of actual confusion, and whether the goods are expensive or impulse purchases (consumer sophistication correlates with price).

Famous marks receive protection beyond the likelihood-of-confusion standard under the federal dilution statute (15 USC §1125(c)). A mark that is “widely recognized by the general consuming public” — KODAK, GOOGLE, MCDONALD'S — can block registration of similar marks even for entirely unrelated goods or services, on the theory that the association weakens the mark's distinctiveness (dilution by blurring) or tarnishes its reputation (dilution by tarnishment). The dilution claim is a significant asymmetric advantage for large brand owners.

The TTAB: Inter Partes Proceedings

The Trademark Trial and Appeal Board is the USPTO's internal administrative tribunal for contested trademark proceedings. It hears two categories of inter partes dispute. An opposition is filed within the 30-day opposition window after publication in the Official Gazette by any party who believes it would be damaged by registration of the applied-for mark — typically a company asserting its own prior registered or common law rights. A cancellation is filed against an already-registered mark by a petitioner claiming the registration should not have issued or that the mark has been abandoned. TTAB proceedings are conducted largely on paper, follow trial-type rules including discovery, and culminate in written decisions.

TTAB decisions are publicly searchable at ttabvue.uspto.gov, which provides full docket history, exhibits, and final decisions. The database has become a significant secondary source for trademark law research. Notable TTAB decisions include the 2015 cancellation of six Washington Redskins registrations on disparagement grounds — a decision subsequently rendered moot by the Supreme Court's Matal v. Tam(2017) holding that the Lanham Act's disparagement clause violated the First Amendment. Craft brewery trademark disputes — often involving similar beer names in adjacent regional markets — have generated a substantial TTAB caseload as the number of licensed breweries in the United States grew from under 2,000 in 2010 to over 9,000 by 2023. Fashion house cancellation proceedings over design marks and trade dress are among the longest-running TTAB matters, sometimes spanning a decade of litigation before a final ruling.

TTAB decisions can be appealed to the US District Court for the Eastern District of Virginia (which may take new evidence) or directly to the US Court of Appeals for the Federal Circuit (on the TTAB record alone). The Federal Circuit's trademark jurisprudence is therefore the primary appellate authority for most registration disputes, with the Supreme Court hearing only the occasional case presenting a novel constitutional or statutory question.

Bulk Data: Structure and Access

The USPTO distributes trademark bulk data at bulkdata.USPTO.gov. Two data loads are available: a full annual file containing every application and registration, and daily incremental files reflecting changes since the previous day. Both are delivered in XML format conforming to the USPTO trademark XML schema, which groups data into application packages and case files.

Each application record contains: the serial number (assigned at filing), the filing date, the mark drawing code (indicating mark type), the mark literal element (the word or phrase for word marks), the owner name and mailing address, the attorney or law firm docketing information, all Nice Classification international class numbers, the full goods and services identification text for each class, any disclaimed matter (non-source-identifying elements that the applicant concedes others may use), the registration number and registration date (once granted), status code, and the prosecution history timeline of key events.

Processing the XML bulk files requires handling nested structures. In Python, bothlxml and the standard library's xml.etree.ElementTreesupport iterative parsing to avoid loading multi-gigabyte annual files into memory at once. The iterparse pattern — yielding individual application elements as the parser streams through the file — is standard practice for the USPTO XML.

For API-based access, the USPTO offers the USPTO Trademark API at api.USPTO.gov/trademark. The JSON REST API supports parameterized searches by mark status, Nice Classification, filing date range, owner name, and mark type. Rate limits apply; the API documentation specifies a cap of requests per minute for unauthenticated access, with higher limits available via an API key. The API returns paginated result sets and is suitable for targeted queries without processing the full bulk XML.

Economic Signals in Trademark Data

Trademark filing volume is a leading indicator of commercial activity. New applications represent a decision by a business to invest in brand identity — a step typically taken when a business is growing or entering a market. Filing volume tends to rise before an economic expansion as small businesses form and launch products, and to contract during recessions as capital tightens and new venture formation slows. The USPTO recorded approximately 650,000 new trademark applications in the peak filing year, up from roughly 400,000 applications annually in the mid-2010s, driven in part by a surge in e-commerce seller filings.

China is the single largest source of foreign trademark filings at the USPTO, accounting for roughly 25 percent of foreign applicant filings in recent years. This reflects both the volume of Chinese manufacturing and the strategic recognition among Chinese exporters that US trademark registration is necessary to protect brands in the American market and to use the Customs recordation mechanism against counterfeit competitors. The USPTO has implemented additional verification requirements for Chinese applicants in response to fraud concerns, including requiring US-licensed attorneys for foreign applicants.

Sectoral analysis of trademark filings maps commercial innovation. Technology products concentrate in class 9 (computers, software, electronic apparatus), while technology services cluster in class 42 (scientific and technological services, software as a service, research and design). A startup building a SaaS analytics platform typically files in both class 9 (the software product) and class 42 (the hosted service). Fashion houses file in class 25 (clothing and footwear) and class 18 (leather goods and handbags), with luxury brands typically covering all adjacent goods classes to prevent dilution by infringing merchandise. Food and beverage brands file across classes 29 through 33, with alcoholic beverages in class 33 seeing a marked increase in registration activity alongside the craft spirits boom. Financial services occupy class 36; media and entertainment span classes 41 (education and entertainment services) and 9 (recordings and software).

Python: Querying Class 42 Technology Trademark Filings

The example below queries the USPTO Trademark API for active registrations in Nice Class 42 (computer technology services) filed between 2020 and 2024. It aggregates filings by month and year, computes year-over-year growth rates, and identifies the most frequently occurring keywords in goods and services descriptions to surface fast-growing technology sub-categories.

import requests
import json
from collections import Counter, defaultdict
from datetime import datetime

# USPTO Trademark API  —  https://api.uspto.gov/trademark
# Fetch active trademark registrations in Nice Class 42
# (computer technology services) filed 2020-2024,
# then compute year-over-year growth and top keyword categories.

BASE = "https://api.uspto.gov/trademark/v1/marks/search"
HEADERS = {
    "User-Agent": "research@example.com",
    "Accept": "application/json",
}

def fetch_class42_page(start_year: int, end_year: int, offset: int = 0, limit: int = 100) -> dict:
    params = {
        "classNumbers": "42",
        "filingDateFrom": str(start_year) + "-01-01",
        "filingDateTo": str(end_year) + "-12-31",
        "markStatus": "REGISTERED",
        "offset": offset,
        "limit": limit,
    }
    r = requests.get(BASE, params=params, headers=HEADERS, timeout=30)
    r.raise_for_status()
    return r.json()

# Collect filings 2020-2024 in manageable pages
by_month = defaultdict(int)
keyword_counter = Counter()
YEARS = range(2020, 2025)

for year in YEARS:
    offset = 0
    while True:
        data = fetch_class42_page(year, year, offset=offset)
        marks = data.get("marks", [])
        if not marks:
            break
        for mark in marks:
            date_str = mark.get("filingDate", "")
            if date_str:
                dt = datetime.strptime(date_str[:10], "%Y-%m-%d")
                key = (dt.year, dt.month)
                by_month[key] += 1
            # Keyword frequency from goods/services description
            gs_text = mark.get("goodsServicesDescription", "") or ""
            for word in gs_text.lower().split():
                word = word.strip(".,;:()")
                if len(word) > 4:
                    keyword_counter[word] += 1
        offset += len(marks)
        if offset >= data.get("totalCount", 0):
            break

# Aggregate by year
by_year = defaultdict(int)
for (yr, mo), count in by_month.items():
    by_year[yr] += count

print("Class 42 registrations by year (2020-2024):")
years_sorted = sorted(by_year.keys())
prev = None
for yr in years_sorted:
    count = by_year[yr]
    if prev is not None:
        growth = (count - prev) / prev * 100
        print(f"  {yr}: {count:,}  ({growth:+.1f}% YoY)")
    else:
        print(f"  {yr}: {count:,}")
    prev = count

print("\nTop 20 technology sub-categories by keyword frequency:")
for word, freq in keyword_counter.most_common(20):
    print(f"  {word:<20} {freq:>6,}")

The keyword frequency analysis on goods and services descriptions reveals recurring sub-categories within class 42: “artificial intelligence” and “machine learning” keywords surged post-2022; “cybersecurity” and “cloud computing” have been perennially dominant since 2018; “blockchain” peaked around 2018–2019 and has since moderated. Year-over-year growth in class 42 filings exceeded 15 percent in 2021 as remote-work software adoption accelerated, then moderated in 2023–2024 as the post-pandemic surge normalized.

For research requiring mark-level granularity — tracking when a specific company filed across multiple classes, or reconstructing the full ownership chain for a portfolio acquired in an M&A transaction — the bulk XML approach is preferable. The serial number is the stable identifier throughout prosecution; the registration number is assigned only on grant and should not be confused with the serial number in historical analysis of the filing pipeline.

Scope and Limitations

The federal register covers only marks registered with the USPTO. Significant brand rights exist entirely at common law in regional markets, particularly for small businesses that have not pursued federal registration. State trademark registrations — available in all 50 states — are not reflected in the USPTO data; they confer rights limited to the state of registration and do not appear in TESS. The federal register is therefore a substantial undercount of all trademark rights in force in the United States, though it captures the great majority of commercially significant brand rights due to the incentives that federal registration provides.

Goods and services identification text is not fully standardized despite USPTO Acceptable Identification of Goods and Services Manual (IDManual) guidance. Applicants routinely customize identification language, making keyword-based classification across the corpus noisy. Natural language processing approaches — topic modeling, semantic embeddings, or fine-tuned classification against the 45 Nice classes — significantly improve categorization accuracy over simple keyword matching when working with the full bulk dataset. Owner name fields are not disambiguated in the bulk data: corporate family relationships, subsidiaries, and holding company structures must be resolved through external sources such as the SEC EDGAR registrant database or commercial entity resolution services.

For the patent side of the USPTO data ecosystem — utility and design patent grants, the PatentsView research dataset, prosecution history, and CPC classification — see USPTO Patent Data: The Federal Database Behind Every US Patent Grant and Application.

For structured financial statement data from public company filings — the SEC EDGAR XBRL API covering 7,000 active filers and every 10-K and 10-Q since 2009 — see SEC EDGAR XBRL: The Machine-Readable Financial Statement Database Behind Every Public Company.