USPTO Trademark Search — A Practitioner’s Tutorial.

Before you search: pick a Nice class.

The single most important number in any trademark search is the Nice class. The Nice Classification is an international taxonomy of goods and services with 45 classes (1–34 for goods, 35–45 for services). Two identical marks can coexist if they are in different Nice classes and the holders’ markets don’t overlap — the canonical example is Delta, registered to Delta Air Lines in Class 39 (transport) and to Delta Faucet in Class 11 (plumbing). They coexist because their markets don’t collide.

For software companies, the killer classes are Class 9 (downloadable software, recorded media, mobile apps) and Class 42 (software-as-a-service, technological consultancy). We break this down at length in Nice Classes for Software. Pick your class before you run TESS, not after. The class drives the search filter; the search filter drives the result set; the result set drives the verdict.

Mode 1 — Basic Word Mark Search.

Open tmsearch.uspto.gov and choose Basic Word Mark Search (New User). This is the entry point everyone starts at, and the one that handles literal-string queries.

1. Type your candidate as a single word, lowercase, no quotation marks. TESS is case-insensitive — stratagem returns the same records as Stratagem or STRATAGEM.
2. Set Field to Combined Word Mark. This searches both the word mark and the translation/transliteration fields. The default is too narrow.
3. Submit. Filter the result list by Live records. Dead marks (status: abandoned, cancelled, expired) don’t bar your registration, though a recently-dead famous mark can still create reputational and dilution risk.
4. Sort by Filing Date descending. The most recent applications are the most likely to be examined alongside yours. Pending applications queue in front of your application during examination.

Basic mode catches literal-string matches and trivial variations. It does not catch phonetic equivalents, translations, or design marks. That’s what modes 2 and 3 are for.

Mode 2 — Structured Word Mark Search.

Structured mode lets you combine fields with Boolean operators. This is where clearance-search attorneys live. Choose Word and/or Design Mark Search (Structured) from the TESS landing page.

The four most useful field codes are:

• [BI] — basic index, the combined word mark field
• [IC] — international class (the Nice class number)
• [LD] — live/dead status indicator
• [GS] — goods and services description

A canonical structured query for a SaaS candidate in Class 42 looks like: stratagem[BI] AND (042)[IC] AND live[LD]. The result is the set of Live marks containing “stratagem” in the word mark, filed in Class 42. This is materially narrower than the basic-mode result — and materially more relevant.

You can union Nice classes when your candidate is genuinely multi-class: stratagem[BI] AND ((009)[IC] OR (042)[IC]) AND live[LD] for a downloadable + SaaS product.

Mode 3 — Free-Form Search (phonetic equivalents).

Free-form mode is the only TESS interface that supports truncation operators. This is where you catch phonetic-similar marks — the marks that fail factor (1) of the DuPont test (similarity of sight, sound, and meaning) but don’t share a single literal character with your candidate.

The two truncation operators:

• $ — the right-truncation wildcard. strat$ matches strata, stratos, strategist, stratagy, stratagem, etc. Use this to fan out the front of the candidate.
• ? — the single-character wildcard. strat?gem matches stratagem, stratogem, stratagem with a typo, etc.

For a candidate like Stratagem, a good free-form sweep is: strat$[BI] AND (042)[IC] AND live[LD] followed by strat?gem[BI] AND live[LD]. The first surfaces every Class 42 Live mark that starts with “strat”; the second surfaces every Live mark across all classes that’s a one-character variant of your candidate.

Most §2(d) refusals cite phonetic-similar marks, not exact matches. If your free-form sweep returns a Live senior mark in your Nice class within a small edit distance of your candidate, treat it as a refusal risk and read on at §2(d) Likelihood of Confusion.

Mode 4 — Browse Dictionary.

Browse mode is the one most founders never use and the one most attorneys occasionally do. It lets you scan an alphabetized list of indexed terms around your candidate — a kind of dictionary view of the federal register. This is how you find marks you wouldn’t have known to search for.

Choose Browse Dictionary (View Indexes) and enter the first few letters of your candidate. TESS returns a paginated list of every indexed term starting with that prefix, with hit counts. If you see strataform (147 hits) two entries away from your candidate stratagem (3 hits), that’s a hint to widen your phonetic sweep.

Browse mode is also the right way to find a candidate that’s descriptive in a way you didn’t notice. If the prefix has hundreds of indexed terms, your candidate likely shares a common root with a category of descriptive marks, which raises §2(e)(1) descriptiveness risk regardless of any specific senior mark.

Reading a TSDR record.

For each hit in your result set, open the TSDR (Trademark Status and Document Retrieval) record. TSDR is the deep view of a single trademark — the application’s prosecution history, the goods and services description, the owner of record, every office action and response. Five fields drive the clearance decision.

1. Status: registered, pending, abandoned, cancelled. Live records bar your registration. Dead records do not, though a famous dead mark can still carry residual rights.
2. International class: the Nice number. If this matches yours, you have a class-collision risk. If it doesn’t, the goods-and-services language may still produce overlap; read on.
3. Goods and services description: the freeform text of what the holder uses the mark for. This is the language an examiner compares to your application’s ID. Overlap here is what triggers §2(d).
4. Owner: the legal entity that owns the registration. Note the owner for the TTABVUE lookup below. An owner who has opposed three junior applications in your class is telegraphing they’ll oppose yours.
5. Prosecution history: the timeline of every office action, response, suspension, publication, and registration. If the mark survived a §2(d) refusal during examination, the response brief is in the prosecution history — and reading it is a free education in how examiners argue.

Five common TESS mistakes (and the corrections).

The most common errors founders make on the TESS interface are not random — they cluster on the same five mistakes. The corrections are quick.

• Searching with quotes. TESS does not treat quoted strings as phrase queries; it treats them literally and rarely returns useful results. Drop the quotes.
• Filtering out pending applications. A pending application bars your registration if the examiner reaches it before they reach you. Always include Live records (pending + registered), not just registered.
• Searching only the word mark. The translation and transliteration fields can match foreign-language marks that sound similar in English. Use the Combined Word Mark field.
• Ignoring the goods-and-services language. Two marks in the same Nice class with completely different goods may coexist. The Nice number is a coarse filter; the GS language is the fine filter.
• Stopping at TESS. TESS is the federal register, but the clearance picture also requires TTABVUE for opposition patterns and Madrid Monitor for international registrations. Three registries, not one.

The two next steps after TESS.

TESS catches roughly 80% of clearance disasters. The remaining 20% live in two other registries that founders systematically miss.

• The TTAB docket at ttabvue.uspto.gov. Roughly 647,000 proceedings indexed. Surfaces opposition patterns by owner. Walk-through at TTAB Opposition Search.
• The Madrid Protocol register at wipo.int/madrid/monitor. Surfaces international registrations that designate the US. Walk-through at Madrid Protocol — One Application, 130 Countries.

How Etymolt verifies trademarks.

The four modes above are what we run in parallel when a candidate hits POST /v1/verify. The phonetic-equivalents sweep is automated against a 12.7M-record TESS mirror plus the 108K Madrid international registrations and the 647K TTAB proceedings. Every flag traces to a serial number. The verdict is attested — every signal carries a citation to the underlying record — and permalinked so it’s citation-grade for diligence. Full methodology at /methodology. The companion guide How to Check If a Brand Name Is Trademarked covers the four-level ladder in full.