Trademark Clearance — The Forensic Guide for Founders.

What is trademark clearance.

Trademark clearance is the evidentiary practice of confirming that a brand name is available for use and registration in the jurisdictions and trade classes that matter to a business. A clearance report does three things: (a) it surfaces senior marks that could bar registration on grounds of likelihood of confusion under 15 U.S.C. §1052(d) — the statute behind every §2(d) refusal; (b) it screens for absolute grounds of refusal — descriptiveness, deceptiveness, surnames, geographic marks; (c) it surfaces oppositions, cancellations, and famous-mark dilution risk.

The U.S. USPTO received 736,778 trademark applications in fiscal year 2024 and refused a material share of them on substantive grounds. The most common refusal basis is §2(d) likelihood of confusion, followed by §2(e)(1) descriptiveness. The refusal rate varies wildly by Nice class: Class 9 (downloadable software) and Class 42 (software-as-a-service) consistently sit at the high end. The clearance check is what catches these refusals before you spend money on the brand.

The cost of skipping clearance is not linear. A refusal in week 4 costs you a naming-storm. A refusal in week 16 — after the trademark application has been filed, the website is live, the customer base has the name committed to memory, the legal hours are billable — costs five figures and eleven weeks. Once is bad luck. Twice is a system problem. Etymolt exists because the second time was the system problem.

For most founders, the right clearance posture is to run a free Level 1 check first, an automated Level 3 check on the finalists, and a Level 4 attorney consult on the candidate you are about to spend money on. The ladder below is the practitioner’s decomposition.

The four levels of clearance check.

Every clearance check sits at one of four levels of rigor. Each level catches a different class of failure. The right level for your candidate depends on stakes, time, and budget. We covered the full ladder in How to Check If a Brand Name Is Trademarked; here is the executive summary.

Level 1 — USPTO TESS direct (free, ~30 min)

The Trademark Electronic Search System at tmsearch.uspto.gov is the canonical federal register. Four search modes — basic, structured, free-form, browse — cover roughly 80% of straightforward conflicts. The limit is the operator: TESS is case-insensitive but otherwise literal, and the phonetic-equivalents query requires you to know the truncation grammar. We walk through every mode in How to Search USPTO Trademarks.

Level 2 — Freemium tools ($0–$29, ~5 min)

Trademarkia, Markify, Trademark.com, and a half-dozen other operators wrap the USPTO public API behind a friendlier UI. They’re fast, they’re cheap, and they catch the obvious head-on collisions. The limits are well-known to attorneys: shallow phonetic matching, opt-out class filtering, no TTAB coverage, no Madrid coverage, no famous-mark dilution screen, no descriptiveness probability.

Level 3 — API / automated check ($0 first 5, ~3 seconds)

The automated tier is what Etymolt occupies. A single API call (POST /v1/verify) runs USPTO TESS, UKIPO, EUIPO, the Madrid Protocol register, TTAB, a phonetic-distance scorer, a Nice-class targeting engine, and four other axes (domain, cultural, sound symbolism, pronunciation resilience) in parallel. The verdict comes back in under three seconds with every flag traced to a record number. Five free per IP. No signup. Every flag is attested and permalinked. The full methodology is at /methodology.

Level 4 — Attorney clearance opinion ($1,500–$3,500, ~1–2 weeks)

The attorney tier is the only level that produces a defensible legal opinion. A clearance-search attorney will typically charge $1,500–$3,500 for a full written opinion letter, which is the unique deliverable. No Level 1–3 tool produces an opinion letter. If a registration is later challenged, the opinion letter is what shifts the burden of proof. For Series A diligence, for Class 9 and Class 42 software disputes, for any brand-defining product, Level 4 is cheap insurance.

The 13 DuPont factors.

Every §2(d) likelihood-of-confusion analysis sits on the same 13-factor framework, first articulated in In re E.I. du Pont de Nemours & Co. (476 F.2d 1357, C.C.P.A. 1973). The factors cover similarity of the marks, similarity of the goods, channels of trade, conditions of purchase, evidence of actual confusion, and eight more. The two factors that drive most refusals are factor (1), similarity of the marks in appearance, sound, connotation, and commercial impression; and factor (2), similarity of the goods or services. If your candidate is phonetically similar to a senior mark in the same Nice class, the refusal probability is high regardless of what the other eleven factors say.

Examiners weigh the factors differently in different cases. The DuPont framework is not a checklist; it’s a totality test. The board and the courts have made clear repeatedly that not every factor is relevant in every case, and the weight of each varies with the fact pattern. A famous senior mark in a related class will dominate the analysis; a niche senior mark in a distant class can be overcome with third-party-use evidence and conditions-of-purchase arguments. Knowing which factors to push on in your specific fact pattern is the practitioner skill that §2(d) responses turn on.

The cluster page §2(d) Likelihood of Confusion — The 13 DuPont Factors walks through each factor in turn, with examiner examples, refusal-rate statistics, and how clearance-search practitioners argue around them. Across the ~220 final TTAB decisions we surveyed from the 2024 docket, factor (1) appeared in 95% of opinions and factor (2) in 93%. The other eleven factors trail behind. Concentrating clearance-search effort on factor (1) and factor (2) is the evidence-driven default.

The seven cluster surfaces.

The clearance space decomposes into seven sub-topics that founders ask about repeatedly. Each has its own dedicated guide; each links back here. Read them in any order — though if you read only one, read the DuPont factors page, because that’s the law every examiner applies.

• USPTO trademark search tutorial

  How to Search USPTO Trademarks — The Four Modes

  A step-by-step walk-through of USPTO TESS — basic, structured, free-form, browse. Phonetic-equivalents, Nice-class targeting, truncation operators, TSDR records.

  Read →

• TTAB opposition search

  TTAB Opposition Search — Reading 647K Proceedings

  What the Trademark Trial and Appeal Board is, how to read ttabvue.uspto.gov, finding §2(d) opposition patterns, anatomy of a TTAB decision.

  Read →

• famous mark trademark

  Famous Marks — The Distance Rule and §43(c) Dilution

  Coca-Cola, Apple, Google, Disney — the marks that bar your registration regardless of Nice class. What “famous” means at law, the §43(c) framework, real refusal cases.

  Read →

• §2(d) likelihood of confusion

  §2(d) Likelihood of Confusion — The 13 DuPont Factors

  The full breakdown of every DuPont factor — similarity of marks, similarity of goods, channels of trade, conditions of purchase, evidence of actual confusion, all 13.

  Read →

• Nice classification software

  Nice Classes for Software — Class 9 vs Class 42

  The killer classes for SaaS and AI startups. How to pick the right ID-of-goods, the USPTO Trademark Manual references, common Class 35 mistakes founders make.

  Read →

• Madrid Protocol trademark

  Madrid Protocol — One Application, 130 Countries

  WIPO’s Madrid System explained. When to use it, when to file national. Cost comparison, the 108K-IR partial corpus Etymolt indexes, the Basic Application requirement.

  Read →

• trademark vs domain name

  Trademark vs Domain Name — The Distinction Founders Miss

  Why registering yourname.com does not protect you. UDRP, URS, sunrise procedures, the difference between trademark rights and DNS records, real cybersquatting cases.

  Read →

When to use Etymolt vs an attorney.

The honest answer: for material trademark questions, you want both. Etymolt is the Level 3 tier on the clearance ladder — we surface signals, we trace every flag to a record number, we publish the methodology, and we permalink the verdict so it’s citation-grade for investor diligence and partner intake. We are not a law firm. The Bureau Model is documented at /legal-ops: we operate as an evidentiary bureau, not a counsel. The output is a verdict with citations; the legal opinion belongs to the attorney.

For naming-storming and for narrowing a longlist to a shortlist, Etymolt is the right tool. The five-axis verdict tells you which candidates are cleared on trademark, domain, cultural, sound symbolism, and pronunciation grounds. The ones that fail you can drop. The ones that pass you advance. This is the workflow for the early-stage founder whose candidate set is a longlist of fifty names and whose budget for clearance is the cost of a few API calls.

For the candidate you’re about to file an ITU application on, for any Class 9 or 42 collision risk, for any famous-mark proximity question, for any Madrid Protocol filing, for any Series A diligence cycle — you want a Level 4 attorney consult on top of whatever Level 3 work we’ve done. We refer to attorney partners; the Bureau Model is explicit about the handoff. The verdict permalink from Etymolt is the artifact your attorney starts from, which compresses billable hours.

The cost arithmetic matters. A typical attorney clearance opinion runs $1,500–$3,500 per candidate and takes one to two weeks. Running ten candidates through the attorney before narrowing to one is rarely the right allocation of budget or time. Running ten candidates through Etymolt, narrowing to one or two finalists based on the Level 3 verdict, and then running the finalists through the attorney is the allocation that most founders actually do once they’ve been through the cycle once.

Clearance signal, not legal advice. Read that disclaimer twice. It is the discipline that keeps us evidentiary instead of promotional. We surface what the registries say; an attorney tells you what to do about it. The Bureau Model exists to make that handoff clean.

The five clearance mistakes we see most often.

Across the first ~1,200 candidates founders have run through Etymolt, a small set of mistakes recurs with depressing regularity. They are the mistakes that caused my own $40,217 bill. They are catchable. The pattern:

• Searching the wrong Nice class. Founders default to Class 35 (advertising and business services) because the word sounds right, missing the actual class of their goods (Class 9 for downloadable software, Class 42 for SaaS). We cover this trap in detail at Nice Classes for Software.
• Skipping the phonetic-equivalents sweep. Founders search the literal string in TESS and stop. Phonetic-similar marks — the marks that actually trigger most §2(d) refusals — require the free-form truncation grammar that founders don’t know about. The walkthrough is at USPTO Trademark Search Tutorial.
• Ignoring TTABVUE. Founders read the senior mark’s registration record and stop, never checking whether the owner has filed oppositions before. A litigious senior-mark owner is a free signal that most founders never read.
• Missing famous-mark proximity. Founders check their Nice class and don’t check whether their candidate brushes against a tier-1 famous mark in any class. The §43(c) dilution risk operates cross-class. Coverage at Famous Mark Rules.
• Treating the domain as the trademark. Founders register the domain, exhale, and assume the brand is protected. The domain is a DNS contract; the trademark is a statutory right. They do not interconvert. The full distinction is at Trademark vs Domain Name.

Each of these mistakes is recoverable if caught in week 4. None of them is cheaply recoverable if caught in week 16. The clearance check exists for the gap between those two weeks.

Further reading.

• How to Check If a Brand Name Is Trademarked — 2026 Guide — the four-level ladder in full, with the TESS walk-through, the TTAB lookup, and the five questions to ask of every hit.
• The Five Ways a Brand Name Dies — trademark is one death, not all of them. The other four kill names at higher rates than founders expect.
• The full five-axis methodology — what Etymolt checks, how we score, the registries we cite, the calibration we publish.