“No results found.” That’s not the same as clear.

The phrase “trademark search” gets used for two very different things, and the confusion between them is expensive.

The first is a knockout search: a quick pass through the USPTO’s federal database to see whether your exact name, or something unmistakably close, is already taken. Knockout searches are useful. I run them constantly. Their job is to kill obviously doomed names fast, before anyone falls in love.

The second is a comprehensive clearance search, and this is the one the law actually cares about. The USPTO’s own guidance describes it: beyond the federal database, a full clearance search covers U.S. state trademark and business registries across all fifty states, domain name registries, and, critically, common-law use, meaning businesses using a name in the marketplace without any registration at all. The USPTO’s guidance is blunt about why: common-law rights that predate yours “may affect the rights provided by your federal registration.” The agency’s own recommendation for anyone who wants depth is to use a screening service or, in its words, work with an experienced U.S.-licensed attorney, because “an attorney can provide a clearance search and interpret the search results.”

Read that again, because it’s the government agency that sells the $350 filing fee telling you its free search tool is the beginning of the job, not the end of it.

Why does the law demand this much looking? Because American trademark rights don’t come from a single register. They come from use, and they’ve worked that way for over a century. Which brings us to the two Supreme Court cases that explain why your search has to look everywhere.

In Hanover Star Milling Co. v. Metcalf (1916) and United Drug Co. v. Theodore Rectanus Co. (1918), the Supreme Court established what lawyers call the Tea Rose–Rectanus doctrine: trademark rights arise from actual use of a mark in the marketplace, and an earlier user keeps its rights in the territory where it built them, even against a later user who registers first.

In practice, that means the bakery in Ohio that never filed a single form, but has sold under its name since 2015, holds real, enforceable rights in its market. If you federally register the same name for the same goods in 2026, your registration takes the rest of the map, but their pocket of rights survives inside it. And if their use predates yours entirely, they can oppose your application or petition to cancel your registration.

No federal database search will ever show you that bakery. It isn’t in the USPTO’s records, because it never asked to be. It shows up in state business registries, in domain registrations, in marketplace listings, in the places a comprehensive search is designed to look and a free search never does.

This is the single most misunderstood fact in all of trademark law, and it’s why I tell clients: the federal register is where registered rights live. The marketplace is where all rights live. A real search reads both.

The USPTO’s search system is a fine tool for what it is: a public window into federal applications and registrations. Here’s what it structurally cannot do.

It can’t see outside the federal register. State registrations, unregistered common-law users, business filings, domains: invisible, for the reasons Chapter 2 explains.

It doesn’t apply the law for you. When you search, you’re matching text. When an examining attorney reviews your application, they’re applying Section 2(d) of the Lanham Act, 15 U.S.C. § 1052(d), which refuses registration to any mark “likely, when used on or in connection with the goods of the applicant, to cause confusion” with a registered mark. Likelihood of confusion is a multi-factor legal analysis, laid out in In re E.I. DuPont DeNemours & Co. (C.C.P.A. 1973) and applied through TMEP § 1207.01: the similarity of the marks in sound, appearance, meaning, and commercial impression; the relatedness of the goods and services; the trade channels; and more.

Notice what’s not on that list: spelling. Under the sound-appearance-meaning test, KWIK and QUICK are the same word. XPRESS and EXPRESS. PHRESH and FRESH. A creative spelling that returns “no results” on an exact-text search can be a dead-on 2(d) conflict in an examiner’s analysis, and conflicts don’t need to be in your exact industry either; they need to be in related goods and services, which is a legal judgment, not a category filter.

It can’t tell you how strong your name is. Trademark law sorts marks along a spectrum of strength, from fanciful invented words down to generic terms that no one can own, a framework from Abercrombie & Fitch Co. v. Hunting World, Inc. (2d Cir. 1976) that examiners and courts still use. A descriptive name faces refusals that have nothing to do with conflicts. The database won’t mention that. An attorney will.

So the free search answers one narrow question: “Has someone federally filed this exact string or something near it?” The question your money is riding on is different: “Can I build a business on this name without someone, somewhere, having the legal power to take it from me?” Those questions are not the same size.

Here’s the part that makes skipping the search genuinely irrational rather than merely risky: a search happens either way. Under TMEP § 704, the examining attorney assigned to your application conducts their own search of USPTO records for conflicting marks as a standard part of examination. The only question is whether the first professional search of your name happens before you’ve paid the government, or after, performed by the one person with the power to refuse you.

If the examiner finds a conflict, you receive an office action, a refusal letter with a three-month response clock. You can argue, and sometimes you win, but USPTO filing fees are nonrefundable. Lose, and the government keeps the money. So does the conflict.

The numbers on going it alone are published and peer-reviewed. In “Do Trademark Lawyers Matter?” (Stanford Technology Law Review, 2013), law professors Deborah Gerhardt and Jon McClanahan analyzed roughly a quarter-century of USPTO filing data and found that 82% of applications filed by attorneys reached publication (the USPTO’s preliminary approval, when a mark clears examination), against 60% of applications filed without one. That 22-point gap isn’t intelligence. It’s clearance and drafting: knowing what will conflict before filing, and writing the application so it doesn’t invite refusals.

And the stakes of getting a confusion question wrong don’t end at the application. In B&B Hardware, Inc. v. Hargis Industries, Inc. (2015), the Supreme Court held that a Trademark Trial and Appeal Board decision on likelihood of confusion can bind the parties in later federal court litigation. Confusion determinations follow you.

One more number belongs here. Filing isn’t just defense; it’s the prize. Under 15 U.S.C. § 1057(c), your filing date becomes your nationwide constructive-use date: from that day, the law treats you as using the mark everywhere in the country. That’s what the search clears the way for, and it’s why the search without the filing is a half-finished play.

Let me show you what I actually do when a search order comes in, because “comprehensive search” should be a list, not a vibe.

Coverage

The federal register, live and dead filings, searched the way an examiner searches: sound-alikes, look-alikes, foreign equivalents, and related goods and services, not just exact text. All fifty states’ trademark registries and business filings. Domain registrations. The marketplace itself: the common-law layer where Tea Rose–Rectanus rights hide. This mirrors, and goes deeper than, the checklist the USPTO itself publishes for a comprehensive clearance search.

Analysis

Raw results are noise; most hits don’t matter. The legal work is applying the DuPont factors to each candidate: how close in sound, appearance, and meaning, how related the goods, how strong the existing mark. Then the strength assessment of your own name on the Abercrombie spectrum, because clearance isn’t only “will someone block me?” but “when I own this, how much will it be worth?”

A recommendation you can act on

Every report I deliver ends with one of three words: proceed, proceed with adjustments, or rethink. Signed by me. Then we get on a call, included with the search, and go through the findings in plain English until you can explain your own risk picture to your business partner without me in the room.

At my firm, that’s the 360 Brand Clearance Report™, step one of every 360 Brand Protection® plan: $679 flat, report in 3 business days, attorney results call included. For brands headed overseas, the international version is $979. You can see an actual report before you spend a dollar, a full sample, every page, on the sample report page, because I think you should read what you’re buying.

See a Sample Report → Order Your Trademark Search →

Whoever performs your search, a report worth paying for should let you answer five questions. Here’s how to interrogate it.

1. What was actually searched? Federal only? Or federal, state, common law, and domains? If the report doesn’t say, that’s your answer.
2. What’s the overall risk call, and who’s making it? A percentage score from an algorithm is not a legal opinion. Look for a named attorney taking a position: proceed, adjust, or rethink. Someone should be accountable for the conclusion.
3. Which findings are real threats? A good report separates the noise (dead registrations, unrelated industries) from the handful of marks that matter, and explains why they matter in 2(d) terms: this one sounds like yours, this one sells related goods, this one is strong and litigious.
4. How strong is my name itself? Where your mark sits on the strength spectrum, from invented word to descriptive phrase, predicts refusal risk that has nothing to do with conflicts, and predicts how much your registration will be worth once you have it.
5. What exactly should I do next, and by when? A report that ends without a dated next step isn’t advice; it’s data. And the date matters, which brings us to the question everyone asks last and should ask first.

A search report is a photograph. It is perfectly accurate about the day it was taken and silent about every day after. New applications arrive at the USPTO continuously, and under § 1057(c), every one of them locks its own priority date the day it’s filed. None of them will call to ask whether you were planning to use your name first.

That’s why my firm runs a rule I’d encourage you to hold any attorney to: we draft applications only on search reports that are 14 days old or less. Past two weeks, the photograph has drifted too far from the register to vouch for. Inside the window, every filing we prepare includes a complimentary before-we-file check: the day I draft your application, I re-check the register one more time, so your application goes in on current information, not a two-week-old snapshot.

The practical translation: a search is not a document you file away while you think. It’s the starting gun. Search when you’re ready to decide, decide while the picture is fresh, and if life intervenes and the window closes, refresh the search before anyone files anything. Paying for the answer twice stings; building on a stale answer costs more.

Prefer to see the work first? Read an actual sample report, every page, before you spend a dollar.