The name that describes your business most clearly is, as a rule, the name you cannot own. That is not a quirk of the registry. It is the deliberate design of trademark law, and the US Patent and Trademark Office states the reason in its own manual: refusing descriptive marks exists “to prevent the owner of a mark from inhibiting competition in the sale of particular goods” and “to maintain freedom of the public to use the language involved.”

Which means the naming question and the clearance question are the same question, asked at different points in the process. Most companies ask them in the wrong order, fall in love with a descriptive name, and discover the constraint after the domain is bought and the logo is drawn.

Everything below is published: the categories, the refusal grounds, the fee schedule and the actual examination figures.

The spectrum decides more than the shortlist does

The USPTO does not evaluate whether a name is good. It places the name on a continuum, and the position on that continuum determines whether the name is ownable at all.

The continuum, in the manual’s own words. “With regard to trademark significance, matter may be categorized along a continuum, ranging from marks that are highly distinctive to matter that is a generic name for the goods or services.”

And the point most naming exercises miss. “The degree of distinctiveness, or, on the other hand, descriptiveness, of a designation can be determined only by considering it in relation to the specific goods or services.” No name is strong or weak on its own. APPLE is arbitrary for computers and generic for apples.

Fanciful means invented. Terms “invented for the sole purpose of functioning as a trademark”, either “unknown in the language” or “completely out of common usage”. The USPTO’s own examples: PEPSI, KODAK, EXXON.

Arbitrary means real words, wrong category. Words “in common linguistic use but, when used to identify particular goods or services, do not suggest or describe a significant ingredient, quality, or characteristic”. APPLE for computers, OLD CROW for whiskey.

Suggestive means the reader has to do a little work. Marks that “require imagination, thought, or perception to reach a conclusion as to the nature of those goods or services.”

And these three are the ownable ones. “Fanciful, arbitrary, and suggestive marks, often referred to as ‘inherently distinctive’ marks, are registrable on the Principal Register without proof of acquired distinctiveness.”

The trademark distinctiveness continuum and the registrability of each categoryThe continuum of trademark distinctiveness as described in the United States trademark examination manual, running from the most protectable category to the least, together with the registrability consequence attached to each position. At the strongest end are fanciful marks, comprising terms invented for the sole purpose of functioning as a trademark, which are either unknown in the language or completely out of common usage, with the office giving the examples of Pepsi, Kodak and Exxon. Next are arbitrary marks, comprising words in common linguistic use which, when used to identify the particular goods or services, do not suggest or describe a significant ingredient, quality or characteristic of them, with the office giving the examples of Apple for computers and Old Crow for whiskey. Next are suggestive marks, which when applied to the goods or services at issue require imagination, thought or perception to reach a conclusion as to the nature of those goods or services. These three categories are together described as inherently distinctive and are registrable on the principal register without any proof of acquired distinctiveness. Below them are merely descriptive marks, which describe an ingredient, quality, characteristic, function, feature, purpose or use of the specified goods or services, and which may not be registered on the principal register absent a showing of acquired distinctiveness under section 2 subsection f, though they may be registrable on the supplemental register. At the weakest end are generic terms, which are the common everyday name for the goods or services, do not indicate source, and are not registrable on either the principal or the supplemental register under any circumstances. The determining principle stated by the manual is that the degree of distinctiveness or descriptiveness of a designation can be determined only by considering it in relation to the specific goods or services, so that no name is strong or weak in the abstract and the same word may be arbitrary for one category and generic for another.Strong to unownable, and where the line fallsInherently distinctiveUnownableFancifulInvented words, unknownin the language.PEPSI, KODAK, EXXONRegistrable as filedArbitraryReal words, unrelated tothe goods.APPLE for computersRegistrable as filedSuggestiveRequires “imagination,thought, or perception”.COPPERTONERegistrable as filedDescriptiveDescribes a quality orpurpose of the goods.”Bronzer” for suntan oilNeeds section 2(f)Generic: not a trademark at all”Not registrable on either the Principal or the Supplemental Register under any circumstances.” Examples the USPTO gives: “Bicycle” for bicycles, “Bagel shop” for a bagel shop, “E-ticket” for ticketing services.The rule that governs all five boxes”The degree of distinctiveness … can be determined only by considering it in relation to the specific goods or services.”No name is strong in the abstract. APPLE is arbitrary for computers and generic for fruit.
Position on this continuum, judged against your specific goods and services, decides whether a name is registrable at all. Source : USPTO, Trademark Manual of Examining Procedure, section 1209.01 (2026)

The descriptive trap, and the exit that is not guaranteed

Most B2B naming shortlists are full of descriptive candidates, because descriptive names feel like they are doing marketing work. They are the ones that fail.

Where the line actually sits. The USPTO’s own illustration: “the word ‘bronzer’ for suntan oil would be descriptive of an oil that helps bronze the skin and wouldn’t be registered alone as a trademark for those goods, while Coppertone is a suggestive, registered trademark for sun-tanning products.”

A useful tell. “Incongruity is a strong indication that a mark is suggestive rather than merely descriptive.” The examples the manual gives are SNO-RAKE, FRANKWURST and TINT TONE. A small mismatch between the word and the thing does real legal work.

The exit exists, and it is slow. Section 2(f) allows a descriptive mark to register on proof that it “has become distinctive”. The test is “that the primary significance of the term in the minds of the consuming public is not the product but the producer.”

And the five-year shortcut is not a right. The statute allows five years of substantially exclusive and continuous use as prima facie evidence, but the manual is explicit that reliance on it “may” be acceptable in “appropriate cases”, that the office “may, at its option, require additional evidence”, and that sufficiency “depends largely on the nature of the mark in relation to the specified goods”.

Which reframes the naming decision commercially. Choosing a descriptive name is choosing to spend five or more years and an evidentiary burden buying something a suggestive name would have given you on filing day.

One thing that is never available. Generic terms are “the ultimate in descriptiveness” and “incapable of acquiring distinctiveness under §2(f)”. No amount of use rescues them.

What actually gets applications refused

The office publishes which failure is most common, and it is not the one most founders worry about.

The single most frequent ground. “If your trademark is confusingly similar to another trademark and the goods and services are related, consumers are likely to mistakenly believe these goods or services come from the same source. This is known as a likelihood of confusion, and it’s the most common reason for refusing registration.”

And the question is narrower than it sounds. “The issue is not whether the respective marks themselves, or the goods or services offered under the marks, are likely to be confused but, rather, whether there is a likelihood of confusion as to the source or sponsorship.”

The two factors that dominate. The manual states them directly: “the similarity or dissimilarity of the marks in their entireties as to appearance, sound, connotation and commercial impression”, and “the relatedness of the goods or services as described in the application and registration(s).”

With no formula behind them. “There is no mechanical test for determining likelihood of confusion and ‘each case must be decided on its own facts.’” And “any single factor may control a particular case.”

A category most people never consider. Common phrases are refused outright, because they identify a sentiment rather than a source. The USPTO’s published examples of unregistrable phrases include “Proudly made in the USA”, “Think green” and “Drive safely”. Any B2B tagline built from ordinary encouragement falls here.

Published United States trademark examination approval rates and pendency figuresExamination performance figures published by the United States trademark office in its own dashboard data file for the third fiscal quarter of 2026, covering the quarter ending on the thirtieth of June 2026. On approval at first action, thirty six point zero three percent of base applications, meaning complete applications filed without surcharges, were approved by an examining attorney without any office action being issued, which implies that approximately sixty four percent received at least one office action. For applications carrying surcharges, meaning applications that are incomplete or that use custom identifications of goods and services rather than the pre approved identification manual, the first action approval rate was ten point five six percent. These are two distinct populations and the percentages do not sum. The office comments that pendency is significantly reduced where an examining attorney approves an application without issuing an office action, and that recent data shows stand alone base applications with no surcharges are more likely to receive first action approval than any other type of filing. On timing, first action pendency, defined as the average number of months from filing to the examining attorney’s first office action, was four point two months against a fiscal year target of five months. Total pendency, defined as the average months from filing to abandonment, allowance or registration excluding previously suspended applications and those involved in proceedings before the trademark trial and appeal board, was nine point seven seven months against a target of eleven months. Total pendency including suspended applications and inter partes proceedings was eleven point three two months against a target of fourteen months. For context, the office’s public facing estimate of the whole process is twelve to eighteen months, and the office states that there is no guarantee a trademark will ever register since it may be refused for various legal reasons.What the office publishes about its own examinationApproved at first action, with no office action issuedBase application36.03%With surcharges10.56%Two separate populations. Roughly two thirds of base applications receive at least one office action.Pendency, FY2026 Q3, against the fiscal year targetFirst action pendency4.2monthstarget 5Total pendency9.77monthstarget 11Incl. suspended and inter partes11.32monthstarget 14And the sentence to keep in mind”There’s no guarantee your trademark will ever register, as it may be refused for various legal reasons.”
The office publishes how often an application clears examination untouched, and how long the stages take. Both are lower than the public estimate implies. Source : USPTO Trademarks Dashboard data file, FY2026 Q3 (2026)
The dominant factors in a United States likelihood of confusion refusalHow the United States trademark examination manual frames a refusal based on likelihood of confusion, which the office identifies as the most common reason for refusing registration. The manual states that although the weight given to the relevant factors may vary, two are key considerations in any likelihood of confusion determination. The first is the similarity or dissimilarity of the marks in their entireties as to appearance, sound, connotation and commercial impression, which means a mark is not compared element by element but as a whole and across four dimensions including how it sounds when spoken. The second is the relatedness of the goods or services as described in the application and in the cited registration, which means the comparison is against the written identification rather than against what the businesses actually do in the market. Four further factors must be considered where there is pertinent evidence in the record, namely the similarity or dissimilarity of established and likely to continue trade channels, the conditions under which and the buyers to whom sales are made distinguishing impulse purchasing from careful and sophisticated purchasing, the number and nature of similar marks already in use on similar goods, and the existence of a valid consent agreement between the applicant and the owner of the previously registered mark. Two constraints govern how these are applied. There is no mechanical test for determining likelihood of confusion and each case must be decided on its own facts, and any single factor may control a particular case. The underlying question is also narrower than it appears, since the issue is not whether the marks themselves or the goods offered under them are likely to be confused but whether there is a likelihood of confusion as to the source or sponsorship of the goods or services.The most common refusal, and what decides it”The most common reason for refusing registration.” Not descriptiveness. Confusion.Key factor 1Similarity of the marks “in theirentireties as to appearance, sound,connotation and commercial impression”.Key factor 2Relatedness of the goods or services”as described in the application andregistration(s)”. Not as traded.Considered where the record supports themTrade channelsBuyer sophisticationSimilar marks in useA valid consent agreement with the owner of the cited registration”There is no mechanical test … each case must be decided on its own facts.”And: “Any single factor may control a particular case.”
The manual names two key considerations out of the DuPont factors, and states that any single factor may control a case. Source : USPTO, Trademark Manual of Examining Procedure, section 1207.01 (2026)

What a database search does not tell you

This is the failure that costs the most, because it happens after launch rather than during examination.

The examiner’s search is narrower than yours needs to be. “Although the trademark examining attorney assigned to your application will also conduct a search, they will only determine whether there are conflicting trademarks in our federal database. It’s your responsibility to search state trademark databases and the internet.”

Because rights exist without registration. “Results in the USPTO’s search database are limited to federal trademark applications and registrations and do not include the trademarks of other parties who may have trademark rights but no federal registration.”

And those unregistered rights can outrank you. Common law rights are “based solely on someone using their trademark in commerce within the United States” and “may affect the rights provided by your federal registration if the common-law use was earlier”.

State registers are partial by design. A state registration “creates rights in that state only”, and “not all states have trademark registration databases, which means that third parties will not be aware of your rights in that trademark.”

So the search list is longer than one database. The USPTO’s own enumeration includes the federal search system, the Official Gazette, state trademark and business registries, domain name registries, Madrid Monitor, the Global Brand Database, EUIPO, TMview, and plain internet searching for common law use.

And it names the three things you are avoiding. A refusal based on likelihood of confusion, an opposition or cancellation proceeding, and a trademark infringement lawsuit. Only the first one is cheap.

Division of trademark clearance search responsibility between the examining attorney and the applicantDivision of responsibility for trademark clearance searching between the examining attorney assigned to an application and the applicant. The examining attorney searches only the federal database of registered and pending trademarks, and the office states explicitly that the attorney will only determine whether there are conflicting trademarks in that federal database and that it is the applicant’s responsibility to search state trademark databases and the internet. Everything outside the federal database therefore falls to the applicant. The office’s own enumeration of a comprehensive clearance search comprises the federal trademark search system, the Trademark Official Gazette, United States state trademark and business registries covering the fifty states plus Washington District of Columbia and Puerto Rico, domain name registries, Madrid Monitor, the Global Brand Database, the European Union Intellectual Property Office, TMview, and multiple internet search engines checked for goods and services that may conflict. The reason this matters is that trademark rights exist without any registration. Results in the office database are limited to federal applications and registrations and do not include the trademarks of parties who hold rights but no federal registration. Such common law rights are based solely on use of the trademark in commerce within the United States with specific goods or services, and may affect the rights provided by a federal registration where the common law use predates the use supporting that registration. State registration creates rights only within that state, and not all states maintain a searchable trademark database, which means third parties may have no way of discovering rights registered there. The three outcomes a comprehensive search is intended to avoid are a refusal to register based on likelihood of confusion, an opposition proceeding preventing registration or a cancellation proceeding against a registration, and a trademark infringement lawsuit.Who searches whatThe examining attorneyThe federal database ofregistered and pendingtrademarks.That is the whole of it.One database.You, per the office’s own listFederal search systemMadrid MonitorTrademark Official GazetteGlobal Brand DatabaseState trademark and business registriesEUIPO and TMviewDomain name registriesInternet, multiple engines”It’s your responsibility to search state trademark databases and the internet.”Why the gap is dangerousCommon law rights arise from use alone, appear in no register, and “may affect the rights provided by your federal registration”.The three things the search avoids: a refusal, an opposition or cancellation, and an infringement lawsuit.Only the first one is cheap.
The examining attorney checks one database. The rest of the search, where the expensive conflicts live, is yours to run. Source : USPTO, Why search for similar trademarks, and Comprehensive clearance search (2023)

The price is not the price on the page

The 2025 fee changes replaced the old two-tier application with a single base fee plus surcharges, which means the published number is a floor.

The base. $350 per class for applications under Sections 1 and 44, effective 18 January 2025. The old TEAS Plus and TEAS Standard tiers no longer exist.

Per class, not per application. The USPTO’s own worked example: golf shirts, dress shirts and t-shirts sit in one class and cost $350; add custom t-shirt printing services and you are in two classes at $700.

The surcharges that catch people. $100 per class for insufficient information. $200 per class for using the free-form text box instead of the ID Manual to describe goods and services, plus $200 per additional 1,000 characters. Writing your own description of what you sell is a paid option.

And it correlates with outcomes. Base applications had a 36.03 percent first-action approval rate in FY2026 Q3; applications carrying surcharges had 10.56 percent. The office notes that “stand-alone Base applications with no surcharges are more likely to receive first action approval than any other type of filing.”

The intent-to-use path costs extra too. $150 per class for a Statement of Use, $125 per class for each six-month extension. Filing before you have launched is legitimate and it is not free.

And renewal is a recurring line. $325 per class for the Section 8 declaration and $325 for the Section 9 renewal, which the USPTO illustrates as $650 per class every ten years, or $1,300 for two classes.

United States trademark fee structure following the 2025 fee changesStructure of United States trademark fees following the fee changes that took effect on the eighteenth of January 2025, which replaced the previous two tier electronic application system with a single base application fee plus a set of surcharges. The base application fee for applications filed under sections one and forty four is three hundred and fifty dollars per class of goods or services, replacing the former two hundred and fifty dollar and three hundred and fifty dollar tiers. The office illustrates the per class nature of the fee by noting that a clothing line comprising golf shirts, dress shirts and t-shirts falls within a single class and costs three hundred and fifty dollars, whereas adding custom t-shirt printing services places the applicant in two classes at a total of seven hundred dollars. The surcharges are one hundred dollars per class where the application contains insufficient information, two hundred dollars per class for using the free form text box rather than the pre approved identification manual to describe the goods and services, and a further two hundred dollars per affected class for each additional group of one thousand characters in that free form box beyond the first thousand. Applications filed through the international system under section sixty six subsection a cost six hundred dollars per class, raised from five hundred. For applications based on intent to use, a statement of use costs one hundred and fifty dollars per class and each six month extension request costs one hundred and twenty five dollars per class. Maintenance costs three hundred and twenty five dollars per class for the section eight declaration of use and three hundred and twenty five dollars per class for the section nine renewal application, which the office illustrates as six hundred and fifty dollars per class every ten years, or one thousand three hundred dollars for a registration covering two classes. The surcharge structure correlates with outcomes, since base applications carrying no surcharges had a first action approval rate of thirty six percent in the third fiscal quarter of 2026 against ten point five six percent for applications carrying surcharges.What an application actually costs, per classThe base$350per class, Sections 1 and 44Replaced TEAS Plus ($250) and TEASStandard ($350) on 18 January 2025.The surchargesInsufficient information+$100Free-form goods description+$200Each extra 1,000 characters+$200If you file before launchingStatement of Use, per class$150Six-month extension, per class$125And every ten years after thatSection 8 declaration, per class$325Section 9 renewal, per class$325Why the surcharges matter beyond their price36.03% of base applications were approved at first action in FY2026 Q3. For applications carrying surcharges: 10.56%.Describing your goods in your own words rather than from the ID Manual is a paid option, and a slower one.
One base fee replaced the old two-tier system, and the surcharges are the part that decides what an application actually costs. Source : USPTO, Summary of 2025 trademark fee changes, and the current fee schedule (2025)

Filing before you launch, and going international

Two decisions that look procedural and are not.

Intent to use is the majority path. In FY2026 Q3, 47 percent of filings were intent-to-use under Section 1(b) against 41 percent based on actual use. It is the normal route, not an exception.

And it buys priority. “You may file even before you use your mark, which means you can get an earlier application filing date than a possible competitor.”

With a hard ceiling. After the Notice of Allowance you have six months, extendable five times, for “a maximum possible extension time of 36 months”. Miss it and “your application will be abandoned and your only option will be to file a new application with new fees.”

On symbols. You may use TM for goods or SM for services “even if you haven’t filed an application”. The R symbol is only available once registered, and “only … with the trademark for the goods or services listed in the federal trademark registration.”

On international filing, the trap has a name. A Madrid international registration “remains dependent on the basic mark for a period of five years”. If the US application dies in that window, “the international registration will be cancelled to the same extent”. This is the central attack, and WIPO notes it is often self-inflicted: the basic mark frequently lapses “due to the inaction of the holder”.

Which produces one concrete piece of sequencing advice, from the USPTO itself. “If you’re using a pending U.S. trademark application as the basis for your international application, we strongly recommend you wait to file your international application until you receive your first USPTO office action.”

What to do with this

Build the shortlist against the spectrum before you build it against the domain. Sort each candidate into fanciful, arbitrary, suggestive or descriptive relative to your actual goods and services, and delete the descriptive ones unless you are willing to fund five years and an evidence file. A shortlist also shortens faster when the brand around it is already defined, because the name, the strapline and the tone of voice are settled together as one verbal territory.

Then run the clearance search the way the office describes it: the federal database, state registries, and the open internet for common law use. The examiner will only do the first of those, and the other two are where the expensive surprise lives.

When you file, use the ID Manual rather than your own wording. It is a $200 per class difference on the invoice and a much larger difference in the odds of clearing examination without an office action.

If a rename is what brought you here, the sequencing matters more than the name: read renaming a company without losing what you built. And if the domain is driving the shortlist, choosing and keeping a domain name covers why it should not.