LogoVerifyLogoVerify

LogoVerify research · October 23, 2026

What blocks a trademark at the USPTO: 4,980 refusals, read

We read 33,246 USPTO examination letters on 31,609 U.S. trademark applications, in equal random samples of names, logos with words and designs, whose first examination (2019–2022) ended in an office action. 4,980 were refused because the mark was likely to be confused with an existing one (Section 2(d)). This page has the findings, the tables, the method and the data.

Key findings

  1. 87% of refused name applications were not an exact match to the mark that blocked them; 10% shared no word with it at all (n = 2,365).
  2. 59% of refused logos with words were blocked by plain word marks alone, with no stylised lettering or design in any cited mark (n = 2,156).
  3. In 65% of refused logos with words, the examiner argued that the words outweigh the design and the design does not avoid confusion (n = 2,044).
  4. 25% of refused name applications were blocked by a mark that has a design element (n = 2,366).
  5. In about 1 in 3 refusals the goods were only related, not identical: 30% for names and 31% for logos with words (n = 2,242 and 2,044).
  6. Half of the refusals of names said the two marks sound alike or are phonetic equivalents (50%, n = 2,242).
  7. 27% of refused name applications cited a mark outside the applicant's classes: 17% only in a related class, 10% in another class (n = 2,365).
  8. About 1 in 6 conflicts involved an earlier application that was not registered yet (16% of names, 17% of logos with words).
  9. The median wait from filing to the likelihood-of-confusion refusal was 128 days for names and 131 for logos with words.
  10. 71% of refused name applications ended abandoned, against 45% of name applications in the same sample that were not refused (n = 2,371).

Source: LogoVerify analysis of USPTO TSDR records, data frozen 2026-10-23 10:32 UTC. What it means for a new brand: searching only for the exact name misses most of the marks that get applications refused.

Most refused names weren’t an exact match

How each refused name related to the closest mark the examiner cited, in our study of USPTO refusals. A search for the exact name finds only the highlighted row: 13.3% of the refusals.

The refused name and the cited markNames · n = 2,365Logos with words
One name inside the othere.g. LOVELY PEPA refused for LOVELY
31.4%
33.1%
A shared worde.g. UNCOMMON BEAUTY refused for UNCOMMON SCENTS
23.9%
22.6%
One or two letters offe.g. FLYOU refused for FLYOO
16.7%
13.4%
The exact same nameexact-name searche.g. MONARCH refused for MONARCH
13.3%
16.9%
No shared word at alle.g. ROADWISE refused for STREETWISE
9.9%
9.9%
Words joined or splite.g. TIPTOP refused for TIP TOP
2.3%
1.8%
A sound-alike spellinge.g. TUFF BOXX refused for TOUGH BOX
1.6%
1.4%
Plural or singulare.g. DUKAR refused for DUKARS
0.9%
0.9%

Each refusal counted once, under the first rule that fits: exact, joined or split, plural, one or two letters off, sound-alike, one name inside the other, a shared word, none. “No shared word” cases were refused on sound, meaning, the design or the overall impression. Public USPTO records, October 2026.

Real refusals, in the examiner’s words

One application from the study for each kind of resemblance, with the mark the examiner cited and a line from the letter. Public USPTO records; marks and serial numbers only.

The data, by mark type

The three mark types were sampled in equal numbers, so every figure is for one type and none are pooled. The unit is the application; the cited marks are every mark in its 2(d) letters. Download all tables (CSV)

How often the first examination ended in a likelihood-of-confusion refusal

Share of sampled applications with a Section 2(d) refusal, by mark type.
MeasureNames (standard characters)n = 10,536Logos with wordsn = 10,536Designs onlyn = 10,537
Refused under Section 2(d)22.5%20.5%4.3%

About 1 in 3 refusal letters cited two or more marks: 31.4% for names, 34.5% for logos with words, 31.7% for designs (share of refusal letters, not of applications).

What kind of mark blocked the application

Refused applications, by the drawing types of the marks the examiner cited.
MeasureNames (standard characters)n = 2,366Logos with wordsn = 2,156Designs onlyn = 448
Only plain word marks cited67.6%58.6%9.8%
Stylised lettering, no design, among the cited marks7.1%9.6%—
A cited mark with a design25.3%31.8%88.6%

USPTO drawing types: plain words are 1 and 4 (standard characters), stylised lettering without a design is 5, marks with a design are 2 and 3.

How the refused name related to the closest cited mark

Refused applications with words, each counted once under the first rule that fits (normalised text: lower case, letters and digits).
MeasureNames (standard characters)n = 2,365Logos with wordsn = 2,141
One name inside the other31.4%33.1%
A shared word23.9%22.6%
One or two letters off16.7%13.4%
The exact same name13.3%16.9%
No shared word at all9.9%9.9%
Words joined or split2.3%1.8%
A sound-alike spelling1.6%1.4%
Plural or singular0.9%0.9%

Rules in order: exact; joined or split; plural (trailing s or es); one or two letters off (names of four letters or more); sound-alike (same Double Metaphone key of three characters or more); one name inside the other; a shared word (two letters or more, not a stop word, or a word of four letters or more inside a compound); none.

Where the cited mark sat in the classes

Refused applications, by the closest class relation between the application and any cited mark.
MeasureNames (standard characters)n = 2,365Logos with wordsn = 2,151
Same class73.1%74.9%
Only a related class17.4%14.5%
Only another class9.6%10.6%

Related classes follow the USPTO's coordinated class table. Some older registrations keep pre-2002 class numbers (restaurant services were class 42, now 43), so “another class” includes renumbered classes.

Conflicts with an application that was not registered yet

Applications with a conflict (a 2(d) refusal or a prior-pending notice) that were told about an earlier application still pending.
MeasureNames (standard characters)n = 2,787Logos with wordsn = 2,561Designs onlyn = 568
Earlier pending application involved16.3%17.3%23.1%

A lower bound: only notices that name a pending application are counted.

How long it took, and how it ended

Refused applications: days from filing to the first 2(d) letter, and status today.
MeasureNames (standard characters)n = 2,371Logos with wordsn = 2,160Designs onlyn = 449
Median days from filing to the refusal128 (n 1,933)131 (n 1,633)142 (n 348)
Middle half of that wait, in days91–22393–233—
Abandoned70.7%67.3%63.0%
Registered26.0%29.4%33.0%
Abandoned, same sample, not refused under 2(d)45.0%35.7%32.5%

What the examiner argued in the refusal

Refused applications whose 2(d) section could be found, by what that section says (estimated from its wording).
MeasureNames (standard characters)n = 2,242Logos with wordsn = 2,044Designs onlyn = 431
The marks sound alike or are phonetic equivalents49.7%46.6%16.2%
The words outweigh the design; the design doesn’t avoid confusion21.5%65.1%12.8%
The goods or services are identical52.9%49.0%58.7%
The goods are only related, not identical30.2%30.8%23.4%
Goods relation not clear from the wording16.9%20.2%17.9%
A weak or crowded mark is mentioned3.7%3.8%6.7%

Phrase counts in the 2(d) section only (from its heading to the next issue), with case-law citation sentences removed; for sound, the boilerplate sentence listing all four factors is removed too. A random check of 20 sound matches found 18 where the examiner applied it. Appearance, meaning and commercial impression are left out: boilerplate made them unreliable.

Other refusals and requirements in the same letters

Sampled applications with the full letter text, by the issues their letters raise (estimated from headings).
MeasureNames (standard characters)n = 9,037Logos with wordsn = 9,025Designs onlyn = 8,959
Likelihood of confusion, 2(d)26.8%24.7%5.3%
Merely descriptive, 2(e)(1)12.4%1.8%0.8%
Primarily a surname, 2(e)(4)1.8%0.6%0.1%
Geographic, 2(e)(2) or (3)1.9%0.5%0.2%
Specimen refused or required24.4%21.9%21.3%
Failure to function or ornamental3.2%2.8%7.3%
Identification of goods to amend22.9%23.4%30.3%
Disclaimer required or advised24.4%35.0%3.7%
Two or more refusal grounds9.8%6.1%3.9%

From short heading lines that name the issue. Refusal grounds counted for “two or more”: 2(d), descriptive, surname, geographic, specimen, failure to function. Of the applications refused under 2(d), 22.7% of names, 19.5% of logos with words and 21.4% of designs also had another refusal ground. The 2(d) share here is of applications with the full text, so it differs from the first table.

Method and limits

  • Sample: a random sample of U.S. applications whose first examination was a non-final office action issued 2019–2022, drawn in equal parts: 10,536 names (standard characters), 10,536 logos with words, 10,537 designs.
  • Letters: 33,246 office-action texts from the USPTO’s TSDR records; 5,629 of them were 2(d) refusals, dated July 2, 2019 to November 29, 2022. Cited marks were looked up in the USPTO’s daily trademark data.
  • Refused under 2(d): 4,980 applications (2,371 names, 2,160 logos with words, 449 designs).
  • Not every application: only those that received a first office action. Shares describe refused applications in this sample, not the whole register.
  • Name relations are rule-based on normalised text, and “related classes” follow the USPTO’s coordinated class table; older registrations can carry pre-2002 class numbers.
  • Counts of public records. Not legal advice, and not a prediction for any one mark.

How to cite

LogoVerify (2026). What blocks a trademark at the USPTO: an analysis of 4,980 likelihood-of-confusion refusals. https://logoverify.com/research/uspto-trademark-refusals

Data (CSV)USPTO, grounds for refusalHow a name is checked

Check your name and logo the way an examiner does

Free preview, no signup, no card. Full check $8.99 one-time, when you want the answer.

Check a name or logo free