There is no trademark similarity percentage defined in Turkish law: Industrial Property Law No. 6769 (SMK) and the practice of TÜRKPATENT (Turkish Patent and Trademark Office) decide whether similarity blocks registration not by a percentage but by comparing the signs visually, aurally and conceptually, together with the similarity of the goods and services. There is no percentage threshold for similarity that prevents registration; the score a search tool displays is only a first-screen indicator.
Below you will find how the scores in search tools are calculated, which legal criteria they cannot see, how the difference between an ex officio refusal by the Office and a refusal following an opposition affects what a score means, and, through fictitious examples, how the same score can lead to very different outcomes.
Is There a Trademark Similarity Percentage in the Law?
No. The SMK uses two separate similarity tests, and both are qualitative. Article 5(1)(ç) makes it an absolute ground for refusal if a sign is "identical or indistinguishably similar" to a mark registered or applied for earlier in respect of identical or same-type goods and services. Article 6(1) provides that an application is refused on opposition where there is a "likelihood of confusion", including a likelihood of association by the public.
The Office's practice does not use percentages either. TÜRKPATENT's Trademark Examination Guidelines (Marka İnceleme Kılavuzu) express the outcome of the sign comparison in a likelihood-of-confusion examination on five levels:
- Dissimilar
- Low degree of similarity
- Average degree of similarity
- High degree of similarity
- Identity
These levels are weighed together with the similarity of goods and services and other factors, and none of them corresponds to a percentage. That is why the thresholds circulating online, of the "anything below X percent is safe" kind, are assumptions with no legal basis.
Part of the reason a percentage does not work lies in the nature of the comparison. According to the Guidelines, a likelihood-of-confusion examination assumes that consumers have no chance to place two marks side by side; the comparison is made on the basis of the impression left in memory. A ratio calculated letter by letter, by contrast, is exactly the product of a side-by-side comparison. Moreover, even when sign similarity is low, identical goods and services can tip the result toward risk.
What Does a Similarity Score in a Search Tool Measure?
A similarity score is a numerical indicator of closeness that software produces by comparing two signs with a particular algorithm; it is not a legal decision. What it measures depends on the algorithm, and three methods are in common use.
Letter-string distance
The Levenshtein distance is the minimum number of single-character insertions, deletions or substitutions needed to turn one word into another. According to its help page, the fuzzy search in TMview, run by the EUIPO, uses this algorithm.
One common way of turning the distance into a percentage is to divide it by the number of letters in the longer word and subtract the result from one. Take a fictitious example: VELORA and VELURA differ by a single letter; the distance is 1, the length is 6, and the score comes out at about 83%. That figure describes nothing but letter overlap.
The method has two blind spots. First, it penalizes additions heavily: when a descriptive word is added to the end of a mark, the distance grows and the score falls, even though the dominant element has not changed. Second, it counts every letter change equally: writing "K" for a "C" that sounds the same affects the score just as much as a letter change that alters the pronunciation completely.
Phonetic algorithms
Phonetic algorithms encode words according to how they sound and match words that receive the same code; Soundex and Metaphone are well-known members of this family. They were designed around English pronunciation rules. Turkish rules, such as "c" producing the sound of the English "j" or "ğ" lengthening the vowel before it, are not automatically reflected in these encodings. For a Turkish mark, it therefore matters which language's rules produced the phonetic score.
Because Turkish is read as it is written, the Guidelines accept that for Turkish words aural similarity runs in parallel with visual similarity. With foreign words, however, spelling and pronunciation can diverge, which can produce two marks with a low letter score but highly similar pronunciation.
Visual-similarity scores
The visual-similarity algorithms used in logo and device searches rank two images by shape, color or the concept they detect. We explain how to use that ranking in our guide to trademark logo search.
Not every tool produces a score at all. Webx's free online trademark search tool shows no percentage; it lists the records found with their application number, mark name, applicant, status, classes and image, so that you can make the assessment on the basis of that data.
What Can a Similarity Score Not Measure?
A score compares the surface of two signs; the law assesses signs in context. The elements software generally cannot see are:
- The dominant element: which part of the mark sticks in the memory and which word consumers use to refer to it.
- Descriptive or weak add-ons: additions such as "GROUP", "TEKNO" or "MARKET" change the letter count but add no distinctiveness.
- The beginning of the word: the Guidelines accept that consumer attention concentrates on the beginning of a word, whereas most algorithms give every letter equal weight.
- Conceptual similarity: marks that mean the same thing in different languages may not resemble each other at all at the letter level.
- The relationship between goods and services: a score does not know the fields in which the two marks are used.
- Contextual factors: the level of attention of consumers and the distinctiveness or reputation of the earlier mark.
The Guidelines also state that two words containing the same number of letters does not by itself lead to a finding of visual similarity; what really matters is how many letters they share and whether those letters appear in the same order. We explain the legal framework of the assessment in detail in our article on the likelihood of confusion.
What Do Examiners and Courts Look At?
Examiners and courts first identify the dominant and distinctive elements of the signs, then compare the signs in three dimensions and weigh the result together with the similarity of the goods and services. Roughly, the order of examination is:
- Visual similarity: the order and position of the letters, the length of the word and any figurative elements.
- Aural similarity: the number and order of syllables, stress and shared syllables. According to the Guidelines, an unstressed syllable added to the middle or end of a word generally does not change its aural character. We cover how to search for pronunciation variants in our article on phonetic similarity.
- Conceptual similarity: the meaning the signs evoke and any translations. The meaning of a foreign word is taken into account only if the average consumer would know it.
- Similarity of goods and services: their nature, purpose and distribution channels, and whether they compete with or complement each other. Under Article 11(4) SMK, goods being in the same class does not create a presumption of similarity, and goods being in different classes does not create a presumption of dissimilarity.
Alongside these four dimensions, the distinctiveness of the earlier mark is also weighed. According to the Guidelines, words that have no connection with the goods and services and whose meaning is unknown have relatively high distinctiveness, while words that carry a particular meaning for consumers may have lower distinctiveness. In practice, this means that choosing a name close to an invented fantasy word is riskier than choosing a name close to a word that is common in the sector.
What Is the Difference Between an Ex Officio Refusal and a Refusal on Opposition?
Under Article 5(1)(ç) the Office acts on its own initiative and applies a high threshold; under Article 6(1) an opposition by the owner of the earlier right is required, and the level of similarity required is lower. This distinction means the same score can carry a different meaning in each channel:
| Criterion | Ex officio refusal (Art. 5(1)(ç) SMK) | Refusal on opposition (Art. 6(1) SMK) |
|---|---|---|
| Who starts it? | TÜRKPATENT, on its own initiative | The earlier right holder, within 2 months of publication in the Official Trademark Bulletin (Art. 18) |
| Similarity required | Identical or indistinguishably similar | Likelihood of confusion, including likelihood of association |
| Goods and services condition | Identical or of the same type | Identical or similar |
| Similarity in pronunciation only | According to the Guidelines, generally not enough without a high degree of visual similarity | Weighed as a separate dimension together with the visual and conceptual ones |
| Ways to overcome it | A notarized letter of consent (Art. 5(3)) | Settlement (Art. 19(4)); a request for proof of use where the earlier mark is more than five years old (Art. 19(2)) |
A clear example of why the threshold cannot be expressed as a percentage is again found in the Guidelines: long words that differ only by "x" and "ks" are, as a rule, considered indistinguishably similar, while in short words the same difference is accepted as enough to set the signs apart. The same letter change leads to a different result depending on the length of the word.
The practical consequence is this: an application that only sounds like an earlier mark but is written in a visually different way can pass ex officio examination, be published in the Bulletin, and then face an opposition. The Office publishing your application does not mean that no similar mark exists. We compare the two channels in detail in our article on whether a similar trademark can cause refusal, and we explain the document that overcomes an ex officio refusal in our article on trademark registration with a letter of consent.
Same Score, Different Outcome: An Example Table
The same letter score can lead to entirely different outcomes depending on the relationship between the goods and services and the structure of the mark. All the mark names in the table are fictitious; the scores were calculated with the Levenshtein formula described above. The "likely assessment" column shows the general tendency and is not a definitive result.
| Earlier mark (example) | New application (example) | Letter score | Goods/services relationship | Likely assessment |
|---|---|---|---|---|
| VELORA — coffee (Class 30) | VELURA — coffee (Class 30) | 83% | Identical goods | A single-letter difference in the middle of a meaningless word: even an ex officio refusal is a risk, and the risk on opposition is high. |
| VELORA — coffee (Class 30) | VELURA — café services (Class 43) | 83% | Different class, related field | An ex officio refusal may not be expected; if opposed, likelihood of confusion will be seriously debated. |
| VELORA — coffee (Class 30) | VELURA — construction machinery (Class 7) | 83% | Unrelated | Likelihood of confusion is low; if the earlier mark has a reputation, Art. 6(5) is assessed separately. |
| KALVIRA — software (Class 9) | KALVIRA GROUP — software (Class 9) | 54% | Identical goods | The score looks low, but "GROUP" adds no distinctiveness; the dominant element is the same, so the risk is high. |
| ZEPO — cosmetics (Class 3) | ZEPA — cosmetics (Class 3) | 75% | Identical goods | In short words a single-letter difference changes perception more easily; the outcome depends on pronunciation and presentation. |
| SARI ZÜRAFA — toys (Class 28) | YELLOW GIRAFFE — toys (Class 28) | 29% | Identical goods | Low letter overlap, identical meaning ("sarı zürafa" is Turkish for "yellow giraffe"). If average consumers know what the English phrase means, conceptual similarity will be argued on opposition. |
The first three rows show how the same score changes with the overlap of goods and services; the last three show how the score and the real risk can part ways.
How Should You Read a Similarity Score?
Treat the score as raw data that has to be translated into legal criteria. The following checklist makes that translation systematic:
- Find out which method produced the score: letter distance, a phonetic code or a visual algorithm?
- Isolate the dominant element; remove weak add-ons such as "GROUP", "TEKNO" or "MARKET" and run the comparison again.
- Look at the beginning of the words: is the difference at the start, in the middle or at the end?
- Read both marks aloud; do they have the same number of syllables and the same stress?
- Compare their meanings and any translations.
- Put the goods and services side by side: are they identical, of the same type, related or unrelated?
- Check the status of the record found: is it registered, still at the application stage, or has its term expired?
- Express the result not as a percentage but on the scale of dissimilar, low, average, high, identical.
Misreading a score is only one of the common mistakes made during a search; we list the others in our article on trademark search mistakes.
What Should You Do If the Score Is High or Low?
A high score is an alarm, and a low score is no guarantee. With a high score, first measure the real risk using the checklist above. If the risk is confirmed, options include adding a distinctive element to the name, removing the conflicting goods and services from the specification, or obtaining consent from the owner of the earlier mark.
With a low score, check conceptual similarity and the descriptive add-on trap in particular. Remember that no similarity score shows the absolute grounds for refusal that stem from your own mark; you will find issues of descriptiveness and lack of distinctiveness in our article on why trademark applications are refused.
A good professional search gives its result as a reasoned risk level rather than a percentage. We describe what such a report should contain in our article on the trademark search report.
Summary: Key Points About Trademark Similarity Percentages
- Neither the law nor the Office's practice sets a percentage threshold for similarity that blocks registration.
- Tool scores measure overlap in letters, sound codes or images; they do not measure the dominant element, meaning or the relationship between goods and services.
- The Office refuses ex officio only marks that are "indistinguishably similar"; lower degrees of similarity are assessed on opposition.
- The same score can mean high risk or low risk depending on the overlap of goods and services.
- Read any score together with a visual, aural and conceptual comparison and a comparison of goods and services.
Would you like to see your similarity risk with reasons instead of a percentage?
Send us your mark, your logo if you have one, and a description of your business through our contact page. Our team, which works with trademark attorneys authorized before TÜRKPATENT, will assess the records found visually, aurally and conceptually, together with the overlap of classes, and report the risk with its level and the reasons behind it.
