Denomination Similarity Engine

Denomination Similarity Engine: how similar is it to your trademark, with a number and a reason

A score from 0 to 100 calculated with six linguistic signals, without generative AI and without randomness: the same pair of names always yields the same result.

“Does it look similar enough for me to worry?” That question, which usually costs a consultation, is answered by aiBrand with a figure and the phrase that explains it.

When a new denomination appears, aiBrand compares it with your registered trademark and with the variants you have registered. It does not look for a literal match: it calculates six distinct signals and retains the highest one, because it only takes one to spike for the risk to exist.

The six are the identity of the name once normalized (without accents, without signs, without the final “S.L.” and with visually deceptive characters translated to their actual letters); phonetic equality in Spanish, with its own rules for b/v, seseo c-z-s, c/k/qu, ll/y, silent h and neutralized plural —it is not the English Soundex, which does not work with “Zafira” and “Safira”—; root containment, when one name fits entirely within the other with at least four characters; Jaro-Winkler, which weighs the first letters more because that is where a trademark is read and remembered; normalized edit distance, which detects changed, added, or lost letters; and dominant element, which in multi-word trademarks compares the word that truly distinguishes and discards the filler (“group,” “premium,” “shop,” “online,” “solutions,” “digital,” and a hundred more).

The result is deterministic: there is no language model in the calculation, it does not consume a single credit, and it is reproducible. Each alert retains the version of the engine used for the calculation, so it is always possible to know with what rules that decision was made. And there is one rule that supersedes all: aiBrand never alerts you about requests from your own proprietor or your own file/application number.

The number does not travel alone. The Nice classes multiplier is applied to it, the traffic light color is decided, and the reason is drafted in plain Spanish: “sounds the same as your trademark in Spanish,” “one denomination fully contains the other,” “they share the dominant element.” If the engine cannot explain why it alerts, there is no alert: an alert without explanation is noise.

Benefits of Denomination Similarity Engine

A number and a phrase, not a suspicion

the score from 0 to 100 always comes with the reason written in plain Spanish, which allows you to decide in ten seconds if it affects you.

Phonetics designed for Spanish

b/v, seseo, c/k/qu, ll/y, silent h, and plurals. The pairs that truly confuse a Spanish-speaking customer are captured here, not in an English dictionary.

Six signals, not a single formula

the score is the maximum of all of them, because identity, auditory similarity, and altered letters are distinct risks, and just one is enough to pose a risk.

Deterministic and reproducible

no AI in the calculation, no results that change from one day to the next, and no credit consumption. Each alert retains the version of the engine with which it was calculated.

Does not alert you about yourself

applications from your own owner or your own file are discarded before scoring, so the inbox only brings external matters.

Filler does not count

in trademarks of multiple words, the dominant element is compared, and generics ('group', 'shop', 'online', 'solutions') are ignored, which is exactly the criterion used to assess similarity in practice.

Elvira Santamaría

Measure similarity before a confused customer does

Register your trademark and its variants, and let the engine compare for you. Monitoring does not consume credits.

Similar Trademark Comparator: Denomination Similarity Score 0-100 | aiBrand