How Does AI Assess Similarity Between Two Trademarks?

When two trademarks “look similar,” the similarity may come from different sources: spelling, pronunciation, structure, meaning, a shared dominant element, or visual composition. AI can help by separating the problem into multiple signals instead of relying on exact character matching.

A similarity score is not a decision that infringement exists or an application will be refused. The legal meaning depends on the marks as a whole, the strength of shared elements and the relationship between the goods or services.

1. Text structure and spelling

A system can compare length, character or syllable order, prefixes, suffixes, dominant components and spelling variants. This is close to fuzzy text search but is most useful when combined with other signals.

2. Pronunciation

Two marks can be spelled differently while sounding close. Phonetic analysis can bring those variants into the candidate set, including reasonable transliterations and pronunciations relevant to the Vietnamese market.

3. Meaning and semantic similarity

Two words may look different but convey a similar concept, translation or association. Semantic similarity can expand the search, but its legal significance depends on context and the distinctiveness of the shared concept.

4. Visual and logo similarity

For figurative and combined marks, image analysis can compare composition, shapes, symbols and overall visual features. A simple color change may not create enough distance where the core device remains highly similar.

5. Goods/services and Nice context

Sign similarity is only half of a conflict assessment. The goods or services may be identical, similar, related or remote. Supplying the actual goods/services and Nice classes makes the screening result more useful.

6. Why a composite score is not a legal verdict

A high score may be driven by a weak descriptive element, while a lower total score can hide a nearly identical distinctive component. Users should open the top references and understand why they are similar instead of relying only on a number.

7. Turning similarity signals into a decision

Use AI to prioritize the top references, then check status, owner, filing or priority dates, goods/services and the strength of shared elements. If one or more references remain material, move to the legal assessment framework or professional review.

Frequently Asked Questions

Does AI use a legal percentage threshold for similarity?

It should not be understood that way. Similarity scores are screening and ranking tools, not fixed statutory thresholds.

Can AI compare logos?

Yes where image-search functionality is available for figurative and combined marks.

Why can a lawyer warn about a mark with a lower AI score?

A lawyer may identify a legally important dominant element or goods/services context that a composite score does not fully express.

Conclusion

AI similarity analysis is most useful when it explains the dimensions of similarity and directs attention to the right records. The user should see not only “how much” but also “where” and “why” the marks may be close.