Trademark Clearance

Trademark clearance is the process of checking whether a proposed mark conflicts with existing registrations before filing an application. The trademark_clearance tool automates this by searching across jurisdictions, scoring similarity, and producing a risk assessment.

What clearance checks

A clearance search identifies:

Exact matches — identical marks in the same or related classes
Phonetic conflicts — marks that sound similar (e.g., “FONE” vs “PHONE”)
Visual conflicts — marks with similar spelling (e.g., “GOOGEL” vs “GOOGLE”)
Conceptual conflicts — marks scoring above the similarity threshold

The tool also checks class overlap and relatedness, so a conflict in class 9 (software) is flagged as related to class 42 (technology services) even without a direct class match.

Running a clearance search

Provide the proposed mark name, target Nice classes, and jurisdictions:

{
  "proposedMark": "SKYFORGE",
  "niceClasses": [9, 42],
  "jurisdictions": ["US", "EU"],
  "includePhonetic": true,
  "includeSimilarSpelling": true,
  "minSimilarityScore": 0.6
}

Response:

{
  "proposedMark": "SKYFORGE",
  "riskLevel": "medium",
  "totalConflicts": 3,
  "conflicts": [
    {
      "trademark": {
        "id": "US-90123456",
        "name": "SKYFORGE",
        "jurisdiction": "US",
        "status": "registered",
        "niceClasses": [28]
      },
      "conflictType": "exact",
      "similarityScore": 1.0,
      "overlappingClasses": [],
      "riskAssessment": "Exact match found: \"SKYFORGE\" (US). This mark is currently registered. No overlapping classes, but exact name match increases risk."
    },
    {
      "trademark": {
        "id": "EU-018999888",
        "name": "SKYFORT",
        "jurisdiction": "EU",
        "status": "registered",
        "niceClasses": [9, 42]
      },
      "conflictType": "visual",
      "similarityScore": 0.82,
      "overlappingClasses": [9, 42],
      "riskAssessment": "\"SKYFORT\" is visually similar (82% similarity). Overlapping classes: 9, 42. Active registration presents high opposition risk."
    }
  ],
  "recommendations": [
    "Moderate risk detected. Consider consulting with a trademark attorney to assess the strength of potential conflicts.",
    "Evaluate whether the overlapping classes involve closely related goods/services.",
    "Visually similar marks exist. Consider marks with more distinct spelling."
  ],
  "searchedJurisdictions": ["US", "EU"],
  "executionTimeMs": 3450
}

Similarity scoring

IPKit uses an ensemble of four algorithms to score how similar two marks are:

Algorithm	What it measures
Jaro-Winkler	Character-level edit distance, weighted toward the beginning of the string
N-gram overlap	Shared character sequences (bigrams and trigrams)
Damerau-Levenshtein	Minimum edits including transpositions (GOOGEL -> GOOGLE is 1 edit)
Phonetic matching	How the marks sound when spoken aloud

The scores are combined into a single similarityScore between 0 and 1.

Jurisdiction-aware phonetic scoring

The phonetic component adapts to the jurisdiction of each existing mark:

Jurisdiction	Algorithm	Why
EU	Cologne Phonetic	Optimized for German/Continental consonant patterns
US, GB, CA	NYSIIS	Optimized for English name pronunciation
AU, NZ	Caverphone	Regional pronunciation patterns
JP, CN	Disabled	Not applicable to CJK character systems
WIPO (default)	Double Metaphone	General-purpose phonetic encoding

This means that when checking “SKYFORGE” against an EU mark “SKAIFORSCH”, the Cologne algorithm handles the comparison, while a US mark uses NYSIIS.

Famous marks pre-check

Before analyzing API results, the clearance tool runs a local check against approximately 200 well-known marks (APPLE, GOOGLE, NIKE, AMAZON, etc.). This catches misspellings that API substring search would miss.

For example, searching for “GOOGEL” via the EUIPO API uses substring matching (verbalElement=="*GOOGEL*"), which will not find “GOOGLE” since “GOOGEL” is not a substring of “GOOGLE”. The famous marks pre-check catches this by running similarity scoring locally.

Famous marks receive special treatment in risk scoring because they have broad protection across all classes and aggressive enforcement programs.

Truncated stem search

When the primary search returns fewer than 10 results and the mark is longer than 4 characters, the tool automatically runs a secondary search using the first N-2 characters of the mark. For example, “GOOGEL” triggers a stem search for “GOOG”, which does find “GOOGLE” via substring matching.

Results from the stem search are deduplicated by ID before conflict analysis.

Risk levels

The overall risk level uses a multi-factor scoring model inspired by DuPont factors:

Level	Score	Meaning
low	Below threshold	Few or no conflicts; safe to proceed with standard due diligence
medium	Above medium threshold	Some conflicts detected; consult an attorney to assess strength
high	Above high threshold	Significant conflicts; modifications to the mark are recommended
critical	Above critical threshold	Near-certain opposition or refusal; choose a different mark

The score is the weighted combination of five factors:

Similarity — how close the marks are (ensemble score)
Class relatedness — how related the goods/services are
Conflict type — exact matches weigh more than phonetic matches
Registration status — registered marks are higher risk than pending
Fame — famous marks get a risk boost across all classes

Interpreting results

Low risk

No blocking conflicts found. This does not guarantee registration — it means no obvious red flags in the jurisdictions searched. Proceed with a formal application, but consider common law searches and domain availability checks.

Medium risk

Conflicts exist but may not be blocking. Common scenarios:

Similar marks in related but not identical classes
Phonetic matches where the spelling is distinct
Pending applications that may not proceed to registration

An IP attorney can evaluate whether these conflicts present real opposition risk.

High or critical risk

Strong conflicts that likely prevent registration. When you see this:

Consider modifying the mark (different spelling, added word, different root)
Use distinctiveness_hints to evaluate alternative marks
Focus on jurisdictions where conflicts are absent

Optional: G&S validation

You can include goodsServicesTerms in the clearance request to simultaneously validate your proposed goods and services specification against the EUIPO Harmonised Database:

{
  "proposedMark": "SKYFORGE",
  "niceClasses": [9, 42],
  "jurisdictions": ["EU"],
  "goodsServicesTerms": [
    {
      "classNumber": 9,
      "terms": ["downloadable computer software for project management"]
    },
    {
      "classNumber": 42,
      "terms": ["software as a service (SaaS) featuring project management"]
    }
  ]
}

The response includes a gsValidation section indicating which terms are harmonized (pre-approved) and which have errors or warnings.

Compact mode

For streaming UIs or large conflict sets, set compact: true to receive only the top 5 conflicts with a totalConflicts count:

{
  "proposedMark": "TECH",
  "niceClasses": [9],
  "jurisdictions": ["ALL"],
  "compact": true
}

Limitations

API-level search is substring-based, not true fuzzy. The EUIPO API matches verbalElement=="*QUERY*" — it cannot find marks where the query is a misspelling rather than a substring. The famous marks pre-check and truncated stem search mitigate this, but some edge cases may be missed.
Single-character and two-character marks have limited search coverage. Manual review of the trademark register is recommended.
Common law marks (unregistered marks with rights through use) are not covered. Only marks in official registry databases are searched.
Image/logo similarity is not supported. The tool analyzes word marks only.

trademark_search — direct search without conflict analysis
distinctiveness_hints — analyze mark strength before clearance
suggest_nice_classes — identify classes from a business description
validate_gs_terms — check G&S terms against the Harmonised Database