What is Search Endpoint?

API & Technical5 min readUpdated Mar 25, 2026

An API endpoint that accepts trademark search queries and returns matching results from across multiple jurisdictions with relevance scoring.

A search endpoint is an API endpoint that accepts trademark search queries and returns matching results from across the trademark database. It is the most fundamental and heavily used endpoint in any trademark API, serving as the primary interface for discovering existing marks that may conflict with a proposed brand name, identifying marks owned by a specific entity, or exploring the trademark landscape in a particular product category or jurisdiction.

Search endpoints support multiple search modalities to address different use cases. Text search finds marks that match or are similar to a given text string, using exact matching, fuzzy matching, or wildcard patterns. Phonetic search identifies marks that sound similar regardless of spelling differences (for example, finding "Klear" when searching for "Clear"). Image search compares visual elements of device marks and logos against a submitted image. Structured search allows filtering by specific fields such as owner name, Nice Classification, jurisdiction, filing date range, and status.

The quality of a search endpoint is measured by several factors. Recall measures the proportion of relevant results that are returned, meaning that the search does not miss important matches. Precision measures the proportion of returned results that are actually relevant, meaning that the search does not return excessive false positives. Response time measures how quickly results are returned. And coverage measures how many jurisdictions and records are included in the search.

Search endpoints typically support a range of parameters that allow clients to configure the search behavior. Common parameters include the search term (text or image), target jurisdictions (which offices to search), Nice Classification codes (which product categories to include), similarity threshold (how similar results must be to the query), sort order (relevance, date, alphabetical), and pagination parameters (page size and cursor).

Why It Matters

The search endpoint is the starting point for virtually every trademark workflow. Clearance searches determine whether a proposed brand name is available for use and registration. Monitoring searches identify potentially conflicting marks on an ongoing basis. Portfolio searches help managers understand the scope of their organization's trademark holdings. Competitive intelligence searches reveal competitors' branding strategies and filing patterns.

The quality of search results has direct legal and business consequences. A clearance search that misses a relevant conflict can lead to a trademark application that is refused, opposed, or challenged, wasting time and money and potentially forcing a costly rebrand. A monitoring search that fails to detect a threatening filing can result in missed opposition deadlines. A portfolio search that returns incomplete results can lead to inaccurate reporting and poor strategic decisions.

For these reasons, the search endpoint's algorithms must balance sensitivity and specificity. Too aggressive a similarity threshold returns many results but creates review burden and may generate false alarm fatigue. Too conservative a threshold misses genuine conflicts. The optimal threshold depends on the use case: clearance searches may warrant broader results to err on the side of caution, while monitoring searches may warrant tighter thresholds to focus on the highest-risk conflicts.

The search endpoint must also perform well under load. Trademark searches are often time-sensitive, whether integrated into a real-time user experience or run as part of a batch clearance process. Fast responses keep search from becoming a bottleneck in the workflow.

How Signa Helps

Signa's search endpoint (GET or POST /v1/trademarks) searches every office Signa indexes in one call. Image search is planned and not available yet.

The q parameter ranks marks that resemble a name through similarity channels: identical, fuzzy (one or two changed letters), embedded (the name as a word or phrase in a longer mark), phonetic (sound-alikes, off by default) and lookalike (look-alike characters). The similarity parameter chooses which channels run. Each result carries a relevance_score from 0 to 100 that orders the results and is not a likelihood of confusion, plus a match object whose tier (identical, near_identical, similar or related) and via say how close the hit is and which channels found it.

Comparing an uploaded logo against the device marks Signa indexes is planned, not available yet. Today, figurative marks are found through their Vienna design codes.

By default the search spans every office in Signa's database, and filters such as jurisdictions, offices, nice_classes, status_primary and filing date ranges narrow it. Text filters (mark_text_is, mark_text_starts_with, mark_text_contains and others) keep or exclude marks by their exact text. Results come back in one normalized format with consistent snake_case fields regardless of the source office.

Typical searches return in under a second: in an August 2026 benchmark of 471 searches against the production API, the median was 208 ms and the 95th percentile 635 ms, timed server-side.

The endpoint pages through results with a cursor. Clients set limit from 1 to 100 results per page, read has_more and pagination.cursor for the next page, and can request include_total for a count. A search with q reaches at most 10,000 rows, so narrow the query with filters for larger result sets.

Example

A brand consultancy uses Signa's search endpoint in the screening phase of its naming projects. For each project it evaluates a few dozen candidate names and wants a first look at potential conflicts before any attorney review.

For each candidate, the consultancy calls the search endpoint with the candidate as q, the relevant Nice classes and its target jurisdictions (the US, EU, UK, Canada and Australia), with the phonetic channel turned on to catch sound-alikes such as "NovaTech" and "NovaTeq" or "SunRise" and "Sonrise."

The consultancy's own triage step reads the match.tier, status and classes of each hit and sorts candidates into those that look open, those that need a closer look, and those with obvious conflicts. The shortlist then goes to a trademark attorney for a full clearance opinion, because a search result is a list of potential conflicts, not a legal conclusion.