A trademark name search is the difference between a $350 filing fee and a $50,000 rebrand. You run one before you file a trademark application, and it tells you whether someone else already owns a mark that's too similar to yours. Skip it, and you're filing blind into a system with over 600,000 new applications per year.
This guide covers how trademark name search actually works: how government databases are structured, why searching for trademarks is fundamentally different from searching the web, and how modern APIs have changed what's possible. Whether you're a developer building search into a product or a founder [checking trademark name availability](/blog/learn-how-to-search-trademarks) before you commit, this is the mechanical reality behind "is this name available?"
## Why a Trademark Name Search Matters Before You File
[Filing a trademark application](/blog/2026-06-26-how-to-file-trademark-uspto) at the USPTO costs $250 to $350 per class, depending on the filing basis. Add an attorney and you're looking at $1,500 to $2,500 all-in. That money is gone if your application gets rejected because an existing registration is too similar.
And rejection is common. The USPTO receives over 600,000 trademark applications per year. That's roughly 1,700 new filings every day, each claiming a slice of the same namespace you want. The odds that your name, or something confusingly close to it, already exists are higher than most people expect.
Here's the part that trips people up: "available" is not binary. You won't get a clean yes or no.
Trademark law doesn't ask whether your exact name is taken. It asks whether your mark is [likely to cause confusion](/blog/learn-likelihood-of-confusion) with an existing one. That test considers how similar the marks sound, how similar the marks look, and whether they cover related goods or services.
A company selling AURORA software in Class 9 might coexist with an AURORA clothing brand in Class 25. But AURORA and ARORA in the same class? That's a conflict most examiners would flag. The question isn't "does my exact name exist?" The question is "could a consumer confuse my brand with something already registered?"
A trademark name search is how you answer that question before you spend money on a filing. It won't give you a legal opinion, but it will show you the field. And if the field is crowded, you'll know before you've committed $2,000 and six months of waiting.
*Consult a trademark attorney for legal guidance specific to your situation. A search surfaces potential conflicts, but interpreting them requires legal expertise.*
## How Government Trademark Databases Work
Every trademark office maintains a public database of its filings and registrations. These are the raw source for any trademark name search, and their structure determines what you can find and what you'll miss.
### TESS (1998-2023)
For 25 years, the Trademark Electronic Search System (TESS) was the primary way to search US trademarks. It supported Boolean operators, design codes, and field-specific queries. The USPTO [shut down TESS on November 30, 2023](/blog/2026-04-03-tess-is-gone-how-to-search-trademarks), replacing it with the Trademark Search tool inside TEASi.
### The Post-TESS World
The USPTO's current Trademark Search tool lives within TEASi (Trademark Electronic Application System, international). It searches active and inactive US trademark records with basic text matching and some filtering by status, owner, and filing date. For detailed status information on a specific application, you use TSDR (Trademark Status and Document Retrieval), which provides the full prosecution history.
Both tools are free and authoritative for US marks. But they only cover US marks.
### International Databases
For international coverage, two databases matter most:
**WIPO Global Brand Database** covers marks registered through the Madrid System and pulls records from 70+ national offices. It's the closest thing to a cross-border search tool that exists at the government level. You can search by text, image, or holder name, and filter by designated office, Nice class, and status.
**EUIPO eSearch** covers European Union trademarks (EUTMs) and registered Community designs. If you're filing in Europe, this is the authoritative source.
Other offices maintain their own databases: CIPO for Canada, IPOS for Singapore, IP Australia, and dozens more. Each has its own search interface, its own query syntax, and its own data format.
### The Silo Problem
The fundamental limitation is that every one of these databases is siloed. If you want to search US, EU, and international marks, you need to run three separate searches across three separate tools with three separate interfaces. There's no single government-run system that lets you do a [cross-jurisdiction trademark search](/blog/2026-08-10-tess-alternatives-trademark-search).
For a developer building a clearance tool, this is the core infrastructure problem. The data is public but fragmented. Each office publishes its records in a different format, with different field names, different status codes, and different update schedules. Unifying that data into a single searchable index is where most of the engineering effort in modern trademark search goes.
<RadarChart data={[ { "Capability": "US Coverage", "TESS": 90, "TEASi": 90, "API": 95 }, { "Capability": "International Coverage", "TESS": 0, "TEASi": 0, "API": 95 }, { "Capability": "Search Sophistication", "TESS": 70, "TEASi": 40, "API": 95 }, { "Capability": "Programmatic Access", "TESS": 0, "TEASi": 0, "API": 100 }, { "Capability": "Filtering", "TESS": 60, "TEASi": 50, "API": 90 } ]} categoryKey="Capability" valueKeys={["TESS","TEASi","API"]} labels={{"TESS":"TESS (retired)","TEASi":"TEASi (current)","API":"API-based"}} title="Search Tool Capabilities: TESS vs TEASi vs API" height={400} />
## What Makes Trademark Search Different from Web Search
If you type "AURORA" into Google, you'll get results about the Northern Lights, a Disney princess, and a city in Colorado. Google optimizes for relevance: what did you probably mean?
Trademark search optimizes for legal risk: what could be confused with your mark? That's a fundamentally different problem, and it's why standard text search doesn't work.
### Phonetic Matching
The biggest gap in amateur trademark searches is phonetic similarity. SIGNA, CYGNA, and SYNNA are spelled differently but sound nearly identical. An exact-text search for "SIGNA" won't return the other two. A trademark examiner, however, will catch all three.
[Phonetic matching algorithms](/blog/2026-08-14-how-phonetic-trademark-matching-works) encode words by their pronunciation rather than their spelling. They reduce marks to phonetic fingerprints and compare those fingerprints. This is where most self-service searches fail: they search for spelling matches in a system that evaluates sound.
### Visual Similarity
Design marks (logos, stylized text) add another dimension. Two marks might look similar even if the underlying words are different. Evaluating visual similarity requires comparing structural elements: shape, color arrangement, dominant features. Government databases support design code searches (the Vienna Classification system), but visual similarity analysis at scale is still an emerging capability.
### Nice Classification Context
Trademark rights are scoped to specific categories of goods and services. The [Nice classification system](/blog/learn-trademark-classes) divides all commerce into 45 classes: Classes 1 through 34 cover goods, Classes 35 through 45 cover services. Class 9 covers software and electronics. Class 25 covers clothing. Class 43 covers restaurant services.
Two identical marks can legally coexist if they're registered in unrelated classes. DELTA airlines (Class 39, transportation) coexists with DELTA faucets (Class 11, plumbing fixtures).
Even within the same class, marks can sometimes coexist if their goods and services descriptions don't actually overlap. Class 9 covers everything from downloadable software to fire extinguishers, so two Class 9 marks with non-overlapping goods descriptions may not conflict. But two similar marks in the same class with overlapping goods will trigger a conflict.
This means a trademark name search isn't just about the mark itself. It's about the mark in context: what goods or services are you selling, and what classes do existing marks occupy?
A search without class filtering will bury you in false positives. A search with class filtering might miss a related-class conflict. Getting this right requires understanding both the classification system and the specific scope of your goods.
### Why Google-Style Search Fails
Web search engines rank results by relevance, popularity, and recency. Trademark search needs to rank by legal risk. A mark with 3 active registrations in your class is more dangerous than a mark with 300 registrations across unrelated classes. A live registration is more threatening than an abandoned application.
Standard search also can't handle the phonetic and fuzzy matching that trademark law requires. It's optimized for finding what you're looking for. Trademark search is optimized for finding what you're hoping doesn't exist.
## How Modern Trademark Search APIs Work
Government databases give you the data, but one office at a time, through a browser, with no way to integrate results into your own systems. Modern trademark search APIs solve this by aggregating records from multiple offices into a unified index and exposing that index through programmable endpoints.
### The Shift to Programmable Search
Instead of clicking through a government website and manually comparing results, you send a request to an API with your query and filters. You get back structured data you can parse, store, and act on. For developers building clearance tools, e-commerce compliance checks, or brand monitoring systems, this is the difference between a manual process and [an automated one](/blog/2026-05-26-automate-trademark-search).
### Four Search Strategies
Modern trademark APIs typically support multiple search strategies, each designed to catch a different type of similarity:
1. **Exact** matches the mark text character-for-character. Fast and precise, but misses spelling variations. 2. **Phonetic** matches marks that sound similar regardless of spelling. This is the strategy that catches SIGNA/CYGNA/SYNNA. Non-negotiable for any serious trademark search. 3. **Fuzzy** matches marks within a certain edit distance, catching typos and minor spelling variations. AURORA would match AURRA, AUROR, and AURROA. 4. **Prefix** matches marks that start with your search term. Useful for finding families of marks built on the same root.
<BarChart data={[ { "Strategy": "Exact", "Matches": 1 }, { "Strategy": "Prefix", "Matches": 5 }, { "Strategy": "Fuzzy", "Matches": 8 }, { "Strategy": "Phonetic", "Matches": 14 } ]} xKey="Strategy" yKey="Matches" title="Typical Matches Found per Strategy for "AURORA"" height={350} />
You can combine strategies in a single request. A thorough search runs at least exact and phonetic together.
### Relevance Scoring
When you search a database of 147 million records across 200+ offices, you need a way to prioritize results. API-based search tools return relevance scores (typically 0 to 100) that reflect how closely a result matches your query. A score of 95 means near-identical; a score of 60 means worth reviewing but probably not a blocker.
These scores aren't legal opinions. They're triage tools. They let you focus your attention on the 15 results that score above 80 rather than reading through 500.
### Filtering
Raw search results without filtering are overwhelming. API-based tools let you narrow results by:
- **Office** (US, EM, WO, and others using ST.3 country codes) - **Nice class** (restrict to relevant goods/services categories) - **Status** (live registrations only, or include abandoned/expired marks) - **Filing date** (useful for monitoring new conflicts)
Beyond filtering individual results, some APIs offer aggregations: counts of matching marks grouped by office, Nice class, or status. These summaries turn a raw result list into a clearance report you can act on.
### Unified Data Model
The engineering challenge behind a multi-office search API is data normalization. The USPTO calls it a "registration number." EUIPO calls it an "EUTM number." WIPO uses "international registration number." Status codes, date formats, owner fields, and classification schemes all vary by office.
A well-built API normalizes these differences into a single schema. You search once and get back consistent results regardless of which office holds the registration. One query, one format, one set of fields to parse.
### Example: Searching Across Offices
A trademark name search through an API looks like this. This example uses Signa's API to search for "AURORA" across US and EU marks in Classes 9 and 42:
```bash curl -G "https://api.signa.so/v1/trademarks" \ -H "Authorization: Bearer $SIGNA_API_KEY" \ --data-urlencode "q=AURORA" \ --data-urlencode "strategies=exact,phonetic" \ --data-urlencode "offices=US,EM" \ --data-urlencode "nice_classes=9,42" ```
Or with the TypeScript SDK:
```typescript
const signa = new Signa({ api_key: process.env.SIGNA_API_KEY });
const results = await signa.trademarks.search({ query: 'AURORA', strategies: ['exact', 'phonetic'], filters: { nice_classes: [9, 42], offices: ['US', 'EM'], }, }); ```
Both return structured results with the mark text, owner, status, Nice classes, filing date, and a relevance score. You get back data you can filter, sort, and feed into your own clearance workflow.
## From Search to Clearance: What Comes After Results
A trademark name search gives you a list of potentially conflicting marks. Trademark clearance is the analysis of that list. Search is mechanical. Clearance is judgment.
<MermaidDiagram chart={` flowchart LR A[Name Idea] --> B[Trademark Search] B --> C[Triage Results] C --> D{Conflicts?} D -->|High risk| E[Revise Name] E --> B D -->|Low/medium risk| F[Full Clearance] F --> G[Attorney Review] G --> H[File Application] `} />
### Triaging by Risk
Not every result is a threat. When you review search results, you're sorting them into three buckets:
**High risk.** Same or nearly identical mark, same or related class, live registration. These are potential blockers. If a phonetic match scores above 90 in your target class, you need to take it seriously.
**Medium risk.** Similar mark, adjacent class, or pending application. These require closer examination. Is the mark actually in use? Are the goods/services descriptions overlapping, or just in the same class header?
**Low risk.** Weak similarity, unrelated class, or abandoned/expired status. Note them for completeness, but they probably won't block your filing.
### Owner Context
Who owns the conflicting mark matters. A mark held by an individual who filed once and never renewed is different from a mark held by a Fortune 500 company with a 200-mark portfolio and a history of sending cease-and-desist letters. Checking the owner's other registrations and enforcement patterns gives you context that a raw search result can't.
### The Human Layer
No API replaces legal interpretation. An API can tell you that ARORA is registered in Class 9 by a company in California. It can't tell you whether your AURORA mark for a different type of software in the same class would survive an opposition. That judgment requires a trademark attorney who understands likelihood of confusion doctrine, the strength of the existing mark, and the specifics of your goods description.
The best workflow combines automated search with human analysis. The API handles the retrieval and initial scoring. The attorney handles the interpretation and risk assessment. For a complete breakdown of that workflow, see the [trademark clearance search guide](/blog/2026-06-01-trademark-clearance-search-guide).
*Consult a trademark attorney for legal guidance specific to your situation.*
## What to Look for in a Trademark Name Search Tool
Whether you're evaluating a [free trademark search tool](/blog/2026-04-24-trademark-search-free-guide) or an API for a production application, these are the criteria that matter:
**Coverage.** How many offices does it search? How many total records? A tool that only covers the USPTO won't catch a EUIPO registration that could block your European expansion. Look for multi-office coverage with a clear list of supported jurisdictions.
**Search strategies.** Exact match is the baseline. Phonetic matching is non-negotiable. If a tool can't find CYGNA when you search for SIGNA, it's going to miss the conflicts that matter most. Fuzzy and prefix matching add additional layers of protection.
**Data freshness.** How quickly do new filings appear in the index? A trademark filed yesterday might be the one that conflicts with yours. If the tool updates weekly or monthly, you're searching stale data. Daily or near-real-time updates are the standard for serious use.
**Programmable access.** Can you integrate it into your own systems? If you're building a product that needs trademark search, a browser-based tool won't work. You need an API with structured responses, pagination, and authentication. The shift from GUI-only tools to [API-based alternatives](/blog/2026-08-10-tess-alternatives-trademark-search) is what makes automated clearance workflows possible.
**Filtering and bulk capability.** Can you filter by class, status, office, and date? Can you run hundreds of searches programmatically? If you're clearing a portfolio of names or monitoring across jurisdictions, one-at-a-time manual search doesn't scale.
## Start Your Trademark Name Search
The tools have evolved from TESS's Boolean queries to APIs that run phonetic matching across 200+ offices in a single request. The fundamentals haven't changed: you're still looking for marks that could be confused with yours, scoped to the classes and jurisdictions that matter.
What has changed is access. You no longer need to search each office individually, learn each database's query syntax, or manually compare results across jurisdictions. Programmable search puts that capability into a single call.
Signa's API indexes 147M+ trademark records across 200+ offices with exact, phonetic, fuzzy, and prefix search strategies. Try a [trademark search with Signa's free API](https://signa.so) (1,000 searches/month on the free plan) to see how API-based search compares to the government databases you've been using.