Trademark Search Strategy: Beyond TESS

On November 30, 2023, the USPTO retired TESS -- the Trademark Electronic Search System that had served as the foundation of countless trademark search strategies for over two decades. The system that launched in February 2000 was replaced by a new cloud-based Trademark Search tool at tmsearch.uspto.gov. The transition was not optional: in the USPTO's words, TESS "relied on a proprietary search language that is no longer supported."

The new system is better in many ways. But free government tools, no matter how modern, have structural limitations that no interface redesign can fix. If your trademark search strategy begins and ends with the USPTO's free database, you are working with incomplete information.

This article lays out a complete trademark search strategy: what the new USPTO system can and cannot do, the three types of searches every brand owner should understand, the mistakes that lead to costly refusals, and how modern AI-powered tools are filling the gaps that free databases leave open.

The New USPTO Trademark Search: What Changed and What Did Not

The replacement for TESS brings genuine improvements. The new system offers two modes -- basic mode and expert mode -- plus a Field tag and Search builder option for complex, multiple-term searches. It runs on cloud infrastructure, which the USPTO says provides a more stable search experience. It carries over the search concepts TESS offered, in a new syntax: coordinated class searches, regular expression matching for phonetic and plural patterns, and field tag searches for live/dead status, design codes, disclaimers, and more.

These are meaningful upgrades. But the new system also introduced friction. Unlike TESS, which could display all pertinent trademark information in a single view, the new system sent users to TSDR (Trademark Status and Document Retrieval) to see the full record; a beta feature now shows most of that information when you open a search result. Practitioners transitioning from TESS have reported a significant learning curve, and results can be overwhelming without careful filtering.

More importantly, the fundamental limitations of any free USPTO tool remain unchanged:

• Federal trademarks only. No state registrations, no common law marks, no foreign registrations (international filings appear only when they designate the U.S.), no domain names, no social media handles.
• No AI similarity scoring for word marks. Searching "SUNFUEL" will not flag "SON FUEL," "SUNFUL," or "SUN FUELS" unless you manually construct those queries.
• No automatic phonetic matching. The burden of imagining how your mark might sound like another mark falls entirely on you.
• No risk assessment. The system shows records, not analysis. It cannot tell you whether a conflict is actually problematic.
• No likelihood of confusion analysis. Understanding the DuPont factors that examiners use to assess confusion requires legal expertise the database does not provide.
• No watch for new filings. Every search is a snapshot in time. The day after you search, someone could file a conflicting mark, and the search system would not alert you.

For a detailed comparison between the new USPTO search and modern alternatives, see our TESS vs GleanMark comparison. For a walkthrough of using the free tools effectively within their limitations, see our free trademark search guide.

The bottom line: the new Trademark Search system is a capable research tool for identifying exact or near-exact matches among the records in the USPTO database. It is not a search strategy. A strategy requires understanding what kinds of searches exist, when each is appropriate, and what sources of conflict the federal database cannot see.

The Three Types of Trademark Searches

Not every search serves the same purpose. A sound trademark search strategy uses different levels of depth at different stages of the branding process. There are three distinct types of searches, each with different scope, cost, and purpose.

Knockout (Preliminary) Search

A knockout search is a quick first pass designed to identify obvious showstoppers -- identical or nearly identical marks already registered or pending for similar goods and services. The goal is not to be exhaustive. It is to eliminate clearly unavailable names before investing further.

Knockout searches are typically limited to the USPTO database and can be completed in minutes to hours. They are cost-effective screening tools: if your dream brand name is already registered by someone selling the same type of product, there is no need for a deeper analysis.

Where knockout searches fail is in their simplicity. They catch "SUNFUEL" if someone already has "SUNFUEL," but they will not catch "SON FUEL" or "SUN FUELS" unless you think to search for those variations. They will not find common law marks, state registrations, or domain names that indicate prior use. A knockout search tells you whether to keep investigating -- it does not tell you whether a name is safe.

Comprehensive (Full) Clearance Search

A comprehensive clearance search is the deep dive. It searches across multiple databases: federal trademarks, state registrations, common law sources (business directories, domain registrations, social media profiles, industry marketplaces), and sometimes international filings. It analyzes not just identical marks but similar-sounding, similar-looking, and similar-meaning marks across related goods and services.

This is where the real analysis happens. A comprehensive search considers the factors that trademark examiners and courts actually use to evaluate conflicts: the similarity of the marks, the relatedness of the goods and services, the channels of trade, the strength of the prior mark, and the sophistication of the relevant consumers. For a full explanation of these factors, see our guide to the DuPont factors.

A comprehensive clearance search is the cost of due diligence -- and it is a fraction of what you will spend if a conflict surfaces after you have filed, built a brand, and gone to market.

Trademark Watch and Monitoring

The third type of search is not a search at all in the traditional sense -- it is ongoing surveillance. A trademark watch service monitors new filings and publications for marks that could conflict with yours. This is critical for two reasons.

First, new applications are filed constantly. Applicants filed 767,138 new trademark application classes with the USPTO in fiscal year 2024 alone (the USPTO counts filings by class, and one application can cover several classes). Any one of those could conflict with your mark, and you will not know unless you are watching.

Second, the opposition window is narrow. After a mark is approved by the examining attorney, it is published in the Trademark Official Gazette for a 30-day opposition period. If you do not file a notice of opposition during that window (or request an extension), the mark proceeds to registration (or, for an intent-to-use application, to a notice of allowance). Monitoring services exist specifically to catch these publications before the window closes. For more on building a monitoring strategy, see our guide to trademark monitoring and alerts.

The following table summarizes the three search types:

Common Search Mistakes

Understanding what can go wrong is as important as understanding what to do right. The following mistakes lead to preventable trademark refusals.

Searching only for identical marks. This is a common error. Trademark law does not require marks to be identical to conflict. "QuickKleen" and "KwikClean" can be found confusingly similar. "Bean House" for a cafe can conflict with "Bean Hut" for a coffee brand. If your search only looks for exact matches, you are missing the marks that actually threaten your application.

Overlooking related industries. Identical marks can coexist in completely unrelated fields -- DELTA Airlines and DELTA faucets, for example. But the inverse is also true: marks that are not identical can conflict when the goods and services are related. A clothing brand and a clothing accessories brand are in related markets, even if they are in different Nice Classes.

Ignoring common law marks. Common law trademark rights arise from use in commerce, not from registration. An unregistered mark that has been used commercially in a geographic area has enforceable rights in that area. These marks do not appear in any government database. A comprehensive search must include business directories, domain registrations, social media platforms, and industry marketplaces.

Not searching state registrations. Each of the 50 states maintains its own trademark registry, separate from the federal database. State-registered marks do not appear in the USPTO's Trademark Search system. A mark registered only at the state level can still block a federal application in certain circumstances.

Relying solely on basic word mark search. The USPTO's Wordmark drop-down search does not account for variations in spelling, word order, or phonetic similarity. It does not catch design or logo similarities. And it does not evaluate whether the goods and services are related closely enough to create confusion.

Not understanding the DuPont factors. The 13 DuPont factors are the framework that USPTO examining attorneys, the Trademark Trial and Appeal Board and the Federal Circuit use to assess likelihood of confusion. They go far beyond mark similarity to consider the relatedness of goods, the channels of trade, the sophistication of buyers, and the strength of the prior mark. A search strategy that ignores these factors will misjudge the risk of conflicts it does find.

Why Common Law and State Searches Matter

The USPTO examiner conducting your application review will search only the federal trademark database for conflicting marks. They will not check state registrations, business directories, domain names, or social media. This means even a mark that clears the federal examination could face challenges from a senior common law user or state registrant after the fact.

Common law trademark rights in the United States arise from use in commerce, not from registration. A small business that has been operating under a particular name in a particular region for years has enforceable trademark rights in that region -- even without any government filing. If that business discovers your federal application or registration, they may have grounds for an opposition (before registration), a cancellation proceeding (after registration), or an infringement lawsuit.

The sources that a comprehensive common law search should cover include:

• Business directories -- Dun & Bradstreet, industry-specific directories, local business registries
• Domain name registrations -- WHOIS databases, parked domains, active websites
• Social media platforms -- Instagram, Facebook, LinkedIn, TikTok, X (formerly Twitter)
• E-commerce marketplaces -- Amazon Brand Registry, Etsy, Shopify storefronts
• General web presence -- Google search results, trade publications, press coverage

No free government tool covers these sources. They require either manual research or access to commercial databases that aggregate common law data.

How AI Is Transforming Trademark Search

The limitations of manual and keyword-based search are well understood. The trademark industry's response has been a wave of AI-powered tools that address the gaps systematically.

Phonetic matching algorithms detect marks that sound alike even when spelled differently. METAPHONE and double METAPHONE encoding, combined with trigram similarity scoring, can identify that "KAIWAYINNO" and "KYWAYNO" are phonetically similar -- something a basic text search will never surface.

Visual similarity analysis uses deep learning neural networks to identify marks that look alike, regardless of how they were described or coded in the trademark record.

Conceptual similarity detection uses natural language processing to identify marks that convey the same commercial impression across languages and synonyms -- catching conflicts that are neither visually nor phonetically similar but that a consumer might still confuse.

Automated risk scoring synthesizes multiple similarity signals into a single assessment of likelihood of confusion, giving applicants and attorneys a quantified view of risk rather than a list of potentially relevant records.

Goods and services classification uses NLP to evaluate whether the goods and services associated with two marks are related closely enough to support a confusion finding -- a critical step that basic search tools leave entirely to the user.

The major commercial platforms in this space -- Corsearch/TrademarkNow, Clarivate/CompuMark, and Huski.ai among them -- offer some or all of these capabilities. The USPTO itself is investing in AI-powered examination tools, including image search capabilities and pre-examination automation. For a broader comparison of available tools, see our 2026 trademark search and monitoring tool comparison.

Building a Complete Search Strategy

A sound trademark search strategy is not a single event -- it is a layered process that matches the depth of the search to the stage of the decision.

Stage 1: Ideation (knockout search). Before you fall in love with a name, run a quick knockout search against the federal database. Use the USPTO's free Trademark Search tool to check for identical and near-identical marks in your goods/services classes. If there is an obvious conflict, stop here and generate alternatives. For tips on running effective knockout searches, see our article on how to conduct a trademark search like a pro.

Stage 2: Shortlist (AI-powered screening). Once you have two or three candidate names that survive the knockout phase, run each through an AI-powered search tool that can evaluate phonetic, visual, and conceptual similarity. This is where automated scoring catches the "SON FUEL" conflicts that basic text searches miss. It is also where you begin to see the landscape of potentially related marks -- not just identical ones.

Stage 3: Final candidate (comprehensive clearance). Before filing an application, conduct a full clearance search on your final candidate. This means federal, state, common law, domain, social media, and potentially international databases. For high-value brands, this search should be conducted or reviewed by a trademark attorney who can assess the legal significance of what it finds.

Stage 4: Post-registration (ongoing monitoring). After your mark is registered, shift from searching to watching. Set up monitoring for new filings that are phonetically, visually, or conceptually similar to your mark. Pay particular attention to publications in the Official Gazette, where you have a 30-day window to oppose conflicting marks.

The Cost of Searching vs. Not Searching

The gap between free tools and enterprise solutions is wide -- and it is exactly in that gap where many brand owners find themselves underserved. GleanMark was built to close that gap, providing AI-powered similarity scoring and phonetic matching across 14 million USPTO records, with monitoring alerts that usually arrive within a day or two of a new filing, at a price point accessible to small businesses and solo practitioners.

What a Modern Search Strategy Looks Like

The retirement of TESS in November 2023 was a watershed, but the real shift is broader than any single tool. The trademark search landscape has moved from static, keyword-based queries against a single government database to multi-source, AI-assisted analysis that evaluates phonetic similarity, visual resemblance, conceptual overlap, and goods/services relatedness simultaneously.

A modern trademark search strategy incorporates all of these signals. It begins with free tools for quick elimination, layers in AI-powered screening for intelligent shortlisting, escalates to comprehensive clearance for final candidates, and transitions to ongoing monitoring after registration. Each layer catches conflicts that the previous one missed.

The goal is not perfection -- no search can guarantee that no conflict exists anywhere. The goal is due diligence: demonstrating that you took reasonable steps to identify potential conflicts before committing resources to a brand. That due diligence is not optional.

Tools that provide AI-powered screening and monitoring sit between the free government database and the enterprise platforms that many businesses cannot afford. With phonetic matching, similarity scoring, and automated monitoring across millions of records, they deliver the kind of search intelligence that was previously available mainly to the largest trademark portfolios.

The best time to build a search strategy is before you need one. The second-best time is now.