Trademark Sentinel | Blog

WIPO Global Brand Database explained

Wed, 06 May 2026 00:00:00 GMT

If you’ve spent any time around international trademark work, you’ve heard of the WIPO Global Brand Database — usually shortened to GBD. It’s the biggest single trademark search interface in the world, and it’s the data source Trademark Sentinel uses today. This post explains what’s actually in GBD, what isn’t, and why we picked it as the MVP source.

What GBD is

The Global Brand Database is a free public search tool maintained by the World Intellectual Property Organization (WIPO). It aggregates trademark and appellation-of-origin records from a number of sources into one searchable index. As of writing, it indexes on the order of 60+ million records, covering both international filings and national contributions from participating offices.

Two things are worth pulling apart in that sentence.

International filings are records from registers WIPO itself runs — most notably the Madrid System, the international trademark registration treaty that lets a brand owner file once and designate dozens of member countries. Madrid records flow into GBD natively. So do filings under the Lisbon System (appellations of origin like “Champagne” or “Roquefort”), the Article 6ter list (state emblems and intergovernmental organisation names), and a few smaller WIPO registries.

National contributions are records that participating national offices send to WIPO for inclusion in GBD. Around 70 national offices contribute, including major ones like the United States (USPTO), the United Kingdom (UKIPO), Australia (IP Australia), Canada (CIPO), and many EU member states. The European Union Intellectual Property Office (EUIPO) contributes its trademark data as well.

The result is the closest thing the world has to a global trademark search. Not the only global search, but the broadest free one.

What GBD does not cover

Equally important — what’s missing.

Not every national office contributes. Several large jurisdictions are absent or only partially represented. China and Japan, in particular, have partial coverage; many smaller offices contribute irregularly or not at all. If your business depends on a specific national office, check whether it’s a GBD contributor before relying on GBD alone.
Update lag varies by source. WIPO’s own registries (Madrid, Lisbon) update with little delay — often within days. National contributions update on whatever cadence the contributing office chooses, which can be weekly, monthly, or quarterly. A “fresh” GBD record for one office may be a week old; for another, several weeks.
No commerce-use data. GBD is a register search, not a market-use search. It tells you what’s been filed; it doesn’t tell you what’s actually being sold under a brand. Use-based research is a separate tool.
No domain or social-handle data. Brand monitoring at large includes domain squats, social-media handle squats, and marketplace listings. None of that is in GBD. It’s a register-level data source, full stop.

The honest summary: GBD is the best single place to start, and you’ll eventually want more sources if your enforcement strategy demands them. We think most teams should start with GBD and add coverage as they hit its limits, rather than starting expensive and broad on day one.

How Trademark Sentinel queries it

At a high level, Trademark Sentinel runs a scheduled search pipeline against GBD: for each watchlist mark, we issue a similarity-aware query against the GBD search interface, normalise the results, deduplicate against records we’ve already seen, score the new ones for similarity, and emit alerts for the matches that cross your threshold.

A few implementation notes that affect what you’ll see as a user, without getting into internals:

Session-bound search. Our queries run inside a managed search session, not as a single one-shot lookup. That keeps result paging consistent between similarity passes and lets us evaluate the same query against the current GBD snapshot end-to-end.
Proof-of-work pacing. GBD is a free public service and we want to keep it that way for everyone. Our query layer paces itself — including a proof-of-work step on each session — so we stay well within fair-use rates no matter how many watchlists are running on our side.
No internal scraping shortcuts. We use the public GBD interface as WIPO publishes it; we don’t try to bypass rate limits or undocumented endpoints. If GBD is down or slow, your alerts arrive when GBD recovers.

If you’re a technical reader interested in the API surface for your own integrations — pulling matches into a case-management system, for example — the API reference has the full schema.

Why GBD is the MVP source

This is a deliberate scope choice. We started Trademark Sentinel with one source for three reasons:

Coverage breadth. Madrid plus 70 national contributions hits more useful registers from one integration than any other single source. For a brand-monitoring product launching today, GBD-only catches the majority of internationally relevant filings on day one.
Operational simplicity. Adapter code for any new register is a real commitment — we have to keep it working as upstream interfaces shift. Shipping one adapter well beats shipping three poorly.
A clear v2 roadmap. Next on the list are direct USPTO and EUIPO adapters (faster updates than GBD’s relayed copies of those offices) and WHOIS-based domain monitoring for teams whose enforcement extends past the registers.

When v2 sources ship, they’ll show up as additional sources you can add to a watchlist alongside GBD.

Practical guidance

Quick fit check against GBD-only coverage:

International enforcement: strong fit. Madrid is excellent, and national contributions catch most of what slips outside Madrid.
Pure-domestic in a market with a v2 adapter on the roadmap (US, EU): workable today, stronger when v2 ships.
Pure-domestic in a non-contributing jurisdiction: not enough on its own — talk to us.

If you want to dig into the source itself, the WIPO Global Brand Database is publicly searchable at branddb.wipo.int. It’s worth ten minutes of exploration, even if you end up using a watch service to do the daily monitoring for you.

Nice classification 101 for non-attorneys

Tue, 05 May 2026 00:00:00 GMT

If you’ve ever filed a trademark or read a watch alert, you’ve run into the Nice classification. It looks like bureaucratic plumbing, and it is — but it’s also the single biggest lever you have over how broad, how expensive, and how enforceable your trademark is.

This post is the version I wish someone had handed me before my first filing. No legal advice; just the mental model.

What the Nice classification is

The Nice classification — formal name, Nice Agreement Concerning the International Classification of Goods and Services for the Purposes of the Registration of Marks — is a single, internationally agreed list of 45 classes that all signatory trademark offices use to categorise what a brand is registered for. Classes 1–34 cover goods (physical products); classes 35–45 cover services. WIPO maintains it, and a new edition lands every five years (we’re currently on the 12th edition; minor revisions ship annually).

Why does it exist? A trademark is never absolute. “Apple” is registered for computers and for music streaming, but a fruit stand can still call itself Apple Orchard without infringing. The Nice classification is the formal mechanism for that: trademarks live inside specific classes, and conflicts between marks generally only matter within or across related classes.

The 45 classes, briefly

You don’t need to memorise them, but knowing the shape of the list helps:

Range	What lives there
1–5	Chemicals, paints, cosmetics, lubricants, pharmaceuticals
6–11	Metal goods, machines, hand tools, scientific apparatus, electronics, vehicles, lighting
12–17	Vehicles, firearms, jewellery, instruments, paper, rubber/plastic
18–23	Leather, building materials, furniture, kitchenware, textiles, yarn
24–28	Fabrics, clothing, lace, carpets, games and toys
29–34	Food, beverages, agricultural products, tobacco
35–40	Advertising/business, insurance, telecommunications, transport, materials processing
41–45	Education and entertainment, scientific and tech services, food/lodging, medical/personal, legal/security

The most useful insight here is that classes are not by industry, but by output type. A software company’s product (downloadable software) lives in class 9 alongside thermometers and circuit boards. The same company’s service (software-as-a-service, hosting) lives in class 42 alongside scientific research. That’s why software brands almost always file in both.

Why classes 9, 25, 35, 41, and 42 are the popular ones

Look at any year’s filing statistics from WIPO and you’ll see the same heavy hitters:

Class 9 — Electronics, downloadable software, scientific apparatus. Any app you can install — mobile or desktop — lives here. So do video games, recorded media, glasses, headphones, and most consumer electronics.
Class 25 — Clothing, footwear, headgear. Fashion is enormous, and the class is narrow enough that you almost can’t avoid it if you sell apparel.
Class 35 — Advertising, business management, retail services. Catches e-commerce, retail stores, marketplaces, and most “we sell other people’s stuff” business models. If you run a Shopify store, you probably belong here.
Class 41 — Education, entertainment, training. Online courses, podcasts, publishing, sports events, gaming services as a service. A surprisingly large basket.
Class 42 — Software-as-a-service, scientific services, design. Where the rest of the software industry lives. Hosting, SaaS, custom development, UX design, IT consulting.

If your business sits anywhere near consumer tech, retail, or media, you’ll end up in two or three of these. That’s normal.

How to pick the right class

Three rules of thumb that won’t replace counsel but will save you from obvious mistakes.

1. File for what you sell, not what you do internally. A coffee company doesn’t file in class 30 (coffee, the substance) just because they buy coffee beans. They file in class 30 if they sell coffee under that brand. The trademark protects what reaches the customer.

2. Distinguish goods from services. “We have a downloadable app and a web app” usually means class 9 and class 42. “We sell a physical product and run a retail website that also sells other people’s products” usually means the goods class and class 35. Pretending you only need one class because the offerings feel like one thing is the most common founder mistake.

3. File where you operate, not where you exist. The class strategy and the country strategy are independent. You can file class 9 in the US, EU, and Japan and not in the UK. Most companies start with their home market plus their two or three biggest export markets — adding more is cheap to plan but expensive to actually file.

Common pitfalls

A short list of things that bite new filers:

Filing too narrowly. “Software for accountants” inside class 9 is enforceable only against software-for-accountants conflicts; a filing of the same mark for “software for dentists” might slip through. Conversely, filing too broadly invites office actions and partial refusals.
Forgetting class 35 for retail. If you sell things online, class 35 protects the retail service, not the goods themselves. Surprising number of D2C brands miss this.
Filing class 41 for SaaS. Class 41 is education and entertainment, not software. SaaS belongs in class 42, even if your product happens to be a learning tool. (The content might be class 41; the platform is class 42.)
Translating the spec verbatim from one office to another. Class specifications — the actual text describing your goods — vary by office. USPTO is famously picky about specificity; WIPO and EUIPO are looser. Don’t copy-paste.

Watching by class

Once you’ve filed, the Nice class also drives how you watch. A clothing brand in class 25 doesn’t need an alert when somebody else files the same mark in class 9 (electronics) — those goods don’t compete. Filtering watches by relevant classes drops the noise floor dramatically.

The user guide on watches walks through how to filter your Trademark Sentinel watchlist by Nice class, including the option to broaden to “related” classes (e.g., 25 plus 18 for leather goods, plus 35 for clothing retail).

What is a trademark watch?

Mon, 04 May 2026 00:00:00 GMT

If you own a brand, the registers never stop moving. Every week, somebody, somewhere, files a new trademark that may sit a little too close to yours. A trademark watch is the standing arrangement that catches those filings while you still have time to act.

The definition

A trademark watch is a continuous monitoring service over one or more trademark registers. You give it a list of marks — your own brand, your product names, maybe a competitor or two — and the watch flags new applications that look similar enough to be worth a closer look. The output is a stream of alerts: “this application published last week, here is the mark, here is the owner, here is the deadline to oppose.”

Watches differ from one-off clearance searches. A clearance search is a snapshot — the answer is “as of today, here’s what’s filed.” A watch is a subscription — the answer is “tell me whenever something changes.” Both have a place; this post is about the second one.

Why monitor proactively

There are two ways to defend a trademark portfolio. The first is reactive enforcement: you discover an infringer in the wild — on a shelf, on a website, in an ad — and you send a cease-and-desist or sue. By the time you’ve spotted the infringement, the other party has already invested in their brand. Pulling them off the market is slow, expensive, and contested. Sometimes you lose.

The second is proactive opposition. Most jurisdictions publish new trademark applications before they grant, and give third parties a fixed window — typically 30 to 90 days — to oppose. If you catch a conflicting application during that window, opposition is dramatically cheaper than litigation. The applicant hasn’t spent money on packaging, marketing, or distribution yet. Often the matter ends with a withdrawal or a coexistence agreement, not a lawsuit.

Trademark watches exist to make proactive opposition workable. Without a watch, you’d need someone to read every gazette in every relevant jurisdiction the day it publishes. Nobody does that by hand any more.

What sources matter

Trademarks are jurisdictional, so a watch is only as good as the registers it actually reads. The sources that matter most:

WIPO Global Brand Database — the closest thing to a global view. It aggregates international filings made under the Madrid System plus contributions from many national offices. If you’re tracking a brand with international reach, this is the one source you cannot skip.
National offices — USPTO for the United States, EUIPO for the European Union, UKIPO for the UK, and so on. National offices catch domestic-only filings that never go international, which is the majority of filings.

For Trademark Sentinel, WIPO Global Brand Database is the source we cover at launch. Direct adapters for USPTO and EUIPO are on the v2 roadmap; we’d rather ship one source well than three sources poorly. If your enforcement strategy depends on national-only US filings today, factor that in.

What an alert looks like in practice

Once you’ve defined a watchlist — say, your brand “Acme” plus the variants “Acme Pro” and “Acme Cloud” — a typical alert might read:

New application: ACMEE CLOUD — Filed 2026-04-28 by Acmee Holdings Ltd in Nice classes 9, 42. Similarity: 0.86 (visual + phonetic). Watchlist match: “Acme Cloud”. Source: WIPO GBD. Opposition window: closes 2026-07-28.

The interesting fields:

Similarity score — a measure of how close the filed mark is to the watched mark, by visual edit distance and phonetic comparison. Modern watch services expose this so you can triage quickly: a 0.95 needs review today, a 0.55 might be a coincidence.
Class match — the Nice classification context. A clothing-class filing for “Acme” matters more if you sell clothing than if you sell software. (We cover Nice in a separate post.)
Deadline — the opposition window for the relevant office. Miss it and your cheapest option disappears.

Good alerts let you make a yes/no decision in under a minute. Bad alerts bury the conflict in noise; that’s why similarity scoring and class filtering matter so much.

What a watch is not

A watch is not a legal opinion. It surfaces things worth looking at; whether to oppose is still a judgment call, often involving counsel. A watch is also not a guarantee — registers have publication delays, and similarity scoring is imperfect by design (too strict and you miss conflicts, too loose and you drown in alerts). Treat it as a daily news feed for your brand, not as infallible truth.

Where to go from here

If you’re setting up watches for the first time:

Read the user guide on watches for how to configure a watchlist, set similarity thresholds, and filter by Nice class.
Start a free watch — the signup form takes about a minute.

The cheapest opposition is the one you file before the other side has even seen the artwork. A watch is how you get there.