Free Check · US register · Brazil from Nov 10

Try the instant check.

Type a name and see the closest marks in USPTO records, looking for the same name, look-alikes and sound-alikes. Brazil opens on Nov 10. It runs in your browser: we don’t send the name you type anywhere. Your browser downloads small parts of our index and compares them on your device.

A preliminary check, not legal advice. No sign-up, no email, nothing stored.

What each check searches

RegisterFree Check (in your browser)Name Report (paid)
United States (USPTO)SearchedSearched, not yet verified
Brazil (INPI)Opens on Nov 10Searched, not yet verified
United Kingdom (UKIPO)Official free check linkSearched by hand, not yet verified
European Union (EUIPO)Official search linkOfficial search link
International (WIPO)Official search linkOfficial search link

“Searched, not yet verified” means we search that register, but it hasn’t passed our public test yet. See the test, misses first.

What this check covers, and what it doesn’t

Covered

  • United States · USPTO: live registrations and pending applications that have a word element, from USPTO records as of 30 September 2026 (a public-domain copy of the USPTO’s bulk data).
  • Brazil · INPI, from Nov 10: live registrations and pending applications that have a word element, from INPI open data dated 17 October 2026 (dispatches up to RPI 2910, 13 October 2026).
  • Same name, the name as the distinctive word of a longer mark, look-alike spellings (small typing differences, c/k, ph/f), and sound-alikes in English (and in Portuguese for Brazil, from Nov 10).

Not covered

  • The UK IPO, EUIPO, EU national offices and WIPO. We link to each office’s own search instead.
  • US applications filed, and US status changes made, after 30 September 2026. Use the USPTO’s own search for anything newer.
  • Brazilian applications filed or changed after 13 October 2026. New applications only appear once INPI publishes them, in any search, including INPI’s own.
  • Brazilian applications that were refused and are now under appeal.
  • Translations, logos and figurative marks without words, marks that died recently, and a name hidden in the middle of a longer word.
  • Very common words: we compare the closest groups your browser can load quickly.

How a result is worked out

Each mark we find gets a similarity score (same name, distinctive word, sound-alike, look-alike or contained name), weighted by how related its classes are to yours and by its status. The highest score sets the register’s label: “High similarity”, “Some similarity” or “No close match”. Where our data can’t support a result, we say “Couldn’t check”. To keep the check small enough to run in your browser, each mark shows its number, classes and status; the owner’s name is on the office’s own record. Marks are shown as their owners published them; a similar mark is information, not a sign anyone did anything wrong.

How we checked the US data

Since June 2026 the USPTO asks for a signed-in USPTO.gov account before you can download its bulk files. So we use a public copy of those files on Hugging Face (illeatmyhat/uspto-trademarks). It is rebuilt from the USPTO’s annual and daily trademark files, runs to 30 September 2026, and is labelled a US government work in the public domain. Because it is a copy made by someone else, we checked it against the USPTO’s own records before using it.

On 22 October 2026 we drew 50 marks at random from our US index, using a published seed (lightfathom-us-spotcheck-v1), and compared each one with its record on the USPTO’s own status page (TSDR). We compared the mark’s wording, whether it is live or pending, and its classes. All 50 matched on all three. In 4 of the 50, the USPTO had updated the case after 30 September (for example, from “new application” to “under examination”); each mark was still live. This check looked at marks in our index. It can’t show a mark that is missing from the copy, which is one reason we still link to the USPTO’s own search.

Is anything missing from our US copy?

We tested it two ways. From the USPTO’s Official Gazette issues before 30 Sep 2026, we drew 100 published marks at random; 100 were in our index. From serial numbers assigned in Jul–Sep 2026, we’re drawing 100 at random and looking each one up on the USPTO’s site by hand; that sample is in progress, and the count goes here when it’s done. Seeds and full lists

The gazette sample used the four issues of 8, 15, 22 and 29 September 2026, downloaded as published files from the USPTO’s Official Gazette page: 41,232 marks published for opposition with words in them. We ranked them by a published seed (lightfathom-us-gazette-v1) and looked up the first 100 by serial number in our own index, with no queries to the USPTO. With none missing in 100, the share missing is likely under about 3.7% (95% upper bound). The gazette only lists marks that reached publication, which is why the second sample looks at recent filings.

Can the data be replayed to a past date?

To test our matching on real conflicts, we rebuild the US register as it stood on the day a later mark was filed, from the copy’s full case files, dead marks included, and each case’s event history. We checked that replay on 100 marks (50 live, 50 dead): it rebuilt today’s status for 99 of them. For 42 marks we compared the full prosecution history with the one the USPTO shows: every entry dated up to 30 September 2026 had an event on the same date in the copy. Records from before about 1980 are thin: some have only a few events and no registration event, so for those we use the registration date on the case file.

How old the US data can get

We count the age of the US data from its date, 30 September 2026, not from when the copy was last updated. We always show that date above US results. From 30 days old we add a notice with the USPTO’s own search next to it. From 60 days old we pause US results and show only the USPTO’s own search, already filled in with your name.

Lightfathom is not affiliated with or endorsed by any trademark office.