Free trademark search across US, EU, UK — under three seconds.

The free trademark search runs four probes in parallel and returns one screening signal:

• USPTO — live registrations, pending applications, and the phoneme-distance adjacency check against the full US register plus the TTAB opposition corpus. If the name sits next to a registered senior mark in the same industry class, the screening finds it.
• EUIPO — the European Union register. Relevant for any brand with EU customers, including remote SaaS.
• UKIPO — the UK register. Post-Brexit UK is a separate jurisdiction from EU; a name clear in Munich isn't automatically clear in London.
• Common law — a web scan for prior use that hasn't made it to any register yet. Surfaces up to three commercial sites using the same string. Additive evidence, not verdict-decisive.

Each source reports its coverage state honestly. If the EU index is a day stale, the freshness block says so. If a source failed to respond, the finding says that instead of pretending the check ran green. This matters because a false-clear on a trademark search is worse than a false-block: false-blocks cost you an iteration, false-clears cost you a launched brand.

There are already free ways to search trademark registers. The USPTO's TESS interface is the canonical one. The EU equivalent is eSearch plus. Each is powerful and free — for one register at a time, with results you have to read like a research report.

The three things this search does that a raw register lookup doesn't:

1. One call, three registers. Enter the name once. USPTO, EUIPO, and UKIPO are hit in parallel. You get one signal that summarises all three, not three tabs to reconcile.
2. A screening signal instead of a hit list. PROCEED, PROCEED_STRATEGIC, or ABANDON — one of three tiers, plus the specific findings that drove the tier. You spend seconds triaging, not thirty minutes reading TESS output.
3. A signed receipt an attorney can verify. The result is Ed25519 signed. If your attorney wants to double-check the search that led to a launch decision, the receipt is independently verifiable at etymolt.com/verify. No trust required.

The opposition-risk model that scores the trademark axis was trained on 177,880 real Trademark Trial and Appeal Board opposition outcomes and measured at AUC 0.6315 on a temporal-holdout test set (filing year 2024+). That's what a data scientist would want to know before trusting the score — details on /trust with methodology and confidence intervals.

The classifier is measurably better than random and honestly calibrated. It is not a substitute for a licensed trademark attorney. A free trademark search shrinks the shortlist and makes the paid consult that follows cheaper — that's the whole product.