NKO DATA LABS
NKO Reality Check #001
Live — free checker, paid batches from €999

NKO Reality Check #001 · Patent legal state

We checked patents the USPTO already declared expired for non-payment. Google Patents still showed 28% as Active.

Every week, the USPTO publishes its own Official Gazette notice of patents expired for failure to pay maintenance fees — a primary official fact, not an estimate (35 U.S.C. 41, 37 CFR 1.362(g)). We pulled the last 52 weeks (179,840 real patent numbers, source below), took a random sample of 50, and checked what Google Patents showed for each one on the same day. 14 of 50 (28.0%, 95% CI 17.5%-41.7%) still showed "Active" — weeks to months after the USPTO's own record already confirmed expiration.

What this is, and isn't. Google Patents doesn't compute legal status itself — it licenses it from a third-party feed (IFI CLAIMS), the same kind of feed shared across much of the patent-data industry, and says so plainly on every page: "The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed." We followed up on how long this specific gap lasts (a real update-lag of roughly 2-5 months on the cases we found, not a permanent error — see Method below), and we checked whether this is a general third-party-data problem or specific to Google Patents: on the same type of sample (100 cases, 50 confirmed-lapsed + 50 confirmed-active), Lens.org (a free, independent patent database) got 99.0% right overall, against 86.0% for Google Patents — see Provider comparison below. This is not "Google is wrong" in general — it's a specific, measured, dated gap in one feed.

Try your own patent — free, no account

Enter a US patent number. We show you both numbers side by side, with the source for each.

Method

Ground truth: the USPTO Official Gazette's own weekly "Notice of Expiration of Patents Due to Failure to Pay Maintenance Fee", published in the open, no account needed (uspto.gov/web/offices/com/sol/og/{year}/week{NN}/TOC.htm). Each patent number listed there is an already-confirmed official fact from the agency that grants the patent — checking a third party against it is a binary match against a primary source, not a statistical inference.

Sample: random.sample() with a fixed, disclosed seed (20260906) over the complete 179,840-record corpus, selected before any of the 50 patents were looked up on Google Patents. We also checked how long the discrepancy lasts: of the 14 mismatches, all were 60-144 days old (2.0-4.7 months); a separate random sample of 20 cases aged 180-365 days came back 0/20 incorrect. That means the gap we measured is a real update-lag window of a few months, not a persistent error — reported here as exactly that, not rounded up to something stronger.

This benchmark covers only one type of legal-status error (expiration from unpaid maintenance fees) — it says nothing about litigation outcomes, assignments, or jurisdictions outside the US. Full corpus and code: github.com/NKO-LLC-TECH/nko.

Provider comparison — is this a Google-specific gap?

We built a second, larger sample to test that: 100 patents with known USPTO ground truth (the same 50 confirmed-lapsed patents above, plus 50 confirmed-active patents as a negative control — recently-granted patents whose first maintenance fee statutorily cannot be due yet, 37 CFR 1.362(d)) and checked two independent sources: Google Patents, and Lens.org, a separate free patent database.

MetricGoogle PatentsLens.org
Overall accuracy (n=100)86.0%99.0%
False-active (says Active, USPTO confirms lapsed — n=50)28.0% (14/50)2.0% (1/50)
False-inactive (the serious error — says not-active on a truly active patent, n=50)0.0%0.0%

The honest finding here is more specific than "third-party patent data is unreliable": Lens.org's single error lands on the exact same patent as one of Google's 14 — there's no case where Lens.org is wrong and Google is right. That doesn't look like every provider having its own scattered blind spots; it looks like Google Patents specifically running behind on this one type of update. We're not going to claim more than that measurement supports — this comparison used two free sources; whether the same pattern holds for paid patent data providers is a separate, unfinished question we're not overstating here either way.

The 20 verified cases

14 real mismatches (highlighted) plus 6 correctly-shown cases from the same random sample, for contrast — every row links to the real Gazette notice and the real Google Patents page, so anyone can verify it independently.

PatentGoogle Patents showedUSPTO Gazette confirmsExpired onEvidence

Batch checks

Free: up to 50 patent numbers per request, no account, same engine as the single check above. Paid: up to 5,000 patents for €999, delivered within 24-48h as a downloadable file — each row carries the possibly-incorrect state, the official USPTO evidence link, and the date it was checked.

€999 — batch up to 5,000 patents

For more than 50 patents at once. Same textarea above works for this — paste your list, confirm, pay, get a signed link to your results within 24-48h.