The ‘Dark Matter’ of Patent Law: Study Finds Nearly 25% of USPTO Office Actions Now Cite Secret Prior Art

A large-scale empirical study has found that approximately one in four USPTO office actions now cites prior art that was not publicly available at the time the applicant filed — a phenomenon the study terms “secret prior art” under 35 U.S.C. § 102(a)(2). Patently-O published the analysis in April 2026, drawing on 233 million citation records from 9 million granted U.S. patents (2002–2026) and a random sample of approximately 10,000 office actions per year from the USPTO Patent Examination Data System API, covering 2008 through early 2026.

Under 35 U.S.C. § 102(a)(2), a prior-filed but unpublished patent application can be cited as prior art against a later applicant once it eventually publishes or issues — even though the later applicant had no way to discover it at the time of filing. This “secret springing prior art” is invisible to applicants during prosecution planning, creating what the study calls the “dark matter” of patent law: prior art that shapes claim scope but cannot be seen until after the fact.

The study distinguishes between two measures of secrecy. The legal measure — the conventional § 102(a)(2) standard — counts any reference that was filed but unpublished on the applicant’s filing date. Under this measure, roughly 25% of examiner-cited references in recent office actions qualify as secret prior art. The practical measure is narrower: it asks whether any family member of the cited reference had published the disclosure by the time the applicant filed. Approximately 79% of legally “secret” references had eventually published as pre-grant applications, with an average secrecy gap of 371 days. The remaining 21% were patents that issued directly without a prior publication, carrying a somewhat longer secrecy gap. The practical measure therefore reduces the apparent prevalence of truly unknowable prior art by roughly one quarter.

The trend is upward. The share of secret prior art citations has grown steadily over the study period, reflecting in part the growing volume of patent applications filed globally and the expanding use of continuation filings that can remain pending — and unpublished — for extended periods. For practitioners, the finding underscores a fundamental limitation of pre-filing prior art searches: no search methodology can identify a reference that has not yet appeared in any public database.

The study has direct implications for patent prosecution strategy. Applicants who receive a § 102(a)(2) rejection based on a reference that postdates their filing date by nearly a year may have little recourse beyond claim amendment. The data also support ongoing policy discussions around publication timing, application pendency, and the adequacy of the current grace period framework under the America Invents Act.

The full analysis, including methodology and data visualizations, is available at Patently-O.