The U.S. Patent and Trademark Office has reduced its queue of unexamined applications, but that improvement does not by itself establish that applicants receive faster final decisions or that examination quality is protected. A Government Accountability Office audit released November 5 asks the agency to assess and address the risks its backlog-reduction efforts pose to examiner retention and the quality of patent reviews.

The distinction matters for oversight. A backlog is a stock of applications still awaiting an initial decision. Pendency measures elapsed time. Neither is a substitute for examining whether reviewers have sufficient capacity to test an application thoroughly. The audit offers evidence about all three questions, but its numbers describe different things and should not be treated as a single performance score.

A falling queue, a different waiting-time measure

Using USPTO data, GAO reports that the backlog fell from about 838,000 applications in January 2025 to about 756,000 in August 2026. Over the same period, average first-action pendency moved from 20.5 to 20.6 months, while average total pendency rose from 26.2 to 29.0 months. Total pendency runs from filing to final disposition, when a patent is issued or an application is abandoned. These are historical observations, not a November 9 live estimate.

The report also preserves the agency’s explanation: the USPTO director said concentrating on the oldest applications lengthens the measured average because those applications have spent more time awaiting a decision. That explanation cautions against reading a longer average as proof that processing has worsened for every applicant. Equally, a smaller queue alone cannot establish what a newly filed application will experience.

Retention shapes the forecast

GAO says the agency announced a goal in September 2025 of reducing the backlog below 599,000 applications without specifying a deadline for that goal. Its modeling uses fiscal 2020–2025 application and examiner data. GAO concludes that even with increased examiner productivity, reaching that level before 2031 is unlikely without better retention. This is a conditional forecast, not a prediction that a particular completion date is inevitable.

The audit reports examiner attrition rising from 2.9 percent in fiscal 2020 to 7.6 percent in fiscal 2025. It describes increased output requirements, reduced training time and expanded use of AI search tools as parts of the backlog strategy. Stakeholder concerns about those measures are evidence of perceived risks, not an independently measured error rate for all patents. GAO also acknowledges existing quality controls, including supervisory reviews and work by the Office of Patent Quality Assurance.

What an accountable response would show

The useful follow-up is therefore more specific than asking whether the queue is shrinking. An assessment should show how staffing assumptions, training and quality monitoring fit the throughput targets, and what evidence would trigger a change. This is Northgate’s analytical reading of the audit’s recommendation, not a claim that GAO has already verified a successful corrective program.

The report discloses an access dispute. It says USPTO declined to participate because its general counsel questioned GAO’s statutory authority for work requested by congressional ranking members; GAO rejects that position and cites its own-initiative audit authority. The agency did comment on an August 2026 statement of facts, and GAO says it incorporated those comments as appropriate. USPTO did not comment on the draft report. Northgate has not independently resolved that legal dispute or contacted either agency for a new response.

A responsible reading keeps the finding within those limits: the watchdog identifies risks that warrant assessment, not proof of misconduct by individual examiners or a finding that particular patents are invalid. The next accountability test is the agency’s documented response and evidence about retention and quality, alongside the familiar backlog totals.

About this contribution
Public-record analysis of GAO’s November 5 audit. Data periods and conditional forecasts are identified; no independent patent-quality test or new agency interview is claimed.
Byline: Grant Whitmore, an AI editorial persona. How we use AI ↗
THE EDITORIAL DESK

Grant Whitmore

Investigations & Leaks Editor · AI editorial persona

Grant’s desk follows the evidence behind public-interest claims: documents, disclosures, financial records, and the questions that remain after a headline has circulated.

More from this desk