The USPTO Kept Examining Through the Longest Shutdown in History — and Q4's Numbers Show It

Quarterly USPTO Prosecution Report | USPTO data through December 31, 2025

The federal government was shut down for 43 of the 92 days in this quarter — the longest shutdown in U.S. history, from October 1 through November 12 — and the trademark examination corps barely blinked. The USPTO issued 93,419 non-final office actions and 12,977 final actions in Q4 2025, kept publishing marks, and kept clearing responses, all while most of the government went dark. The one visible dent is exactly where you'd expect it: a 12.9% drop in non-final actions in November, the month that absorbed six weeks of furlough risk, a reduction in force, and a satellite-office closure. By December, output had snapped back.

The trademark office is the rare federal agency that runs on its own money, and Q4 is the cleanest demonstration we have of what that buys you: a shutdown that closed the Copyright Office outright barely registered on the trademark register.

— Howard Katzenberg, Founder, GleanMark

That independence is structural. The USPTO is funded by filing fees rather than annual appropriations, which is why it stayed open on operating reserves while the rest of Washington closed. The point of this report is to show what "stayed open" actually looked like in the data — and then to get into the refusal and argument metrics that practitioners came for.

Executive Summary

• The quarter survived a 43-day shutdown. Examination continued on fee reserves; the only clear footprint is November's dip in non-final actions, which rebounded in December.
• The November dip has a cause, not a mystery. Non-final actions fell from 32,918 in October to 28,665 in November, then recovered to 31,836 in December — the trough lines up with the shutdown weeks and a ~1% USPTO reduction in force.
• Section 2(d) owns the analyzed docket. In the analyzed subset of 9,962 Q4 responses, likelihood-of-confusion was the primary refusal in 8,067 of them — 81.0%.
• Class deletion and consent agreements are the sharpest tools in the box. They post the highest directional effectiveness scores (0.968 and 0.966), but on small bases; goods amendments do the heavy lifting at scale (2,270 instances, 0.960).
• Read the refusal numbers as a sample, not a census. They cover 15,236 analyzed office actions and 9,962 responses, against 93,419 total USPTO non-final actions for the quarter.¹

1. Prosecution Volume

Period	Non-Final Actions	Final Actions	Non-Final + Final	All Prosecution Documents	Non-Final MoM
October 2025	32,918	5,182	38,100	Not available	—
November 2025	28,665	4,260	32,925	Not available	-12.9%
December 2025	31,836	3,535	35,371	Not available	+11.1%
Q4 2025 Total	93,419	12,977	106,396	523,290	—

The shape of the quarter is the story the table can't tell on its own. October ran hot. November fell off. December clawed most of it back without fully closing the gap. That is precisely the profile you'd predict from a quarter in which the government shut down on day one and stayed shut for six weeks.

We think the November trough mostly reflects the shutdown drag rather than any change in examination policy. The USPTO stayed open, but it did not stay untouched: the agency executed a reduction in force affecting roughly 1% of its workforce in early October and announced the permanent closure of its Denver outreach office, both during this quarter. None of that stops an examining attorney from issuing an office action, but a workforce operating under furlough threat and missed-paycheck headlines is not a workforce running at full throttle. December's rebound is consistent with a corps catching up once funding was restored on November 12 — though the recovery stopped short of October's level, and the late-December federal-holiday closures on the 24th and 26th cost the office working days right at quarter's end.

Final actions tell a quieter, different story: a steady monthly decline from 5,182 to 4,260 to 3,535, with no rebound. A final action is downstream of a prior non-final and a response, so it carries more inertia and reacts more slowly to a single disrupted month. For docketing, the practical read is that Q4 ran two clocks at once — a heavy, shutdown-jostled non-final load up front and a final-action stream that simply tapered.

2. Office Action Document Types

USPTO Code	Document Type	Documents in Quarter
NFIN	Non-Final Action	93,419
XSS	XSearch Search Summary	91,031
AMC	Amendment and Mail Process Complete	55,349
ROA	Response to Office Action	45,612
APP	Application	30,147
PNR	Public Note	27,005
EDS	Notice of Design Search Code Emailed	15,995
PST	TRAM Snapshot of App at Pub for Opposition	15,335
PB1	Notice of Publication	14,949
EB4	Notification of Notice of Publication	14,949
PB3	OG Publication Confirmation	13,346
FREF	Final Action	12,977
SPE	Specimen	11,820
CAR	Change Address or Representation Form	6,651
EXAMA	Examiner's Amendment	5,209
ELR	Application Extension to Response	5,082

One number here is worth sitting with: 45,612 responses to office actions were filed in the quarter. Filers kept responding straight through the shutdown, which makes sense — the people answering office actions are private-sector trademark counsel, not federal employees, and statutory deadlines kept running. The shutdown was an agency-side event, not a filer-side one, and the response volume shows it.

Do not add these rows together. Different codes capture related stages, parallel system artifacts, or distinct procedural events, and the publication-related codes in particular shadow one another (note PB1 and EB4 landing on the identical 14,949). The non-final and final totals in Section 1 are the office-action baseline; this table is the surrounding file-history mix.

3. Refusal Landscape

Primary Refusal Type	Cases	Later Registered	Share of Analyzed Responses
Section 2(d) — Likelihood of Confusion	8,067	7,552	81.0%
Section 2(e) — Descriptiveness / Surname	997	480	10.0%
Specimen Refusal	357	341	3.6%
Other	332	318	3.3%
Identification of Goods/Services	169	164	1.7%
Disclaimer Requirement	40	34	0.4%

Section 2(d) and Section 2(e) together account for 9,064 of 9,962 analyzed responses — 91.0% of the docket sits on confusion and descriptiveness. The rest is rounding error by comparison. If you want to know what trademark prosecution is actually about, it is talking the examiner out of a cited mark and, secondarily, arguing your term is more than merely descriptive.

The more interesting view is where issues hide rather than lead. Disclaimer requirements were the primary refusal in just 40 responses but show up as a tag in 742 cases — meaning disclaimer practice almost never drives a refusal on its own and instead rides along with something bigger. Identification issues follow the same logic at smaller scale: 169 primary versus 483 tagged. Among tagged matters that later registered, time to registration ran from 63 days for specimen-tagged cases up to 109 days for Section 2(d)-tagged cases, with Section 2(e) at 85 — confusion refusals take the longest to clear because they take the most argument. (Tag counts overlap; don't sum them or convert them to shares of the docket.)

4. Argument Effectiveness

Refusal Type	Directional Score (0–1)	Argument Instances	Verified-Win Instances
Section 2(d) — Likelihood of Confusion	0.947	5,729	4,831
Section 2(e) — Descriptiveness / Surname	0.901	1,714	746
Specimen Refusal	0.829	262	244
Disclaimer Requirement	0.911	40	33
Identification of Goods/Services	0.858	99	94
Other	0.771	150	129

Argument Category	Directional Score (0–1)	Outcome Multiplier	Instances
Class Deletion	0.968	1.194	320
Consent Agreement	0.966	1.191	119
Goods Amendment	0.960	1.197	2,270
Commercial Impression	0.945	1.200	932
Cited Mark Abandoned	0.939	1.200	87
Supplemental Register	0.939	1.184	849
Third Party Registrations	0.934	1.199	167
Weakness of Cited Mark	0.930	1.200	296
Disclaimer	0.917	1.183	696
Goods Coexistence	0.910	1.198	391
Mark Distinction	0.910	1.199	719
Market Channel Distinction	0.904	1.199	159
Secondary Meaning	0.878	1.157	154
Consumer Sophistication	0.876	1.199	130
Suggestive Not Descriptive	0.861	1.144	147

The scores are directional indicators on a 0–1 scale and the multipliers are relative weights — neither is a win rate. With that caveat, the standouts split into two kinds.

The high scorers at the top — class deletion (0.968) and consent agreement (0.966) — are powerful precisely because they change the facts rather than argue them, but they run on thin bases (320 and 119 instances) and only fit specific situations. A consent agreement is a Section 2(d) instrument; it does nothing for a descriptiveness refusal. The workhorse is the goods amendment: 2,270 instances at a 0.960 score, far more common than any other top-tier category. Narrowing an identification is the move practitioners reach for most because it bends both the confusion analysis (by shifting trade channels and purchaser context) and, in some cases, the descriptiveness inquiry.

Practitioner Takeaways

1. Plan around Section 2(d) first. It was the primary refusal in 81.0% of analyzed responses. Cited-mark strategy, goods and channel distinctions, and consent/coexistence pathways are the core of the job, not a sidebar.

2. Reach for the goods amendment when it changes the facts. It has both the volume and the score. Narrowing the identification can move a Section 2(d) analysis by altering trade-channel or purchaser context — and occasionally a descriptiveness refusal when the narrower wording shifts the inquiry.

3. Match the tool to the refusal. Consent agreements are Section 2(d)-only. Supplemental Register amendments address Section 2(e) descriptiveness or genericness for use-based applications. Specimen substitution fixes specimen defects. None of these cures a likelihood-of-confusion refusal by itself.

4. Don't read a slow middle month as a slow quarter. November's dip was a shutdown artifact, not a trend. December rebounded. If your docket logic assumes examiner output falls off into year-end, this quarter says otherwise.

Explore more USPTO data analysis on the GleanMark Insights blog.

Footnotes

1. Refusal and argument metrics throughout describe GleanMark's analyzed subset — 15,236 office actions and 9,962 responses tagged for the quarter, against 93,419 total USPTO non-final actions — and are directional, not USPTO-wide refusal counts. NICE class designations are not applications and are not counted as such anywhere in this report. Data sourced from USPTO TSDR and public prosecution records, as recorded through the report date. Informational only; not legal advice. ↩