The AI land grab has reached the trademark register

Start with the raw demand. First-half filings rose from 290,538 in 2024 to 342,760 this year — up 18% in two years — and AI is doing far more than its share of the lifting. Applications whose goods-and-services text mentions “artificial intelligence” nearly tripled, from 5,098 to 16,433, rising from 1.75% to 4.79% of everything filed — about one in twenty-one. Crypto, the last vocabulary to sweep the register, peaked quietly in 2025 at 0.67% and is fading.

The more interesting question is who is filing, and the register answers it two ways. First: 63.4% of AI filings are intent-to-use — the box you check when the product doesn’t exist yet — versus 48.1% for the register as a whole, and that gap has widened every year (53.7% → 59.3% → 63.4%). This is name-staking ahead of launches: claims on an imagined product line, not labels on a shipping one. Second: they are overwhelmingly new names. Only 32.5% of 1H AI filers had ever filed before under the same name — almost exactly the register-wide rate of 34%. The AI rush is not big incumbents relabeling their catalogs; it is a wave of first-time entrants, arriving at the Office at the same rate every gold rush does.

Where does all that AI language land? Overwhelmingly in Class 042 — the software-and-technology-services class — and it has rewritten the class’s vocabulary. Five years ago, the most common substantive words in Class 042 descriptions were the timeless furniture of tech filings: research, medical, apparatus, processing, devices. “Intelligence” and “artificial” sat outside the top 25. In the first half of 2026 they are the #1 and #2 words in the entire class — each appearing roughly two-and-a-half times as often as the next word (26,555 and 25,145 mentions vs. 10,789 for “research”).

And here is the punchline the register delivers on schedule: once every brand wants a word, the Office stops letting anyone own it. Among applications filed this half, “AI” is now the single most-disclaimed term at the USPTO — 563 applicants so far have had to formally renounce exclusive rights to it, more than perennial disclaimer champions like “company” or “coffee.” Five years ago it sat outside the top 40. A disclaimer is the examiner’s way of saying a word belongs to the language, not to you. AI just became that kind of word.

All of this demand lands on an agency that spent the early 2020s buried. The pandemic-era filing surge left the Office with a queue it has been working through ever since — and the first half of 2026 is the clearest evidence yet that it is winning. Every stage we can measure, measured by when the event actually happened, is faster than a year ago, and much faster than two:

The headline is the first clock: an applicant who filed into the 1H 2024 queue waited a median 251 days for an examiner’s first Office Action; for actions issued this half the median is 134 days — nearly twice as fast. Marks that hit an Office Action and fixed it reached publication 21 days sooner than in 2024. End to end, marks registering this half got there in a median 297 days, down from 439 — a third faster — and the Office issued 223,740 registrations, up 42% over 1H 2025. That last pair is the tell: faster medians and higher volume at the same time is what a backlog being cleared looks like, not a quiet quarter.

For applicants the practical read is simple: the years of “file and wait a year to hear anything” are ending. A mark filed today gets in front of an examiner in roughly four and a half months — which also means conflicts, refusals and deadlines all arrive sooner.

Speed cuts both ways: more examination means more Office Actions, sooner. Across the 214,987 refusal-bearing Office Actions in our analyzed corpus from this half, four issues dominate — and they arrive in a near-tie:

The pattern matters more than any single number. Only two of the top objections — likelihood of confusion with an existing mark (§2(d), 27%) and descriptiveness (§2(e)(1), 10%) — are substantive judgments about whether the mark can be protected at all. The rest of the top of the list — disclaimers, goods-and-services wording, specimens, descriptions — are fixable formalities: things a careful application avoids and a competent response cures. In other words, most of the gauntlet is paperwork, not verdicts. The refusal an applicant should actually fear is §2(d) — the one you cannot paperwork your way out of, and the one the next section is about.

When an examiner refuses a mark under §2(d), they cite the specific registration in the way. Tally those citations across our corpus and you get a map of where American brand ambitions pile up — and the top of the map is not a tech giant. It is a 1975 registration on the word PICKLE-BALL, cited against 51 applications in our corpus this half, up from 38 a year ago. America’s fastest-growing sport keeps minting paddle brands, leagues and academies, and every one of them files into the same fifty-year-old wall.

Read down the list and it is a portrait of the consumer economy’s crowded corners: a sport growing faster than its naming space, a protected spirit riding an agave boom, a burger word half the industry wants. And one entry that reads like a preview: an AI-named software registration already blocking newer AI applicants. Sixteen thousand AI filings a half will do that — the crowding this report opened with is starting to show up as §2(d) citations, right on schedule.

Figures come from GleanMark’s structured copy of the public USPTO record, queried directly. Four instruments, each labelled where used:

• Filing demandEvery U.S. application by filing date, first halves (January 1 – June 30) of 2024–2026 compared like-for-like. AI/crypto terms are case-insensitive matches in the goods/services text — the literal phrase "artificial intelligence" for AI (bare "AI" excluded, so the share is conservative). Intent-to-use is the application's ITU flag; "repeat filer" means the exact owner name appears on any earlier application, which undercounts repeats for both groups equally (name variants), so the AI-vs-overall comparison is fair even though both levels are floors.
• USPTO speedMedians from the complete prosecution-event history (247.9 million events): "first Office Action" is the earliest non-final-action event per application; "publication" is the published-for-opposition event; each metric is grouped by the half-year the event occurred. Filing→first-OA measures the Office queue; OA→publication includes the applicant's own response time; filing→registration is for marks whose registration issued in that half. Cohort sizes: ~173K–178K first Office Actions per half; ~88K–93K publications after an OA per half.
• Refusal mix & cited marksGleanMark's parsed Office-Action corpus (214,987 refusal-bearing OAs dated 1H 2026). Shares are of OUR corpus, not the USPTO total; grounds are parser-derived and advisory-only sections are excluded. Citation counts for the "blockers" list are distinct OAs in our corpus citing the registration under §2(d); registration families are consolidated where noted.
• Vocabulary & disclaimersClass 042 word ranks compare goods/services text of applications FILED in 1H 2021 vs 1H 2026 (same sampling, boilerplate removed). Disclaimer comparison is by RANK among disclaimed terms of applications filed in each period — ranks are comparable across periods even though older cohorts have accumulated more disclaimer entries.
• What we did not measureHow registered marks overcame each refusal type needs response-side outcome data our corpus does not yet support cleanly; it is on the roadmap for a future edition. A filing is intent, not a granted right. Chart data and term lists are available to reporters on request.