Refusals 101, Part 2: Section 2(d) — The Register's Biggest Fight

Pick a name. Build a logo. Print the business cards. Then an examining attorney at the U.S. Patent and Trademark Office tells you someone already owns something close enough that customers might mix you up — and your application dies.

That refusal has a number: Section 2(d). By the USPTO's own description, likelihood of confusion is the most common reason it refuses to register a trademark — and it is a "substantive" refusal, meaning it's about the mark itself, not a paperwork fix like a mistyped goods description. Part 1 of this series drew that line between fixable paperwork and the refusals that decide whether you keep your name; 2(d) is the biggest of the real fights. Founders should care because it's the single most preventable naming mistake, and it's preventable before you spend a dollar on branding. Junior practitioners should care because reading how examiners apply it is the fastest way to build the instinct that clients pay for.

What Section 2(d) actually forbids

The rule is short. The USPTO will refuse to register a mark that so resembles a mark already registered that using both would be likely to cause confusion, mistake, or deception among consumers. An earlier-filed application that's still pending can block you too, just on a delay: the examiner suspends your application while the earlier one works through the system, and if it registers, the 2(d) refusal follows.

Read that carefully, because two words do a lot of quiet work.

"Resembles" — not "is identical to." The examiner is not looking for a copy. Marks are compared as consumers meet them: in appearance, in sound, in meaning, and in overall commercial impression. A respelling that sounds the same as an existing mark — LYFT against LIFT, say — can be found confusingly similar even though no letter sequence matches. QREW (serial 97351996) is the kind of coined spelling that has to clear the sound-alike hurdle against ordinary words like CREW — and it did: it's now a live registration.

"Likely" — not "certain," and not "someone was actually confused." No one needs to have been fooled yet. The examiner is predicting future confusion in the marketplace, which is why so much of a 2(d) fight is argument rather than evidence.

The register is enormous, and that's the problem

The reason 2(d) refusals are so common is arithmetic. There are roughly 4.6 million live trademark records in the federal system right now — registrations plus still-pending applications — and every one of them is a potential obstacle to your new name. "Live" means the mark is registered or still moving through examination; a live registration can be cited against you today, and a live earlier-filed application can be cited the day it registers. Every name you invent is competing for space in a field that's already crowded to the horizon.

This is why the cleverest available name and the registrable name are often not the same name. The plain descriptive words are taken. The obvious puns are taken. What's left is either genuinely distinctive or already occupied by someone who got there first — and "got there first" is the whole ballgame under 2(d).

What examiners weigh: the du Pont factors

Examiners don't decide "likely confusion" by gut feel. They work through a checklist that comes from a 1973 appeals-court case, In re E.I. du Pont de Nemours & Co. — the case that gave us the du Pont factors, the standard menu of considerations for weighing confusion. There are thirteen of them, but two carry most of the weight in practice.

The first is how similar the marks are — in appearance, in sound, in meaning, and in overall commercial impression. Not dissected letter by letter, but taken as a whole, the way a distracted shopper would take them in.

The second is how related the goods or services are. This is where classes matter. The USPTO sorts every product and service into 45 numbered classes — categories like class 25 for clothing or class 33 for alcoholic drinks other than beer. A common misunderstanding among founders is that a different class means safety. It doesn't. Goods in different classes can still be found related if consumers would expect them from the same source, and closely related goods inside overlapping trade channels are exactly where 2(d) bites hardest.

The remaining du Pont factors — the strength of the earlier mark, whether buyers are careful or impulsive, whether the two brands have coexisted without confusion, and more — each get their own treatment in our du Pont factor series, which walks through them one at a time.

A refusal is not a death sentence

Here is the part the printed-business-cards panic misses: a Section 2(d) refusal arrives in an office action — the examiner's official letter explaining what's wrong — and an office action is a conversation, not a verdict. You get to respond. You can argue the marks aren't as close as the examiner thinks, that the goods travel in different channels, that the earlier mark is weak because dozens of similar marks already coexist.

Sometimes that argument works even after the examiner digs in. The USPTO's timestamped prosecution logs — the public record of what happened during examination — show applications that drew a final refusal, the examiner's "I've heard you and I'm not moving" letter, and still walked away registered. Here are five where the final refusal was a 2(d), taken straight from the office actions in their public files.

MIXIBLE (serial 97300288), a streaming-media mark, drew a final Section 2(d) refusal in January 2023, answered it with a request for reconsideration, and registered on the Principal Register in June 2024 as registration 7409344. ZOO (serial 97300330) is the harder case: a one-word mark for audio-and-video-production training that sat suspended for over a year, then drew a final refusal in August 2024 citing three earlier ZOO-family registrations, including ZOOMEDIA — and still registered in April 2025 (registration 7749603). One of the three marks cited against it, ZOOMEDIA, has itself since gone dead on the register. GFX LABS (serial 97300421), C & C CARWORX (serial 97300500), and Z ZELOS (serial 97300640) each carried a final 2(d) refusal in their files and registered anyway.

These were found by searching the public event feed for applications that logged a final-refusal event and nevertheless issued as registrations, then reading the office actions in each file to confirm the refusal was a 2(d) — a check any reader can reproduce against USPTO records. Two honest caveats. First, the event log alone doesn't say why an examiner went final; you have to open the letters, and many final refusals that get beaten turn out to be descriptiveness fights or paperwork disputes, not 2(d). Second, not every save is a full win: CAMERA TO CLOUD (serial 97300518) drew a final refusal in 2023 and landed on the Supplemental Register — a secondary register with real but weaker protections — rather than the Principal Register.

What the pattern teaches is not that final refusals are easy to beat. It's that "final" is a term of art, not a wall. A well-built response, or an appeal, or a consent agreement with the owner of the cited mark can move a refusal that looked immovable.

What to actually do with this

Search before you fall in love. The most expensive 2(d) problems are the ones discovered after the brand is built, because by then the sunk cost pressures you toward the losing fight instead of the free rename. A knockout search — a quick pass over the live register for close marks in related goods — costs almost nothing compared to rebranding a launched company. If you've never run one, start with a basic availability check before you commit to anything.

Think in sound and meaning, not just spelling. If your clever respelling reads or sounds like an existing mark, the coined letters won't save you. Examiners read marks the way customers hear them.

Don't assume a different class is a moat. Ask instead whether a normal buyer would expect your product and the older one to come from the same company. If the honest answer is "maybe," treat the older mark as a real obstacle.

And if a refusal lands, read it as the opening of a negotiation. The marks above prove the point: an examiner's "no," even a final one, is a position you can answer — sometimes successfully.

The register has rules that quietly decide who keeps their name. Section 2(d) is the loudest of them. Knowing how it works before you name anything is the cheapest brand protection there is, and knowing how to answer it is a skill an early-career practitioner can build a career on.

---

GleanMark's search tools run the same kind of knockout check an examiner runs — surfacing close marks in related goods before you commit to a name — and its monitoring flags new prosecution events on marks you track twice each business day. This analysis is based on public USPTO records and is not legal advice.