What is trademark infringement? The likelihood-of-confusion test

The most common misconception about trademark infringement is that it requires copying. It doesn't. A competitor who has never seen your brand, who chose their name independently, and whose spelling differs from yours can still infringe — because the legal question isn't did they copy you, it's would ordinary buyers be confused about who's behind the product.

That standard is called likelihood of confusion, and it's the centre of almost every US trademark dispute.

What has to be true

To make out infringement, an owner generally needs to show three things:

1. They own a valid mark with priority — they used it in commerce before the other party did (registration helps prove this, and extends it nationwide).
2. The other party is using the mark in commerce, in connection with goods or services.
3. That use creates a likelihood of confusion.

The first two are usually straightforward. The third is where the argument happens.

The factors courts actually weigh

There's no single formula. Federal courts apply multi-factor tests that vary by circuit — Polaroid in the Second, Sleekcraft in the Ninth — while the TTAB uses the DuPont factors. The labels differ; the substance is broadly consistent:

• Similarity of the marks — in appearance, sound, and commercial impression. Sound matters as much as spelling, which is why Kwikset and Quickset are a problem and why comparing letters alone misses real threats.
• Relatedness of the goods or services. Identical goods are the strongest signal, but the goods don't have to be identical — only related enough that buyers might assume a common source.
• Strength of the senior mark. Coined or arbitrary marks (KODAK, APPLE for computers) get wide protection. Descriptive marks get very little.
• Channels of trade. Same shelves, same search results, same distributors.
• Purchaser sophistication. Buyers of industrial equipment are held to more care than buyers of a $4 drink.
• Evidence of actual confusion. Powerful when it exists — and not required.
• The other party's intent. Bad faith hurts them; good faith doesn't save them.

No factor is decisive and they aren't scored equally. A very strong mark on closely related goods can win despite meaningful spelling differences; a weak descriptive mark can lose even against something quite close.

Things that are not required

Worth stating plainly, because each of these is assumed by someone every week:

• Identical marks. Nobody files a straight copy of a registered mark. Real disputes are about similar-but-not-identical.
• Identical goods. Related is enough.
• Actual confusion. Likelihood is the standard. Proof that someone was actually confused is strong evidence, not a prerequisite.
• Intent to copy. Independent creation is not a defence.
• A federal registration. Unregistered marks have rights too, from use — just narrower and harder to prove.

Dilution: a separate route for famous marks

If a mark is genuinely famous — nationally recognised by the general public, not just well known in its industry — its owner can act against uses that blur its distinctiveness or tarnish it, even with no confusion and no competition at all. That's why an unrelated business can't call itself Rolex.

The famousness bar is high and most marks don't clear it. For nearly everyone, confusion is the operative theory.

Why timing decides your options

If someone files an application for a confusingly similar mark, the cheapest moment to act is before it registers. Once an application clears examination it's published for opposition, which starts a 30-day window in which anyone who'd be damaged can oppose or request an extension.

That window is short and it doesn't reopen. Afterwards you still have options — a cancellation petition, a demand letter, litigation — but you're arguing against a registrant holding a certificate rather than an applicant hoping for one. The same dispute, considerably more expensive.

This is the entire practical argument for monitoring new filings continuously rather than checking now and then. A quarterly search will reliably miss a 30-day window.

What to do when you find something

Options run from cheapest to most severe: do nothing and document it; send a cease-and-desist; oppose the application; petition to cancel a registration; sue. Sometimes coexistence is the commercially sensible answer, particularly where the goods genuinely don't overlap.

What you shouldn't do is nothing, repeatedly and without a record. Trademark rights weaken when they aren't policed — a mark whose owner has tolerated years of similar uses is harder to enforce against the next one. Keeping a documented record of what you found and what you decided matters as much as the enforcement itself.

A note on similarity scores

Any tool that ranks conflicts — including ours — produces a similarity score. That number is a way of directing your attention, not a legal conclusion. Likelihood of confusion is a multi-factor legal test applied to specific facts about your market, your buyers and your goods. A high score means look at this one first, not you have a case.

This is general information, not legal advice. Whether a specific use infringes depends on facts that a qualified trademark attorney should assess.