The 13 likelihood-of-confusion factors, and which this tool assesses

Each factor below is marked Assessed, Partly assessed, or Not assessed. "Assessed" means the tool measures evidence relevant to that factor — never that it decides the factor. Deciding likelihood of confusion is a legal judgment that weighs all thirteen together; this page is only about which raw material a name check can and cannot supply.

#	DuPont factor	Coverage here
1	Similarity of the marks — appearance, sound, meaning, commercial impression	Assessed — the core engine
2	Relatedness of the goods or services	Partly — the goods-class gate
3	Similarity of established, likely-to-continue trade channels	Not assessed
4	Conditions of purchase — impulse buy vs. careful, sophisticated buyer	Not assessed
5	Fame of the prior mark	Not assessed
6	Number and nature of similar marks in use on similar goods	Partly — shown, not scored
7	Nature and extent of any actual confusion	Not assessed
8	Length of concurrent use without evidence of actual confusion	Not assessed
9	Variety of goods a mark is used on (house mark / family of marks)	Not assessed
10	Market interface — consent agreements, assignments, prior dealings	Not assessed
11	Extent to which the applicant has a right to exclude others	Not assessed
12	Extent of potential confusion — de minimis or substantial	Not assessed
13	Any other established fact probative of the effect of use	Not assessed

One factor assessed, two partly, ten not. A screening score is a read on name similarity, not a likelihood-of-confusion determination.

This is the factor the tool is built around. Every candidate mark is compared to your name on the three axes that make up mark similarity for a word mark: sound (a phonetic signal), spelling (an orthographic signal), and meaning (a semantic signal). How those combine into the 0–100 score is documented in full on the methodology page.

Even here the coverage has edges. It reads word marks only: "appearance" is measured letter-by-letter, not as the look of a logo, stylization, color, or design. And "commercial impression" — the overall idea a mark conveys in the market — is only partly captured by the meaning signal. The tool measures how similar the names are. Whether that similarity rises to likely confusion is the legal question it does not answer.

Confusion only bites when the goods overlap, so the tool applies a goods-class gate: a data-mined matrix of relatedness across the 45 Nice classes, which dampens a strong name match in an unrelated class and keeps its full weight in the same or a related one. That is a real, measured read on this factor — but a partial one.

What it can't do: it works off Nice class relatedness, not your actual goods description or the specific way two products meet in the marketplace. And it only engages when you enter a class — with no class, overlap can't be assessed, so the gate stays neutral rather than assumed. An examiner reads the full identification of goods; the gate reads the class.

Every result shows how many similar names already exist in the register. That count is the raw evidence behind a crowded-field argument: a mark surrounded by many coexisting look- or sound-alikes may be legally "weak," so a new near-match matters less than it would in an empty field.

We surface that evidence but deliberately do not let field density move the score. We measured it, and a denser field did not reliably mean lower confusability — folding it into the number would have cost more accuracy than it bought. So this factor is shown, not scored: you can see the crowd, but the tool won't weigh it for you.

The remaining factors turn on marketplace, business, and legal facts that simply aren't in the public name record. A few have faint echoes in context shown elsewhere in the product — a prior owner's portfolio size (9) or their history of opposing others (5) — but that context is presented as public record and is never folded into the score. One line each on why the tool is blind to them:

• 3 · Trade channels — where and how the goods are actually sold. Not in the register.
• 4 · Conditions of purchase — whether buyers are casual or sophisticated, cheap impulse buy or careful purchase. A marketplace fact.
• 5 · Fame of the prior mark — fame is proven with sales and advertising evidence, and famous marks get a wider berth (and separate dilution protection). None of that is name data.
• 7 · Actual confusion — real-world evidence that consumers were actually confused. The strongest evidence there is, and it comes from the market, not a string comparison.
• 8 · Concurrent use without confusion — a track record of two marks coexisting. A historical fact about use, not names.
• 9 · Variety of goods / family of marks — whether an owner uses a mark across a family of products. A use fact.
• 10 · Market interface — consent agreements, assignments, and prior dealings between the parties. Private legal arrangements.
• 11 · Right to exclude — the strength and scope of the prior owner's rights.
• 12 · Extent of potential confusion — whether any confusion would be trivial or substantial.
• 13 · Any other probative fact — the catch-all for anything else that bears on the effect of use.

Because the tool measures one axis, its score can point the opposite way from a full §2(d) analysis — in both directions:

• A low similarity score is not clearance. A name you'd never call similar can still be blocked — most plainly by a famous prior mark (factor 5), which is protected well beyond close look-alikes.
• A high similarity score is not a refusal. Two similar names can lawfully coexist when other factors outweigh the resemblance — sophisticated buyers (4), separate trade channels (3), a consent agreement between the owners (10), or a long history of concurrent use without confusion (8).
• Actual confusion (7) — the most persuasive evidence of all — lives entirely in the marketplace, where no name comparison can reach it.

The bottom line: this tool tells you where a name sits on the single axis it can measure. A real likelihood-of-confusion analysis weighs all thirteen factors, several of which sit completely outside the public name record.

This page explains the framework the USPTO and the TTAB apply. It does not apply that framework to your name, your goods, or your situation, and nothing here is a legal opinion about any mark. Which factors matter in a given case, and how they balance, is a legal judgment — that is what a trademark attorney and a full clearance search are for.

This tool is preliminary automated screening built on public USPTO data: signals, not a verdict, and not legal advice. It is not affiliated with or endorsed by the USPTO. See the methodology for how the score is computed and the benchmarks for what we measured, including the negative results.