No black box. Here is exactly how MCInvent computes a registrability score, and every rule maps to a real ground the USPTO refuses marks on. It is a transparent heuristic that surfaces known refusal grounds — not a machine-learning prediction, and not a substitute for an attorney's clearance.
Trademark strength runs along the settled spectrum: fanciful → arbitrary → suggestive → descriptive → generic. We classify each element of the mark against the goods you enter and score accordingly:
A distinctive mark can still be refused if it's confusingly similar to a prior mark in the same or a related class. We search your mark against a trademark index (a baseline of well-known marks by default; the full USPTO corpus when enabled) using sight/sound/meaning-style similarity, and lower the score for close matches in related classes. Nothing reads "clear" on a partial index — that's stated plainly in every result.
To separate marks of similar type, small adjustments reward distinctiveness texture — uncommon letters, length, letter variety. These are labeled here as heuristic: they nudge, they don't drive, and they carry no legal weight of their own.
"Strong — highly distinctive", "Moderate", "Risky", "Likely refused" describe the strength and risk of the mark. They are not a prediction that the USPTO will or won't register it — that decision turns on a full examination and clearance we don't perform.
MCInvent · Millennials Creatives LLC. This is an analysis and preparation tool, not legal advice, and not a law firm.