Trademark + domain verification for AI startups, SaaS founders, and dev-tool builders.

Across 4,080 unique LLM topic clusters representing 43.1 million monthly queries (Semrush Prompt Research, January 2026), the Tech / SaaS / AI vertical accounts for 14.5 million queries — 33% of the entire corpus, with a 28% buying-intent score and 1.75× year-over-year growth. It is the largest vertical, the highest-growth vertical, and the vertical with the founder-shaped buying behavior. Every other vertical we measure is one-third the size or smaller and none grows faster.

The implication for founders: the namespace is crowded and getting more crowded every week. The same LLMs that founders use to brainstorm names are training on the same English-language web that already contains the trademark filings; the candidates an LLM returns are systematically biased toward the saturated register. The verification step is no longer optional for any AI / SaaS / dev-tool brand that intends to ship and scale.

The full demand intelligence breakdown — the five-axis test, the founder journey sizing, the competitor landscape — is in the underlying strategy memo (§12). The TL;DR is that trademark and domain are the demand-side lead, cultural / sound / pronunciation are the moat depth, and Tech / SaaS / AI is the vertical that funds the build-out.

The Bard problem. When a major LLM company launches a consumer product under a dictionary noun, the senior users of that mark — independent musicians, indie game studios, storytelling apps — get crowded in search and app stores. A founder without an AmLaw-100 legal team to absorb the friction cannot use dictionary nouns; the register is famously crowded. The reflexive lesson is pick a coined word, not a dictionary word.

AI-prefix saturation. USPTO Class 9 has more than 12,000 live and pending registrations beginning with the “Ai-” phoneme as of May 2026; Class 42 adds another 8,000. There are fewer than 50 distinct onset-and-vowel clusters under that prefix. Any new “Ai-prefix” candidate is almost certainly in a crowded phonetic neighborhood. The escape is to stop signaling “AI” in the brand token — Cursor, Anthropic, Mistral, Replit, Vercel all signal it through their product and copy, not their name.

Famous-mark families. The Federal Trademark Dilution Act protects famous marks against junior uses even when the goods don't overlap. Claude, OpenAI, Anthropic, Mistral, Gemini, and Cohere are all either famous or trending famous. The families matter too — Claude / Claude Code / Claude Desktop / Claude Sonnet — and the family rule (McDonald's family of “Mc-” precedent) treats junior uses adjacent to any family member as at risk.

For the full walkthrough of the five-step founder-side clearance checklist that addresses each of these failure modes, see the spear piece at /content/naming-ai-startup-trademark-clearance.