Brand naming in 2026: AI namesakes, .com scarcity, trademarks

Naming a brand in 2026 is harder than it has been in any prior decade. Three structural pressures compound the problem: .com scarcity is now near-total in the 4-9 letter range, AI naming tools have produced tens of thousands of generically-named startups, and trademark thickets exist in every category as the prior wave of well-funded startups defends its marks. Founders routinely spend six to twelve weeks on naming and end up with a compromise they don’t love. The brands that get to a strong name in 2026 follow a different process than the one most founders use.

This post is the practical naming framework: why the 2020s naming playbook stopped working, the five-criteria filter that actually identifies strong candidates, and the 30-day process that gets a startup to a defensible name without the typical decision paralysis.

Why naming got harder in 2026

Three things happened in parallel between 2023 and 2026:

AI naming tools flooded the market with generic candidates. Namelix, Brandmark, ChatGPT, Claude — every founder now runs the same 50 prompts and surfaces the same kinds of names. The result: thousands of “Lumi” / “Volt” / “Nova” / “Synth” / “Forge” brands launched in the same 24-month window. The names look fine in isolation but are interchangeable in the market.
.com inventory in usable lengths is effectively gone. Five-letter .coms are extinct in any pronounceable form. Six-letter pronounceable .coms cost $10K-100K on the aftermarket. Seven-letter compound .coms are increasingly priced. Most founders now have to choose between an expensive .com, a worse domain (.co, .io, .ai), or a longer compound name.
Trademark thickets thickened. Every major category — fintech, productivity, AI, design tools, ecommerce, food, fitness — has 50-500+ active trademarks. Clearing a name through legal trademark search is now standard, and the search frequently surfaces blocking marks the founder did not know existed.

The naive 2020s approach — brainstorm 20 names, pick the favorite, register the .com — is now structurally broken. The names you can register are usually weak; the names you want are usually taken or trademarked.

The 5-criteria filter

A name in 2026 needs to clear five separate filters. Most candidates fail at least one. The brands that ship strong names systematically score candidates across all five.

1. Memorable phonetic shape

A strong name has a distinctive sound that’s easy to say once and remember twice. Indicators:

Two-to-three syllables, ideally
A consonant pattern that’s pronounceable in your target markets
Not phonetically similar to a known competitor
Easy to spell after hearing it once

This is the criterion AI tools handle reasonably well. The other four criteria are where AI-suggested names typically fail.

2. Domain availability (or affordable acquisition)

The honest position in 2026: you probably will not get a 4-7 letter .com for free, and you probably should not pay $50K+ for one as a pre-revenue startup. Acceptable alternatives:

A compound .com (tryacme.com, useacme.com, acmehq.com) for $10-50
A 7-9 letter .com that’s available
The .co (acceptable globally, especially for B2B/SaaS), .ai (acceptable for AI products), or .io (acceptable for developer products)
Buying the .com aftermarket if the brand has clear runway to justify the cost

Avoid: .net (reads as second-tier), country TLDs you have no connection to, hyphenated domains (reads as cheap), unusual TLDs (.app, .so, .one — fine if the brand is strong enough to carry it, weak if used to compensate for a weak name).

3. Trademark clearance

Run a USPTO TESS search for your candidate name in the relevant classes. The hard filter: any active mark in your category or related categories blocks the name unless the prior mark is clearly distinguishable.

For early-stage startups, the realistic process:

Self-search the USPTO database for your top 3-5 candidates
Search EU and key foreign jurisdictions where you plan to operate
Search common-law uses (Google search, Crunchbase, Product Hunt) for unregistered prior uses
Engage a trademark attorney for a formal clearance opinion before committing — usually $500-2,000

Skipping the trademark search and committing to a name produces the worst possible outcome: discovering the conflict after you’ve launched, when changing names costs you months and significant goodwill.

4. Search distinctiveness

A name that is impossible to find on Google is a name that compounds your customer acquisition cost for years. Test:

Search the candidate name. Are there established sites/brands using it for unrelated purposes?
Is there a Wikipedia entry for an unrelated meaning?
Is there a famous person with the same name?
Are there active social media accounts that would compete for the brand handle?

Names that share the candidate phonetic with multiple meaningful entities are search-distinctiveness disasters. Even if the trademark clears, you’ll spend years fighting for the SERP.

The strongest names in 2026 are either invented (Notion, Linear, Webflow — all coined in their original founder eras) or evocative pairings (Stripe, Figma, Plaid — common English words used in a novel category context).

5. International phonetic safety

In any market you plan to expand into, the name should:

Be pronounceable in the local language
Not sound like a slur, joke, or unrelated word
Not have negative cultural connotations

This filter eliminates more candidates than founders expect. A common Anglophone name can be offensive, awkward, or accidentally meaningful in Spanish, Mandarin, Japanese, or Arabic. Native-speaker verification is the only reliable check; AI translation tools miss the cultural/colloquial layer regularly.

The 30-day naming process that works

For a founder starting from zero:

Week 1: Generate volume

The goal is volume, not quality. Generate 100-300 candidate names through varied methods:

AI generation with structured prompts (positioning, category, target customer, vibe)
Word combinations: noun + noun, adjective + noun, verb + noun
Suffixes and prefixes: -ly, -ify, -er, -io
Borrowed words from other languages, mythology, science, geography
Acronyms expanded into pronounceable forms

Do not filter during generation. Filter happens in week 2.

Week 2: Apply the five-criteria filter

Score each candidate against the five criteria above. Use a binary pass/fail per criterion. Names that pass all five make it to the shortlist; names that fail any criterion drop out.

For a typical 200-candidate generation, expect 5-20 names to clear all five criteria. This is the realistic yield in 2026 — most candidates fail at trademark clearance or .com availability before personal preference even enters the equation.

Week 3: User test the shortlist

The shortlist of 5-20 names gets tested with real humans in your target market. Tests that produce useful signal:

“If you saw this brand for the first time, what would you guess they do?” — tests intuition fit
“How would you spell this if you heard it on a podcast?” — tests phonetic legibility
“What kind of company is this — premium, mass, technical, friendly?” — tests connotation
“Have you heard of this name before?” — tests novelty and conflict risk

Test with 10-20 real people from your target customer demographic, not your founder peer group. The signal sharpens fast.

Week 4: Commit and protect

Pick a name. Register the domain. File the trademark in your home jurisdiction. Reserve social handles. Build a minimal brand identity around it. Stop second-guessing.

The single biggest naming mistake in 2026 is not picking the wrong name — it is staying in the naming process too long. Founders who stretch naming past four weeks routinely converge on weaker names than the candidates they had at week three.

For the broader brand foundation work, see the startup brand checklist — naming is one piece of a larger brand system. The connection to brand voice in the AI era matters too: a strong name is the first signal of a brand’s voice and posture.

What strong 2026 names actually look like

Patterns we see in brands that named well in 2024-2026:

Coined evocative names — Cursor, Linear, Granola, Cognition. Real English or quasi-English words used in unexpected category contexts.
Founder-name brands — Anthropic, Mistral, Perplexity. Strong intellectual register, hard to copy, often coined to evoke specific concepts.
Pair compounds — Notion, Stripe, Plaid. Short, distinctive, evocative without being literal.
Founder names that double as brand names — Vercel, Replit. Founders’ personal authority transfers to the brand.

What we don’t see in strong 2026 names:

AI-generated -ly names (Quickly, Smartly, Easily) — saturated and undifferentiated
Tech-buzzword compounds (SyncAI, FlowOps, DataX) — read as bot-generated
Vowel-stuffed Silicon Valley clichés (Foovo, Glooba, Lumica) — feel dated

FAQ

Should I use AI to generate names? For volume, yes. For final selection, no. AI is a generation tool, not a selection tool. The five-criteria filter has to be applied by humans.

How much should I spend on a domain? Pre-revenue startup: $0-200. Funded seed-stage: up to $5K. Series A and beyond: up to 1-2% of cash on hand. Spending $50K on a domain pre-revenue is almost always a mistake.

Can I name a brand the same as a famous one outside my category? Sometimes. Apple (computers) and Apple (records) coexisted for decades. But the risk is real, and the search-distinctiveness penalty is severe. Avoid where possible.

Do I need a trademark before launching? File the trademark application within the first 1-3 months of operating under the name. The “intent to use” filing protects the name even before commercial use. Operating under a name for 12+ months without filing creates risk that someone else files first.

What about renaming? Renaming a brand is expensive and risky — see when to rebrand for the framework. The right time to invest in naming is at the start; renaming should be a last resort.

How do I evaluate AI-suggested names without falling for hallucinated trademark / availability info? AI tools regularly hallucinate domain availability and trademark status. Never trust the AI’s claim. Manually verify every shortlist candidate at the registrar and on the USPTO database.

The honest 2026 framing

Brand naming in 2026 is a structurally harder problem than it was in 2018, and most founders are unequipped for it. The framework above does not make naming easy — it makes it tractable. Generate volume. Apply the five-criteria filter rigorously. Test with real humans. Commit and protect.

Most founders who follow this process get to a name they love within 30 days. The ones who skip the trademark and domain filters early often spend 60-90 days and end up with a weaker name. The discipline is in the filtering, not the generation.