Vision paper · Paris · April 2026

The Naming Revolution

Why the internet's address book is being rewritten — and what AI is doing to domains, brands, and creators.

A vision paper by Cyril Frémont · CEO — internet.Best · domains.Best · web.Best · ai.Best

The third era of naming has begun

The internet has lived through two naming eras. We are now entering the third, and it is the most consequential of the three.

The first was the Directory Era — roughly 1985 to 2005. You typed an address and arrived at a destination. .com became a reflex because human memory was the discovery layer. Domain equals door.

The second was the Handle Era, born inside social platforms. Discovery migrated to feeds, algorithms, and hashtags. Your @name on Twitter, Instagram, TikTok, LinkedIn mattered more than your domain. The address bar emptied. Platforms became the directory.

We are now in the Agent Era. An AI intermediary — chatbot, voice assistant, agentic browser, autonomous system — sits between the user and the web. It reads, summarizes, recommends, books, buys, and increasingly chooses on behalf of humans. Discovery is no longer typed or swiped. It is prompted.

That single shift rewrites what a name is for.

Era I · 1985–2005
Directory
Discovery layer
Human memory. You typed an address and arrived.
Domain = door
Era II · 2005–2022
Handle
Discovery layer
Feeds, algorithms, hashtags. The address bar emptied.
@name on the platform
Era III · 2022 →
Agent
Discovery layer
An AI intermediary reads, recommends, and chooses.
Prompted, not typed
Where discovery happens
Three eras, one re-sorting
Illustrative share of discovery by channel across the three naming eras

What AI actually does to naming

Three shifts matter.

Discovery becomes probabilistic. When users asked Google, the answer was a ranked list. When users ask Claude, ChatGPT, or Gemini, the answer is a synthesis — and inside that synthesis, a handful of names get mentioned. The distance between "mentioned by the model" and "invisible" is the new distance between existence and non-existence.

Agents need semantic anchors. An AI assistant booking a hotel, buying a ticket, or deploying a workflow does not care about your clever pun. It cares whether your name describes what you do with enough clarity for a model to resolve it without ambiguity. A domain that means something becomes machine-readable real estate. A domain that means nothing becomes friction.

Memorability goes verbal. Voice interfaces, chat interfaces, and dictation have moved the internet's primary input from fingertips to vocal cords. Names that are phonetically clean, spellable on the first try, and category-obvious now compound faster than clever ones.

The consequence: the naming layer of the internet is being re-sorted. The names that survive the Agent Era are not necessarily the ones that won the Directory Era.

The .ai precedent

If you want to see the Agent Era's impact on naming before it becomes conventional wisdom, look at .ai. In 1995, the ITU assigned it to Anguilla, a 15,000-person British Overseas Territory in the Caribbean. For twenty-seven years it was a quiet ccTLD with a few thousand registrations. Then ChatGPT launched in November 2022, and the string came alive.

0×
Registrations growth in under six years
0M
Names registered — crossed in January 2026
0%
Of new tech startups chose .ai in H1 2025 (7% in 2021)
$0K
Top single-name sale — you.ai

The trajectory is unambiguous. Registrations went from roughly 144,000 at the end of 2022 to 354,000 a year later, and crossed the one-million mark in January 2026 — a twenty-fold increase in under six years.

.ai registrations
A string that came alive
Active .ai registrations, thousands · 2026 = January

For Anguilla, the fiscal impact was civilizational. .ai revenue rose from $2.9M in 2018 to more than $100M in 2024, with 2025 run-rates pointing toward another record year. Domain registrations now contribute roughly 20% of the island's government revenue, up from around 5% before the AI boom.

Anguilla registry revenue
More value than decades of tourism
Government .ai registry revenue, USD millions

The secondary market tells the same story. Single .ai names have sold at auction for $405,000 and, in the case of you.ai, $700,000. Google reclassified .ai from a country-code to an effective gTLD for ranking purposes in May 2023, and the Government of Anguilla transferred operational control to Identity Digital in January 2025, professionalizing the registry for long-term scale.

Strip away the Caribbean-windfall framing and a clean economic signal remains. That is not a tourism story. It is a semantics story.

The lesson for the next cycle is not "go buy .ai names" — the premium inventory is already taken, priced, and earning renewals. The lesson is why it happened. .ai worked because the extension carried meaning that both humans and machines could resolve without ambiguity. It worked because when an AI startup writes its name, what comes after the dot finally costs less attention than what comes before. It worked because the naming layer rewards strings that look like the category they serve.

.ai is the first TLD to price the Agent Era correctly. It will not be the last.

Superlatives win in AI mediation

Here is a practical observation from running an ICANN-accredited registry since 2018. Generic .com names are abundant but semantically weak. Descriptive gTLDs — .hotels, .agent, .ai, .best — carry meaning at the extension itself. In an AI-mediated world, that meaning is not decoration. It is structural signal.

.best is a superlative. It is the only TLD that carries category leadership in its extension. When an AI agent resolves paris.best or hotels.best, it is not pulling a random string — it is pulling a commitment. That commitment becomes the unit of comparison the machine can reason about.

This is why I do not believe the next decade of domain value creation will come from stockpiling .com refugees. It will come from owning the semantic infrastructure of an AI-first web.

Three audiences, three opportunities

For domainers

The valuation logic is mutating. For two decades, domain value was driven by typing traffic, SEO advantage, and brand-match acquisition fees. All three are being compressed by AI mediation.

What is rising instead: agent-addressable assets (domains that AI systems can use as canonical endpoints for a category, a workflow, or a capability); semantic premium (names whose meaning is unambiguous inside a model's training distribution); and superlative verticalshotels.best, restaurants.best, agents.best, prompts.best — category leadership expressed at the naming layer.

The portfolios that outperform in the next cycle will be the ones rebalanced toward meaning, not toward legacy extensions. Domainers who treat their inventory as AI-readable infrastructure, and price it accordingly, will be ahead of the curve by the time the rest of the industry notices the shift.

For brands

Brand identity has been visual for forty years. The logo, the palette, the typeface. In the Agent Era, verbal identity becomes equally critical — because the first surface of your brand is often a line of text inside a chatbot's answer.

Three implications. Category ownership through naming: a brand that sounds like its category survives synthesis; a brand that relies on visual recognition alone disappears when the visual layer is stripped by an agent. Prompt-resistant positioning: when users ask "what's the best X?", brands whose names align with the question phrase surface first — naming becomes the SEO replacement, but for models instead of search engines. Verbal identity as infrastructure, not as marketing — meaning the thing without which everything downstream is slower, more expensive, and less defensible.

Brands that treat domains as corporate hygiene are about to relearn the lesson travel brands learned when Booking.com ate their direct distribution: the naming layer is not neutral ground.

For AI enthusiasts, creators, and builders

This is where it gets exciting, and where ai.best comes in.

The AI creator economy — prompt engineers, agent builders, workflow designers, skill authors — does not yet have a native distribution layer. GitHub is for code. Hugging Face is for models. Gumroad is for PDFs. Nobody has built the canonical marketplace for the objects AI enthusiasts actually create every day: prompts, agents, skills, workflows.

ai.best is being built as that home — all the best AI models in one subscription, for less than the price of one, with a marketplace for prompts, agents, skills and workflows sitting right inside the platform. You run every model in one place, and you sell what you make in the same place. The marketplace economics are deliberately tilted toward the creator — a flat commission of 10% + €1, Stripe fees absorbed by the platform — because without creators, there is no marketplace worth naming.

Why does the naming layer matter here? Because a prompt without packaging is a file. A prompt published under a credible, categorical domain is a product. Same text, same work, same output — but different distribution and different economics depending on the surface it lives on.

For AI enthusiasts that means two things. Your work has a home that isn't a screenshot on X. And your reputation compounds inside a namespace designed for it.

The web.best angle

There is a second layer to this vision that does not get talked about enough.

AI is changing not just how people find sites, but what a site is. Static pages optimized for Googlebot are yesterday's artifact. The sites that will win the next cycle are cinematic, video-first, and built for the attention patterns of TikTok-native audiences and AI-summarizing agents alike.

web.best is the infrastructure for that shift: a website builder that starts from video, not from a text template. Because when an AI agent surfaces your site and a human lands on it seconds later from that recommendation, the element that converts is not your copy. It is the moving image that communicates your category leadership in three seconds.

.best is DNS infrastructure. web.best is video infrastructure. ai.best is model infrastructure — every leading AI behind one login. These are not three products. They are three layers of the same thesis:

In an AI-mediated internet, infrastructure that carries meaning is immunity against commoditization.

What happens next

Three predictions for the next 24 months.

The ICANN 2026 new gTLD round will be the largest naming redistribution since .com itself was introduced. New strings will be delegated across AI, creator, and vertical clusters. Brands and domainers that do not participate in some form will be on the wrong side of the semantic shift.

Agent-native domains will develop their own pricing curve. A name that an AI agent resolves cleanly for a high-intent query will be worth multiples of a name that merely looks good on a business card.

Superlative and vertical TLDs will outperform generic ones in relative growth. Not because .com disappears — it won't. But because the marginal dollar of domain investment will move toward meaning.

The call

If you are a domainer, rebalance for meaning. The inventory that pays off in 2028 is being assembled now.

If you are a brand, audit your verbal identity the way you audit your visual one. You are about to be rendered by machines more often than by humans.

If you are a creator or AI enthusiast, stop publishing your best work into feeds that own your audience. ai.best is opening precisely because this gap needed to be filled.

The internet is being rewritten at the naming layer. This is a window, not a trend. Windows close.

We built .best for this decade. We are building web.best and ai.best for this window.

Cyril Frémont, CEO of internet.Best
Cyril Frémont
CEO — internet.Best
domains.Best · web.Best · ai.Best
Paris · April 2026