AI-Era Naming: Which Words Are Gaining Value, Which Are Losing It

How the AI boom reshaped domain demand, from .ai price surges to saturated GPT and agent keywords, and which naming patterns look durable versus bubble-shaped. Research current as of 12 November 2026.

The short answer

The AI boom did not make "AI-related" names valuable. It made a handful of specific things valuable, and mislabeled everything near them.

The evidence in 2026 points to a split. Scarce, short, real-word names (the single-word .com and the single-word .ai) have repriced upward, in some cases spectacularly. Meanwhile, generic AI vocabulary is cooling as a share of new registrations, trademark-adjacent terms like GPT have become crowded and risky, and event-driven coinages spike and fade within days. The durable pattern is scarcity plus brandability. The bubble pattern is a literal keyword that only makes sense while one technology's current label is fashionable.

How the boom reshaped demand

The .ai story

.ai is the clearest case of a technology boom repricing a namespace. Registrations grew from roughly 148,000 in late 2022, when ChatGPT launched, to over one million in January 2026 (2026 industry trend reports). Anguilla's government earned about EC$230 million (around US$85 million) from .ai in 2025, and reported roughly US$80 million in the first seven months of 2026 alone. Officials and the IMF have cautioned that this share of government revenue is likely to stabilize lower as the initial rush slows.

The registry also moved prices up in March 2026: the wholesale fee reportedly rose to $160 for a mandatory two-year term, up from $140, so two-year retail prices typically start around $160 and run higher, since registrars set their own margins on top. A reported renewal rate near 90% suggests most holders treat the cost as a branding expense rather than a speculative gamble, though that figure comes from secondary sources and should be verified against registry data.

Adoption among new companies is real. One analysis of startup funding data found nearly a quarter of pre-seed companies funded since January 2025 use .ai, while later-stage companies mostly built their brands on .com before .ai was fashionable. Established leaders still tend to hold .com: OpenAI's brand lives on openai.com and chatgpt.com, and it acquired chat.com.

The aftermarket split

The aftermarket tells a more nuanced story than the registration numbers. According to DNJournal's year-to-date chart through early July 2026, the top three public sales of the year were all .com names: AI.com at $70 million (a 2025 deal whose price was disclosed in February 2026), Club.com at $10 million and Green.com at $7.5 million. Premium .com still owns the top of the market.

But .ai has become the strongest challenger. Bot.ai sold for $1.2 million in February, and Free.ai reached $350,000 in April. Over the summer, .ai names such as OF.ai ($250,000) and Perform.ai ($200,000) filled the top of the country-code chart. Two things stand out in these sales: they are overwhelmingly single dictionary words, and they were sold through professional marketplaces. Domain investors consistently describe the same lesson, that the boom lifted good names a lot and mediocre ones barely at all.

One report circulating in September claimed average prices for exact-English-word .ai names now slightly exceed those for .com. Treat that carefully: it comes from a secondary summary, and averages in thin, skewed markets can be driven by a few outliers.

Saturation: AI, GPT and agent

"AI" as a keyword

.ai is the clearest case of a technology boom repricing a namespace. Registrations grew from roughly 148,000 in late 2022, when ChatGPT launched, to over one million in January 2026 (2026 industry trend reports). Anguilla's government earned about EC$230 million (around US$85 million) from .ai in 2025, and reported roughly US$80 million in the first seven months of 2026 alone. Officials and the IMF have cautioned that this share of government revenue is likely to stabilize lower as the initial rush slows.

The registry also moved prices up in March 2026: the wholesale fee reportedly rose to $160 for a mandatory two-year term, up from $140, so two-year retail prices typically start around $160 and run higher, since registrars set their own margins on top. A reported renewal rate near 90% suggests most holders treat the cost as a branding expense rather than a speculative gamble, though that figure comes from secondary sources and should be verified against registry data.

Adoption among new companies is real. One analysis of startup funding data found nearly a quarter of pre-seed companies funded since January 2025 use .ai, while later-stage companies mostly built their brands on .com before .ai was fashionable. Established leaders still tend to hold .com: OpenAI's brand lives on openai.com and chatgpt.com, and it acquired chat.com.

"GPT" as a keyword

GPT shows what happens when a technical term becomes a brand-adjacent magnet.

• Legal ambiguity. OpenAI's attempts to register GPT as a trademark in the US were rejected on the grounds that the term describes a type of technology, and hundreds of products adopted it. OpenAI did secure an EU trademark. That leaves a messy position: the term is hard to own, yet owners of ChatGPT-style names face enforcement risk.
   
• Abuse concentration. Palo Alto Networks' Unit 42 found that GenAI-themed new domains are suspicious far more often than ordinary new registrations, and gpt was the single most abused keyword, with a 76% suspicious rate. Registrars and investors should expect scrutiny, takedowns and reputation drag on names carrying it.
   
• Escalation to the namespace level. OpenAI applied in ICANN's 2026 round for 15 extensions, reportedly including .gpt, .chatgpt and .agi. None is approved, and other applicants are contesting some of them.

"Agent" and "agentic" as keywords

Agent is the interesting case, because the vocabulary is still spreading. In the same analysis, registrations containing agent grew 14.6% and agentic 16.7% in the second half, well ahead of the AI theme as a whole. In ICANN's 2026 round, .agent is the most contested string, with 13 applicants including Google, Meta and OpenAI. Aftermarket buyers have paid real money too: AgenticIntelligence.com reportedly sold for $150,000 in May.

But the same data carries a warning. In the 2026 .com analysis, agent names were listed for sale at 10.8%, nearly five times the market baseline, and far higher than the broader ai keyword. The narrower and more fashionable the keyword, the higher the speculative share. The same study documented a sharper version of this when OpenClaw rebranded on 30 January 2026: claw domains went from near zero to hundreds of registrations a day within hours, with a for-sale rate over five times the baseline.

Scorecard tally

Durable patterns

1. Scarcity of the string itself. AI.com selling for $70 million is the outlier, but it fits a longer pattern: the shortest, most memorable names reprice with each technology wave, whatever the wave is. The 2026 premium-domain data from registries also points to rising demand for short, brand-defining names and healthy renewals.
    
2. Coined or ordinary words that stand on their own. Startups across tech increasingly describe a return to real words and coined brands over labels like AI. By this view, the best AI companies do not need the label. The names that last are those that still read as a brand if the category word changes.
    
3. Names tied to something people actually do. The fastest-growing registration themes were home, energy and creative, not AI. People name what they are building. That is a reminder that demand follows use, and that a name anchored to a concrete noun (a city, a trade, an object) carries value that does not depend on a technology narrative.
    
4. Category nouns that become infrastructure vocabulary. Chat and agent have a shot at being durable in the way app and cloud were, because they describe a function, not a product version.

Bubble patterns

1. Literal, stacked keywords. Commentary from domain investors on their biggest losses names ultra-literal chains built around prompt engineering and similar roles, on the assumption that every profession would launch an AI tool under a descriptive name. Real startups, the argument goes, want short, flexible names, and hundreds of weak .ai names with expensive renewals are a carrying-cost problem. This is investor commentary rather than a dataset, so weigh it accordingly.
    
2. Version, product and trademark terms. Anything tied to another company's product or a specific model generation is exposed to a single decision by that company: a rebrand, a trademark action or a new product name. GPT, ChatGPT and Sora-style names fit here.
    
3. Event-driven coinages. Claw names after a rebrand, superintelligence-themed names, and the .si registration surge in September all show the same shape: a burst of registrations within hours, mostly speculative, mostly unrelated to what anyone is building. In the .si case, only about 3% of one registrar's new names even contained an AI term. That points to buyers hedging on a label, not building products.
    
4. Extension-as-signal. .ai carries real value today, but it has structural exposures. It is a single country's registry with unilateral pricing, a two-year minimum, and a repricing already in 2026. The IMF itself expects the revenue boom to level off. If AI becomes the default feature of every product, a label that says AI-first may signal less. That is a risk to watch, not a prediction.
    

A stress test for any AI-era name

Before paying a premium, ask:

1. If the industry renames this technology in three years, does the name still work?
    
2. Is the name a brand word or a description? Descriptions compete with everyone; brands do not.
    
3. Does it depend on another company's trademark or model name?
    
4. Would a real company, not just a flipper, plausibly use it as its primary brand?
    
5. Can the holding cost (especially .ai's two-year minimum and renewals) be carried for years if there is no buyer?
    
6. Is the buying signal an organic trend, or a single announcement driving a burst of registrations?

What this means for domain investors

• Sell scarcity, not keywords. The aftermarket data rewards single-word, brandable inventory. Merchandising and search should surface short real words and coined brands ahead of keyword chains.
   
• Be careful with GPT-class terms. Higher abuse rates and trademark exposure argue for tighter screening and clear messaging on names that borrow another company's marks.
   
• Prepare for new extensions. ICANN's 2026 round drew 1,615 applications. The application window ran from 30 April to 12 August, and most new extensions are not expected to reach general availability until 2027 or 2028 if the 2012 timeline is a guide. .agent contention will take time to resolve. Pre-launch interest lists are a way to measure demand without promising availability.
   
• Watch signals beyond registrations. Resale-listing rates, renewal behavior and active-DNS rates separate real use from speculation better than raw volume does.
   
• Expect event spikes. The .si and claw episodes suggest that a single news event can move registration volume by orders of magnitude in a day. Pricing, fraud screening and capacity planning should assume this can happen again.

Caveats

• The registration-trend analysis covers one registrar's customers and uses keyword matching, so it does not represent the whole market, and its AI theme counts names by keyword rather than TLD.
   
• Public sales charts only capture reported transactions, which skew toward high prices. They do not measure the median domain.
   
• Several claims here are inferences, flagged as such: the GPT value decline, the durability of agent, and the long-run effect of .ai's pricing.
   
• Prices, counts and ICANN timelines change quickly. Check current sources before using any figure in a decision.