The registrar used to process instructions. Now it gives them. Every other asset market met that shift with a wall of disclosure. Domains have no wall yet.
Seventy-four domains cleared above $70,000 in a single quarter. Not one of the ten largest sold through GoDaddy or Afternic — the counters where most owners keep their names. The largest transaction GoDaddy reported in that sample ranked twenty-fourth. That is not a soft quarter. It is the top of a market routing around its own default intermediary.
For forty years the registrar stood behind a counter. You chose the name. You set the price. You picked the buyer, the broker, the marketplace. The registrar processed the instruction and took the fee.
That counter is gone.
The next registrar is in the room before the domain exists. It proposes the name, ranks the alternatives, registers it, configures the DNS, connects the email, publishes the site, measures the crawlers, solicits the buyers, estimates the value, and advises the owner whether to sell. This is not a forecast — an AI-native registrar already advertises exactly that, and states the ambition plainly: the conversation is the registrar. So the governing question is no longer which registrar renews cheapest. It is: whose benefit is the machine optimizing when it tells the owner what to do?
A registrar earns when a name is registered, renewed, parked, listed, financed, transferred, or sold through an affiliated marketplace. Put an AI agent on top of that revenue and the agent inherits an objective function: more registrations, more renewals, more inventory, faster velocity, higher commissions. Not one of those goals is illegitimate. Not one of them is identical to maximizing the owner's outcome.
In securities markets, an institution that held the asset, appraised it, recommended the trade, made the market, and took the other side would face disclosure at every step. In domains, that same stack is sold as convenience. The industry never separated custody from execution, research from brokerage, market-making from advice — because it treated domains as a technical subscription. They are becoming property. The separations did not arrive with them.
An owner holds 2,000 names — some excellent, some speculative, some commercially dead. The registrar's agent reviews the book and recommends. If it says renew everything, it can call that optionality, patience, portfolio protection. Every unnecessary renewal is also revenue for the institution giving the advice, and the owner cannot tell conviction from compensation. So the test of alignment is almost embarrassingly simple: how often does the agent tell the owner not to buy its product? This name is available and not worth registering. These 318 renewals will not pay for themselves. Transfer this one — another provider is cheaper.
An institution reveals whom it serves when serving that person costs it money. Everything else is advertising.
The old appraisal was a number on a screen. The new one initiates conduct. A model estimates $4,200. A system sets a $4,999 buy-now. The affiliated marketplace lists it, a buyer's agent finds it, a negotiation agent settles at $4,300, the registrar processes the transfer, the marketplace books the commission. The estimate looks vindicated. But it was never tested against the market — it built the road the market drove down.
A platform can then report that its estimates strongly predict sales without disclosing how many of those sales it anchored. That is not manipulation. It is measurement becoming market structure — and once an estimate can cause the outcome, independence is no longer a feature of valuation. It is a requirement.
WECKETT prices from evidence and takes no side of the transaction. DomainGame adds what an appraisal cannot: it captures an independent human estimate before the institutional number is shown, turning the crowd into a control group. That yields three measurements the industry has never published together — accuracy against the realized sale, the anchor (how far estimates move once an institutional number appears), and the alignment of what the institution recommended after producing its number.
The number can be identical across providers; the allegiance is not. A $5,000 estimate paired with "sell now, through us" is a different instrument than the same $5,000 paired with "hold" or "get a second opinion." Which lets the market finally keep two scoreboards: one for who priced it best, and one for the question no registrar has had to answer — who behaved best after pricing it?
The institution that registers the asset should not automatically control its price, its distribution, its machine-access rights, and the advice its owner hears. Every mature market either separates those functions or discloses the conflict. Domains are about to find out which of their institutions grew into that responsibility — and which were built, all along, to serve someone other than the owner.
Captures the independent estimate before the institution's. The control group.
domaingame.io ↗Evidence note: quarterly volume and venue distribution from WECKETT's recorded-sales corpus (sample and dates on file); the ~65× institutional miss (GoValue $2,141 estimate vs. $140,000 realized) from disclosed transaction records; AI-native registrar claims from the provider's own public materials. Analysis and inference are the author's; no intent is alleged where the record shows only outcome.