The signal underneath the ranking

Entity verification

Before a discovery system can describe a business, it has to establish that the business is real. That check is older than AI search and mostly invisible, and it is becoming the thing worth owning. This page records what carries the signal, what the largest platform actually requires, and where the cost of proof now falls.

SEPARATE RECORDS, ONE ENTITY STATE REGISTRY PLATFORM PROFILE STRUCTURED DATA THIRD PARTY VERIFIED ENTITY

Entity verification  /  noun

The process by which a discovery system establishes that a business or organisation is a real, operating entity. It is separate from how well that entity ranks, and separate from how accurately it is described.

The AIOFacts usage. The underlying practice long predates AI search and has no settled name across the industry.

What carries the signal

Four records, and only one of them is cheap.

Verification is not a single check. It is the agreement between independent records, each controlled by a different party, each costing the business something different to establish and maintain.

RecordWho controls itWhat it costs the business
Company or state registryThe governmentA filing fee and an annual obligation. Lapses quietly and is usually discovered late.
Verified platform profileThe platformWhatever the platform decides to require, on terms the business does not set.
Structured data on an owned domainThe businessClose to nothing, and it is the only one of the four fully under the owner's control.
Third-party corroborationOther publishersTime, and no guarantee. Cannot be bought without becoming the problem it is meant to solve.

The third row is the one most businesses have never populated, and the only one they can fix in an afternoon.

What the largest platform actually requires

Supported

Every quotation in this section is taken from Google's own published Business Profile documentation and can be checked at the linked sources.

Google Business Profile is the largest single entity-verification system a small business will encounter, so it is the clearest available example of how the cost of proof is distributed. Three things in its published documentation are worth recording, because they are routinely misunderstood.

The business does not choose how it is verified. Google's documentation states: "Verification methods are automatically determined by Google and can't be changed." It further notes that "Verification through postcard isn't available for all businesses" and that "Live video verification may not be available for all businesses." A business offered only one method has not missed a setting. There is no setting.

Video verification must be captured live. Google's documentation states: "You must capture the footage live as pre-recorded videos aren't allowed, and you can't record it offline and upload it later." The video must be at least 30 seconds long. There is no opportunity to review the submission before it becomes the submission.

For a business without a storefront, documents are explicitly accepted. For service-area businesses, Google asks the recording to show "street signs, nearby landmarks, or other identifiers at your business address", then "the professional tools, equipment, or products you use for work", and finally proof of management by "performing a service or accessing business-only assets". It adds: "You can also show business documents like a business permit, invoice, or utility bill."

That last provision matters more than its placement suggests. A great deal of published advice treats branded signage or a branded vehicle as the requirement, which is expensive and, for a business that has neither, impossible. The documentation does not say that. A printed permit, invoice or utility bill carrying the registered name satisfies the same requirement at close to zero cost.

The problem these requirements answer

Supported

Figures published by Google covering calendar year 2025, in its own announcement dated 16 April 2026.

Any account of verification that omits the fraud it exists to prevent is incomplete. Google reports that in 2025 it "placed posting restrictions on more than 782,000 policy-violating accounts and removed over 13 million fake Business Profiles." Over the same period it "blocked or removed over 292 million policy-violating reviews" and "blocked 79 million inaccurate or unverified edits to help keep Business Profiles factual."

Thirteen million fabricated business profiles in a single year is not a rounding error and not a pretext. A discovery system with no verification layer would be substantially worse for the businesses complaining about the verification layer. Both of those things are true at once, and the second one does not dissolve the first.

AIOFacts position

AIOFacts position

What follows is an argument this site is making. It is not a finding, it has not been measured, and no one else has adopted it.

As AI systems answer more questions directly, the unit of value moves. Ranking assumed a list a person would scan. A recommendation is singular, and what a recommending system needs first is not a better page but a confident answer to a prior question: is this entity real, and is it the one being asked about.

That makes the verified entity record the durable asset and the ranking a derivative of it. Our position is that the cost of establishing that record has been transferred onto the business at precisely the moment the record became valuable, and that the transfer has happened without a corresponding obligation in the other direction. A business can be asked to prove itself on camera, in one continuous take, on terms it did not choose, with no reason stated in advance.

The second observation is the one we would defend hardest. Verification of this kind is a real obstacle to a compliant single-location business and a rounding error to an operation generating listings at scale, which absorbs the cost and automates around it. Enforcement whose burden falls hardest on the party least likely to be the problem is not, in its effect, a fraud control. It functions as a tax on legitimacy. We do not claim that is anyone's intention, and we have no evidence about intention.

What remains unsettled

Unverified

Open questions this site cannot currently answer with evidence.

Whether AI answer systems weight a verified platform profile above a state registry record, or treat them as interchangeable corroboration, is not publicly documented by any major system. Whether verification status propagates between platforms, so that being verified in one place makes verification elsewhere more likely, is unknown to us. And whether the share of small businesses abandoning verification after a failed attempt is material is not something we have measured, though it is the number that would decide how serious the effect described above actually is.

Anyone holding data on the third question is invited to send it. We will publish it, including if it contradicts the position above.

Sources

Documentation on this subject changes without notice. Quotations were read directly from the linked pages on 25 July 2026. If one of them no longer matches, that is a correction we want: see the corrections policy.