PlatformHow it worksPricingFull Intelligence ReportFor small businessEnterprise Contact
Measurement standards

How OG01 measures,
and what it refuses to assert

Every OG01 reading is made from outside your business, from publicly observable information, using an independent scoring methodology. Nothing you tell us about yourself is accepted as an input, no value that could not be measured is filled in, and no claim is made about the internal ranking logic of any AI system — because that is not observable from outside and nobody who publishes a factor list can show you how they verified it. What makes the output useful is not that it is comprehensive. It is that it is repeatable.

See a reading of your domain → The five standards
Standards

Five rules the reading is held to

1 · Observed from outside, never self-reported

The measurement takes the same vantage point a search engine or an AI assistant has: publicly available pages, publicly available records, and what assistants actually return when prompted. No analytics access, no CMS access, no credentials, and no figure supplied by you can move the result. That is what allows a reading to be handed to a third party — a client, a board, an acquirer — as something other than a self-assessment.

2 · Unmeasured is reported as unmeasured

Where a value could not be obtained, the reading says so rather than substituting a zero. This sounds like a detail and is the single most important rule on the page. A zero is a measurement; an absence is not. Treating one as the other manufactures a deficiency the evidence does not support — and, worse, sometimes manufactures a reassurance, because an unmeasured risk score read as zero produces a clean bill of health nobody checked. Nobody disputes good news about themselves, which is why that direction of the error is the dangerous one.

3 · A sample says it is a sample

Where coverage is partial, the reading gives the ratio rather than the raw count. Twenty pages read out of forty-eight discovered is reported as exactly that, because "twenty pages" alone overstates coverage in precisely the direction that flatters the vendor, and "one page" elsewhere understates it in the direction that flatters a different vendor. The ratio removes the incentive.

4 · An unreadable site is Not Rated, not badly rated

When a domain refuses automated requests, the result is Not Rated. It would be trivially easy to return a low score instead, and it would be wrong: banding a domain against evidence nobody has is the same error as reporting an unmeasured value as zero, one level up. The measurements that do not require fetching your pages are still returned, and the explanation describes the evidence from this scan. It does not establish which other crawlers can access the site. The linked FAQ covers retry options and support.

5 · The method does not move between readings

A rescore runs the identical measurement against the same domain. Because nothing about the method changed in between, the difference between the two readings is attributable to what happened to the domain. This is the entire basis on which OG01 claims a change is real, and it is why the measurement is not quietly tuned to make results look better over time.

Scope

What is measured, and what is deliberately not

Measured
Whether your organisation resolves as one clear entity; how you are retrieved and positioned in conventional search; whether your pages can be fetched, parsed and quoted; what assistants say when prompted about your business; and what inaccurate, stale or contradictory information about you is sitting in the sources these systems draw on.
Not measured, and never asserted
The internal weighting any AI system applies; whether a specific answer will name you tomorrow; traffic, revenue or conversion; the contents of any page that could not be read. A reading that cannot see something says nothing about it.
Why there is no published list of ranking factors here
Because a confident list of unverifiable factors is the most common piece of bad advice in this category, and publishing one would make this page part of the problem. The systems that matter do not expose their selection logic to anyone outside the companies that build them. What can be established from outside is whether the conditions those systems are known to depend on — retrievability, entity clarity, corroboration, absence of contradiction — are present on your domain. That is what is measured, and it is reported as a measurement rather than as a theory about the model.
Independence

Why the diagnosis is separated from the fix

OG01 sells remediation, and the same organisation producing a diagnosis and quoting for the cure is a conflict worth naming rather than hoping nobody notices. Three things keep it honest. The measurement takes no self-reported input, so it cannot be talked into a worse result. The findings name specific things in specific places, so an inflated list is checkable. And the rescore uses the same method, which means overstating a problem in the first reading produces an embarrassing second one rather than a bigger invoice.

You are also under no obligation to buy the fix from us. Every finding is specified to be executed by someone else, and plenty of customers take the roadmap to their own developer or agency. A specification that only your author can implement is not a specification.

Related

Where to go next

The customer journey: score, findings, fix, rescore → What a reading contains → How findings are ordered → What the fixes involve → What it costs →
Questions

Measurement questions

Will you publish the exact factors and weightings behind the score?
No. The score is produced by an independent scoring methodology, and publishing its internals is not ours to do. What is published is the evidence: under each area sit the measured signals it was built from, so any claim in a reading can be checked against something observable rather than taken on trust.
Does OG01 have inside knowledge of how ChatGPT or Google rank sources?
No, and neither does anyone else outside those companies. OG01 measures observable conditions and records what assistants actually return when prompted. It makes no claim about internal selection logic, and it will not publish a factor list it cannot show you the verification for.
Is the reading repeatable? Would I get the same number twice?
The method is held constant between readings, which is what makes a before-and-after meaningful. The world is not held constant, though: your own site changes, third-party sources change, and what assistants return varies. That is why a rescore is framed as a comparison of two measurements rather than as a claim that a single number is exact.
Why does OG01 refuse to score a site it cannot read?
Because a score built from a reading that never arrived is a fabrication, and it would place your domain on a scale using evidence nobody has. Not Rated is returned instead, with any supported off-site findings and the explanation available. Read the Not Rated FAQ before deciding whether to retry or contact support. The result alone does not establish a configuration fault.
How is this different from an SEO audit?
An SEO audit is usually built around position on a results page. This measurement is built around whether systems that write answers rather than lists can understand, retrieve and safely repeat information about your business. Search retrieval is one input to that rather than the whole of it, which is why conventional rankings can be strong while the reading is not.
Who can I show a reading to?
Anyone. Because nothing in it is self-reported, it works as third-party evidence in a way an internal assessment does not — for a client, a board, a partner or an acquirer. That property is the reason the no-self-reported-input rule is absolute rather than a default.
The standards matter most when the news is bad.

Run the free reading and see what it declines to claim as well as what it reports. Ninety seconds, no card.

Get your free score → See the full evidence — $399 →