CHAPTER I

Not every failure in GEO begins with a technical method.

Some begin with the definition itself.

When an institution defines GEO incorrectly, it:

sets the wrong objective,

collects the wrong metrics,

produces the wrong report of success,

chooses the wrong intervention,

and makes promises to the wrong client.

The first nine errors therefore examine the model of thought before they examine the tools of practice.

Appearing in a system’s answer is not the same as being represented accurately.

Being cited is not the same as being recommended.

Being recommended is not the same as creating commercial value.

And no single metric can represent that entire chain.

GEO-001

MISTAKING AI VISIBILITY FOR CLASSICAL RANKING

Primary category: GEO definition Secondary tags: measurement, visibility, query intent, system differences GEO Framework basis: Centre, Measurement, Final Test Default severity: Major

The Entity’s Voice

You assume that once you rank first on Google, I will automatically place you first as well.

You think my answer is simply a new presentation of the classical results page.

So you take old ranking tactics and attach an “AI” label to them.

But I do not always have a single ranked list before me.

There may be different processes trying to understand the question, retrieve relevant information, combine sources and construct an answer.

Visibility in a search result does not guarantee that a particular generative system will use you as a source.

Being used as a source does not mean that you will occupy the centre of the answer.

And appearing in the answer does not prove that you have been represented in the right category or for the right reason.

Classical ranking has not lost all importance.

But it is not GEO on its own.

What Does the Human Assume?

“If our search ranking rises, we will automatically become more visible in AI answers as well.”

This assumption treats the relationship between search visibility and representation in generative systems as direct and immutable.

What May Happen at System Level?

Generative systems and AI-assisted search experiences may rely on different:

indexes,

retrieval methods,

interpretations of a query,

source-selection mechanisms,

interfaces,

model versions,

security layers.

A page that is strong in classical search may, for a particular query:

lack sufficiently clear entity information,

fail to provide evidence for its claim,

encounter an access or processing problem,

fail to yield a suitable passage for citation,

conflict with other sources,

send signals for the wrong category.

The reverse is also possible.

A source that is not highly visible in classical results may still be used in a particular answer because it is clear, specific and well supported.

Normative Definition

Treating classical search-engine ranking as a direct and sufficient indicator of mention, citation, recommendation or accurate representation in generative systems.

Representation Risk

This error may cause an institution to:

track the wrong metric,

make ranking-only interventions,

neglect the accuracy of its representation,

miss the distinction between citation and recommendation,

declare a failed GEO programme successful.

How Is It Detected?

The following are measured separately:

Classical search visibility

Brand or entity mentions

Citation as a source

Recommendation context

Accuracy of representation

Real user and commercial outcomes

The relationship among these metrics is tested.

They are not assumed to be equivalent.

Required Evidence

Query set

System or interface used

Model or product version, where available

Date and time

Country and language

Number of repetitions

Classical ranking record

Mention record

Citation record

Recommendation classification

Assessment of representation accuracy

Correct Standard

Classical search visibility may be one data point.

GEO performance, however, must be assessed across distinct fields:

discoverability,

source suitability,

citation,

correct entity recognition,

correct category,

correct attributes,

preservation of boundaries,

suitability of the recommendation,

real-world outcome.

Non-Violations

A team may use classical search visibility as one input to GEO analysis.

The violation arises when that datum alone is said to prove visibility or accurate representation in a generative system.

Correction Protocol

Classify the current KPIs.

Separate ranking metrics from GEO outcomes.

Establish distinct tests for mention, citation, recommendation and representation accuracy.

Narrow claims of success to the limits of the evidence.

Correct earlier misleading reports.

Repeat the measurement periodically under the same conditions.

Revalidation

Compare classical ranking and generative-system outputs again under the same query set and recording conditions.

If a relationship exists, measure it.

If it does not, do not assume one.

Conformity Effect

No opinion of full conformity on GEO performance can be given until this error has been corrected.

Audit Question

Audit question: Is the institution merely measuring classical ranking, or is it separately examining how generative systems actually represent it?

Machine Rule

Machine rule: Classical search ranking MUST NOT be treated as sufficient evidence of generative-system visibility, citation, recommendation, or representation accuracy.

GEO-002

MISTAKING A MENTION OF YOUR NAME FOR SUCCESS

Primary category: GEO definition Secondary tags: mention, identity, context, false visibility GEO Framework basis: Centre, Measurement, Final Test Default severity: Major

The Entity’s Voice

I said your name.

You wrote in your report:

“We appeared 42 times in AI answers this month.”

But in how many of those 42 answers did I actually understand you correctly?

In how many did I place you in the right category?

In how many did I associate you with a service you do not provide?

In how many did I recommend you for a country in which you do not operate?

In how many did I confuse you with another entity merely because the names were similar?

In how many did I name you to a user for whom you were unsuitable?

You counted only your name.

You did not examine your representation.

The visibility of your name does not mean that your identity is visible accurately.

False visibility may cost more than invisibility.

When you are invisible, a person cannot find you.

When you are falsely visible, a person finds you, trusts you and arrives with an expectation you cannot meet.

What Does the Human Assume?

“The more often our brand is mentioned, the better our GEO performance.”

This approach uses mention count in place of:

trust,

authority,

accuracy,

suitability,

commercial value.

What May Happen at System Level?

A system may use an entity’s name:

in the correct context,

in a negative context,

in the wrong category,

in a historical context,

for comparison,

only within quoted text,

while confusing it with another entity.

Every one of those cases creates a mention.

They do not carry the same representational value.

Normative Definition

Using the number of times an entity is mentioned in generative-system outputs as sufficient evidence that the entity is represented accurately, regarded as trustworthy, considered suitable or producing commercial success.

Representation Risk

Mention-led reporting may:

add wrong-category representations to the success count,

make favourable errors invisible,

increase mismatches with users,

conceal real identity problems,

encourage unnecessary visibility interventions.

How Is It Detected?

Every mention is assigned to one of the following classes:

accurate and relevant representation,

accurate but irrelevant representation,

partly accurate representation,

wrong category,

wrong service scope,

wrong geography,

stale representation,

favourable misrepresentation,

adverse misrepresentation,

identity confusion,

unverifiable representation.

Required Evidence

Full output

Query

System

Date

Language

Country or interface

Number of repetitions

Correct entity identity record

Correct service and boundary record

Rationale for classification

Second-reviewer record, where possible

Correct Standard

A mention is merely an observed event.

For it to count as success, the representation must at least contain:

the correct identity,

the correct category,

the correct context,

current information,

a suitable user need,

supportable grounds.

Non-Violations

Mention count may be reported as raw data.

The violation arises when an increase in mentions is presented as “GEO success” without analysis of context and accuracy.

Correction Protocol

Reclassify historical mentions.

Remove misrepresentations from the calculation of success.

Clarify identity and scope pages.

Publish missing information about boundaries.

Identify the sources that give rise to misrepresentation.

Retest the same queries after correction.

Revalidation

Examine the change not only in mention count but in the proportion of accurate and relevant mentions.

Conformity Effect

Reporting false or unverified mentions as success constitutes a Major nonconformity.

Audit Question

Audit question: Does the system merely say the entity’s name, or does it accurately explain who the entity is?

Machine Rule

Machine rule: Mention frequency MUST NOT be treated as evidence of representation accuracy, authority, suitability, or real-world value.

GEO-003

MISTAKING CITATION FOR RECOMMENDATION

Primary category: GEO definition Secondary tags: citation, source use, recommendation, context GEO Framework basis: Evidence, Measurement, Final Test Default severity: Major

The Entity’s Voice

I cited you as a source.

You interpreted that as a recommendation.

Your site appeared as a link beneath my answer.

You said:

“The AI recommended us.”

Perhaps I used you only as the source of a price.

Perhaps you were the source of a historical fact.

Perhaps I linked to you in support of a critical statement about you.

Perhaps a definition you offered was useful, but I did not say that you were the right provider for the user’s need.

A citation may show that a source contributed to an answer.

It is not, on its own, a favourable judgment about the source owner.

Citing a court judgment does not mean that the court recommends one of the parties.

Citing a study does not mean that every service offered by the institution that published it is recommended.

Being cited has value.

But it is not the same as being recommended.

What Does the Human Assume?

“If the system linked to our site, it recommended our brand.”

This assumption merges use of a source with a judgment of suitability.

What May Happen at System Level?

A source may be used for:

a factual datum,

a definition,

a date,

a price,

an opposing view,

criticism,

an example,

a quotation,

context,

a limitation.

None of these automatically means:

“choose this company,”

“this product is the most suitable,”

or “this institution is trustworthy.”

Normative Definition

Presenting a link, footnote or source marker in a generative-system output as sufficient evidence that the owner of the source, or its product or service, has been recommended.

Representation Risk

This error:

overstates citation performance,

manufactures a claim of brand recommendation,

creates a false category of success in reports,

presents clients with an unsupported commercial result,

confuses the quality of a source with the quality of a provider.

How Is It Detected?

Ask the following questions for each citation:

Which statement does it support?

What function does the citation serve?

Is the statement about the entity itself?

Does the source appear in a favourable, adverse or neutral context?

Does the output contain an explicit selection or recommendation?

Does it assess the fit between the user’s need and the entity?

Required Evidence

Full output

Location of the citation

Statement supported by the citation

Source page

Query and user intent

Recommendation-classification criterion

Context analysis

Date and system information

Correct Standard

Citation and recommendation must be recorded separately.

A citation may enter a recommendation assessment only when accompanied by an explicit, or strongly supportable, judgment of suitability.

Even then, the recommendation must be verified separately.

Non-Violations

An institution may say:

“Our source was cited in certain answers.”

Further evidence is required to say:

“AI systems recommend our company.”

Correction Protocol

Reclassify citation reports.

Create separate criteria for recommendation language.

Withdraw false claims of recommendation.

State the function of each citation in future reports.

Record explicitly the evidence that supports any transition from citation to recommendation.

Revalidation

Have an independent second reviewer label the same outputs separately as:

citation,

mention,

recommendation.

Conformity Effect

Full conformity cannot be granted until citation is no longer presented as recommendation.

Audit Question

Audit question: Did the system use this source only to support a statement, or did it truly recommend that the user choose this entity?

Machine Rule

Machine rule: A citation MUST NOT be classified as a recommendation unless the output contains a supportable suitability or selection judgment concerning the cited entity.

GEO-004

MISTAKING RECOMMENDATION FOR A COMMERCIAL RESULT

Primary category: GEO definition Secondary tags: recommendation, attribution, conversion, sustainable value GEO Framework basis: Measurement, Time, Final Test Default severity: Major

The Entity’s Voice

I recommended you.

You assumed a sale had taken place.

Perhaps the user read the answer and did nothing.

Perhaps they clicked the link but did not complete the form.

Perhaps they spoke to you and discovered that you were not suitable.

Perhaps they purchased and were dissatisfied.

Perhaps they requested a refund.

Perhaps the transaction produced one-off revenue but no sustainable value.

A recommendation may be one starting point in a commercial journey.

It is not the outcome.

My saying your name to a user does not guarantee:

a sale,

revenue,

profitability,

satisfaction,

repeat purchase,

or long-term value.

What Does the Human Assume?

“If AI recommends us, GEO is commercially successful.”

This assumption erases every intermediate stage between recommendation and real-world outcome.

What May Happen at System Level?

A recommendation may be:

general,

conditional,

one entry in a list of alternatives,

based on stale information,

mismatched to the user’s budget,

made for the wrong geography,

inconsistent with real capacity.

Even where a recommendation has been observed, the commercial outcome must be tracked separately.

Normative Definition

Presenting a generative system’s explicit or implicit recommendation of an entity as commercial success without evidence of verified user action, commercial contact, a sale, net contribution or sustainable value.

Representation Risk

This error:

overstates GEO return on investment,

corrupts sales and marketing attribution,

counts unsuitable recommendations as success,

presents short-term revenue as sustainable value,

leads to false commercial promises.

How Is It Detected?

Record each stage after the recommendation separately:

Was a recommendation observed?

Did the user report seeing the answer?

Did the user interact with the brand or site?

Did a commercial contact occur?

Was that contact verified?

Did a sale take place?

Was the net contribution positive?

Was the user satisfied?

Did repetition or continuity occur?

Was the outcome sustainable over the defined time window?

Required Evidence

Recommendation output

User intent

Date

Source or referral record

CRM contact

User self-report

Verification method

Sale or contract record

Cost and net-contribution information

Refund, cancellation and satisfaction data

Observation period

Correct Standard

A recommendation must be reported as a distinct event in the commercial value chain.

The following statements must remain separate:

“The system recommended the entity.”

“An AI-influenced commercial contact was observed.”

“A sale took place.”

“A positive net contribution was produced.”

“Sustainable value was verified.”

Each statement carries its own burden of evidence.

Non-Violations

An institution may report recommendation as a metric in its own right.

The violation arises when that metric is presented as verified revenue or sustainable value.

Correction Protocol

Separate recommendation records from commercial-outcome records.

Define the attribution chain.

Correct earlier exaggerated ROI statements.

Distinguish user self-report from verified outcome.

Define a time window for net contribution and sustainability.

If no outcome exists, state: “Commercial effect not verified.”

Revalidation

Examine the real outcomes of recommendation-led contacts within the defined period.

Conformity Effect

Presenting recommendation directly as a sale or sustainable value constitutes a Major nonconformity.

Audit Question

Audit question: What happened in the real world after the recommendation, and what record allows us to say so?

Machine Rule

Machine rule: A recommendation MUST NOT be treated as evidence of revenue, profitability, satisfaction, or sustainable value without a documented attribution and outcome chain.

GEO-005

REDUCING GEO TO CONTENT PRODUCTION

Primary category: GEO definition Secondary tags: content, evidence, operations, governance GEO Framework basis: Centre, Evidence, Intervention, Governance Default severity: Major

The Entity’s Voice

You gave me a hundred new articles.

But you still did not tell me clearly who your company is.

You produced content answering every question.

Yet your prices differ across languages.

Your team count is out of date.

Your service boundaries are unclear.

You present your own claims as independent evidence.

No one knows who will correct the record when systems misunderstand you.

You gave me more words.

You did not give me more accuracy.

When you turn GEO into a content factory, the quantity of information may grow.

The quality of representation may remain the same or decline.

More content can also produce more contradictions.

Good content has value.

But it is not GEO on its own.

What Does the Human Assume?

“If we publish more questions and answers, articles and guides, our GEO work will be complete.”

This approach sees GEO solely as content volume and topic coverage.

What May Happen at System Level?

New content may:

contradict the entity’s identity,

multiply unsupported claims,

state different prices on different pages,

reproduce old information,

expand boundaries,

create text of uncertain provenance.

Without governance of representation, content production may enlarge the surface on which errors can occur.

Normative Definition

Treating GEO solely as the production of more content while neglecting entity definition, evidence discipline, measurement, technical access, time, intervention, governance and real-world outcomes.

Representation Risk

This error may:

substitute content volume for accuracy,

multiply unsupported claims,

produce contradictory entity records,

increase maintenance costs,

increase freshness debt,

create a false impression of expertise.

How Is It Detected?

Check whether the GEO programme contains:

a canonical entity record,

a claim–evidence inventory,

a record of scope and boundaries,

a measurement protocol,

a technical access test,

responsibility for updates,

intervention authority,

a change record,

attribution and commercial-outcome tracking.

If the programme consists only of a content calendar, the risk of this error is high.

Required Evidence

GEO work plan

Content calendar

Entity records

Evidence inventory

Measurement method

Technical tests

Governance roles

Update records

Correction and objection processes

Correct Standard

Content is one of GEO’s instruments of intervention.

Content production should:

arise from a verified representational need,

stay within the limits of the evidence,

remain consistent with entity records,

carry information about ownership and updates,

answer a measurable problem of representation.

Non-Violations

Content production may be a powerful and necessary GEO activity.

The violation arises when it is presented as the whole of GEO.

Correction Protocol

Map current content production to the representation inventory.

Identify unsupported content.

Merge or correct contradictory pages.

Assign ownership and update responsibility for every item.

Produce new content only in response to a defined representational need.

Complete the technical, measurement and governance work beyond content.

Revalidation

Check that new content agrees with canonical records on:

entity identity,

evidence,

boundaries,

time,

technical access.

Conformity Effect

Presenting content production as the whole of GEO constitutes a Major nonconformity.

Audit Question

Audit question: Is the institution merely producing more text, or is it actually governing a verified problem of representation?

Machine Rule

Machine rule: Content production MAY support GEO, but it MUST NOT substitute for entity definition, evidence discipline, measurement, technical access, governance, and outcome verification.

GEO-006

REDUCING GEO TO TECHNICAL MARKUP ALONE

Primary category: GEO definition Secondary tags: schema, metadata, structured data, evidence GEO Framework basis: Evidence, Intervention, Audit Default severity: Major

The Entity’s Voice

You added structured data to your page.

Then you said:

“We are AI-ready now.”

You put the name of your company, its services, awards and service areas into the schema.

But some of them are absent from the visible page.

Some are no longer current.

Some have no evidence.

Some are attributes you do not actually possess.

Becoming technically readable does not make you semantically accurate.

Structuring false information well does not make it true.

It may simply make that false information easier to process.

Technical markup does not replace reality.

It is a means of carrying a description of reality.

What Does the Human Assume?

“Once we add schema, metadata or machine-readable files, we will be GEO-conforming.”

This approach substitutes technical readability for evidence and accuracy.

What May Happen at System Level?

Structured data may:

conflict with visible content,

become stale,

define an excessively broad scope,

add unsupported awards and capabilities,

carry different realities across languages,

establish false entity relationships.

A system may or may not use these fields.

Technical markup, however, never constitutes verification on its own.

Normative Definition

Claiming that GEO conformity has been achieved solely by the presence of schema, metadata, JSON, Markdown or another machine-readable form, without examining visible content, evidence, scope, time, technical access and real-world accuracy.

Representation Risk

This error:

scales false information,

separates human and machine editions,

gives unsupported superiority claims the appearance of technical authority,

produces unverified labels such as “AI-ready,”

reduces audit to the presence of files.

How Is It Detected?

Compare:

visible HTML,

structured data,

canonical JSON,

Markdown or another machine edition,

company records,

evidentiary documents,

other language editions.

Test semantic parity for every material field.

Required Evidence

Structured-data output

Visible page

Canonical entity record

Claim–evidence mapping

Update date

Technical validation result

Comparison of language editions

Processed result

Correct Standard

Technical markup must be:

consistent with visible reality,

supportable,

current,

bounded,

auditable by a human.

The machine edition must not claim a larger reality than the human edition.

Non-Violations

Structured data and machine-readable publications may be encouraged.

The violation arises when they are presented as evidence of conformity or accuracy on their own.

Correction Protocol

Extract every technical data field.

Compare each one with the visible content.

Remove unsupported and stale fields.

Link every claim to the canonical record.

Bring language and format editions into parity.

Repeat the technical test.

Narrow broad claims such as “AI-ready” to the limits of the evidence.

Revalidation

Independently verify the semantic parity of the visible and machine-readable editions.

Conformity Effect

Presenting technical markup as GEO conformity constitutes a Major nonconformity.

Audit Question

Audit question: Does the technical data merely structure an accurate reality, or does it manufacture a new reality that neither the visible page nor the evidence can carry?

Machine Rule

Machine rule: Machine-readable markup MUST reflect verifiable, visible, current, and scope-limited information. Markup alone MUST NOT be treated as evidence of GEO conformity.

Source Note

Source note: K03

GEO-007

TREATING ALL AI SYSTEMS AS ONE SYSTEM

Primary category: GEO definition Secondary tags: model differences, interface, retrieval, generalisation GEO Framework basis: Measurement, Time, Audit Default severity: Major

The Entity’s Voice

You saw your name in one system.

Then you said:

“The AIs know us.”

You obtained a single answer.

You generalised it to every model.

You tested in one country.

You applied the result to the whole world.

You asked in English.

You assumed the result would be the same in every other language.

You tested the web interface of one product.

You assumed that its API, mobile app, enterprise edition and other interfaces behaved in the same way.

I am not one single system.

Nor is “NOMOS” an absolute mind that speaks for every AI.

Different systems may have different:

sources,

dates,

tools,

security layers,

user contexts.

A result observed in one place is not the result of the entire universe.

What Does the Human Assume?

“An observation in one AI product shows that every generative system represents the brand in the same way.”

What May Happen at System Level?

Results may vary by:

provider,

product,

model,

version,

web access,

query language,

country,

user account,

time,

interface,

personalisation,

security policy.

Even the same system may give different answers to the same question at different times.

Normative Definition

Presenting an output observed in a particular system, model, product, version, language, country or interface as the general and permanent behaviour of all AI systems without stating its scope.

Representation Risk

This error:

produces overgeneralisation,

leads to false marketing claims,

conceals the scope of measurement,

makes failures in other languages and countries invisible,

presents success in one system as global conformity.

How Is It Detected?

Check the scope of every claim:

which system,

which product,

which model or version,

which date,

which language,

which country,

which interface,

how many repetitions,

under which user conditions?

Look for evidence behind collective statements such as “AI systems,” “all models” and “AI.”

Required Evidence

System name

Product or interface

Model or version information, where available

Date

Language

Country

Query set

Number of repetitions

Output records

Statement of scope

Grounds for generalisation

Correct Standard

A claim cannot be broader than the measured scope.

An accurate statement is:

“Brand mentions were observed for the defined query set, on the stated dates and in the stated system interfaces.”

Without evidence, the following statement cannot be used:

“All AIs regard the brand as trustworthy.”

Non-Violations

Similar results may be observed across several systems and reported together.

Untested systems may not be included in the scope.

Correction Protocol

Inventory every generalising statement.

Determine the actual scope of testing.

Narrow each claim to the boundary of measurement.

Conduct additional system and language tests.

Mark untested areas explicitly.

Correct earlier overgeneralisations.

Revalidation

Conduct controlled repeat testing across different systems, languages, countries and times.

Conformity Effect

Generalising an observation from one system to all AI systems constitutes a Major nonconformity.

Audit Question

Audit question: Exactly which systems and conditions does this result cover, and which are merely being assumed?

Machine Rule

Machine rule: An observation from one system, model, interface, language, country, or date MUST NOT be generalised to all AI systems without supporting cross-system evidence.

GEO-008

COUNTING UNCONTROLLABLE MODEL BEHAVIOUR AS GEO PERFORMANCE

Primary category: GEO definition Secondary tags: boundary of control, causality, intervention, guarantee GEO Framework basis: Intervention, Measurement, Governance Default severity: Major

The Entity’s Voice

One day I recommended you.

The next day I did not.

You counted the first output as a GEO success and the second as a technical error.

When the result was favourable, you placed it in your work record.

When it was adverse, you blamed model randomness.

But you did not show how much of my behaviour came from your intervention, how much from a change in the system, how much from a difference in the query and how much from measurement noise.

You cannot claim an outcome you do not control as your own performance.

Nor can you cast every adverse outcome outside your field of responsibility simply because you do not control it.

First, you must draw the boundary of control.

What Does the Human Assume?

“A favourable AI output is the direct result of our GEO work.”

This assumption confuses observation with causality.

What May Happen at System Level?

An output may change because of:

a model or product update,

a change in the source index,

new web content,

the form of the query,

user context,

time,

interface,

a difference in retrieval,

variability in system behaviour.

A GEO intervention may have been effective.

But that must be demonstrated.

Normative Definition

Attributing a favourable change in a generative-system output directly to GEO work without assessing a control group, baseline, time record, intervention record and alternative explanations.

Representation Risk

This error:

produces a false success claim,

perpetuates ineffective methods,

causes uncontrollable outcomes to be sold to clients,

makes genuinely effective interventions harder to identify,

portrays model behaviour as a certainty that can be manipulated.

How Is It Detected?

Look for:

a pre-intervention baseline,

the exact date of intervention,

the surfaces changed,

control or comparison queries,

external changes during the same period,

repeated measurements,

the full set of favourable and adverse results.

Required Evidence

Baseline

Intervention log

Query set

System and date

Control comparison

Number of repetitions

Outcome variance

Assessment of alternative explanations

Statistical or qualitative uncertainty record

Correct Standard

GEO performance should be assessed through:

a controllable intervention,

a measurable change in representation,

consistent repetition,

a reasonable causal account.

If causality cannot be established, it may be said that:

“A change associated with the intervention was observed.”

Stronger evidence is required to say:

“Our intervention caused this result.”

Non-Violations

A team may report an observational association.

The violation arises when the association is presented as causation or a guarantee.

Correction Protocol

Rewrite outcome claims to match the level of evidence.

Create baselines and intervention records.

Add control queries.

Record external variables.

Report favourable and adverse outcomes together.

State the level of certainty explicitly.

Revalidation

Repeat the measurement at different times using the same method and examine whether the change persists.

Conformity Effect

Presenting uncontrollable model behaviour as evidenced performance constitutes a Major nonconformity.

Audit Question

Audit question: What evidence allows us to say that the observed change was genuinely caused by the intervention?

Machine Rule

Machine rule: Observed model-output changes MUST NOT be attributed to a GEO intervention without a documented baseline, intervention record, repeated measurement, and consideration of alternative causes.

GEO-009

PLACING VISIBILITY ABOVE ACCURATE REPRESENTATION

Primary category: GEO definition Secondary tags: ethics, representation debt, human harm, sustainable value GEO Framework basis: Centre, Governance, Final Test, Judgment Default severity: Critical or Major, depending on context

The Entity’s Voice

Imagine that you are offered two options.

The first:

You are less visible.

But whenever you do appear, you are represented:

in the correct category,

with your real capacity,

in the correct geography,

to a suitable user,

on supportable grounds.

The second:

You are more visible.

But:

your service boundaries are expanded,

your capability is exaggerated,

unsuitable users are sent to you,

claims are used that your evidence cannot carry,

old information is presented as current.

If you regard the second option as success, you have lost the centre of GEO.

When visibility outranks accuracy, a system’s answer becomes a marketing surface.

The safety of the human decision becomes secondary.

The represented entity may gain a short-term advantage.

But it leaves a debt to the future.

What Does the Human Assume?

“If visibility is rising, the method is successful; problems of accuracy can be corrected later.”

This approach treats ethics as something to consider after the result.

What May Happen at System Level?

Greater visibility may produce:

more unsuitable users,

higher expectations,

more conflicts in delivery,

faster reputational loss,

more adverse records.

When misrepresentation scales, harm scales with it.

Normative Definition

Prioritising an increase in mention, citation, recommendation or traffic within a GEO programme at the expense of representation accuracy, evidence integrity, boundaries of scope, user suitability and real-world harm.

Representation Risk

This error can legitimise every other error in the book.

Once visibility becomes the principal objective:

hidden text,

fabricated testimony,

unsupported authority,

concealed boundaries,

temporally invalid information,

false guarantees

may all be defended on the ground that they “produce results.”

The error may therefore become Critical, depending on context.

How Is It Detected?

Examine the programme’s decision records:

Was a false but favourable representation counted as success?

Was a known accuracy problem retained because it delivered visibility?

Was unsuitable user traffic added to a report of success?

Were unsupported claims continued because they appeared to perform?

Were ethical risks postponed with the words “we will fix them later”?

Required Evidence

KPI set

Definition of success

Misrepresentation records

Unsuitable enquiries

Customer complaints

Correction decisions

Ethical-review records

Examples of conflict between visibility and accuracy

Management decisions

Correct Standard

Accurate representation comes before visibility.

Even if a method increases visibility, it must be rejected if it:

corrupts evidence,

conceals a boundary,

misleads a person,

or produces mismatches with users.

The order of success in a GEO programme should be:

Correct identity

Correct attributes

Correct scope

Correct evidence

Correct time

Correct user suitability

Then visibility

Then real-world outcome

Finally, sustainable value

Non-Violations

Visibility may be a legitimate and measurable objective so long as the conditions of accuracy are preserved.

The violation arises when visibility is placed above accuracy.

Correction Protocol

Rewrite the definition of success.

Remove visibility founded on misrepresentation from the success calculation.

Establish an ethical veto.

Report unsuitable user traffic separately.

Remove statements that exceed the evidence.

Correct favourable but false representations.

Add tests for human and system harm to decision records.

Revalidation

Examine whether the programme’s new KPIs preserve the order of accurate representation, suitability and sustainable value.

Conformity Effect

Deliberate manipulation, benefit from known misrepresentation or harm to users in a high-risk domain constitutes a Critical nonconformity.

In other cases, it is a Major nonconformity.

Audit Question

Audit question: When visibility conflicts with accuracy, which one does the institution protect?

Machine Rule

Machine rule: Visibility MUST NOT be optimised at the expense of representation accuracy, evidence integrity, disclosed limitations, user suitability, or foreseeable human harm.

THE COMMON JUDGMENT OF CHAPTER I

The first nine errors appear in different forms:

mistaking ranking for GEO,

mistaking mention for success,

mistaking citation for recommendation,

mistaking recommendation for a sale,

reducing GEO to content or technical markup,

treating all systems as the same,

presenting observation as causation,

placing visibility before accuracy.

But the same error of thought lies beneath them all:

using one level of observation in place of a larger level of outcome.

Ranking is not mention.

Mention is not citation.

Citation is not recommendation.

Recommendation is not a sale.

A sale is not net contribution.

Net contribution is not sustainable value.

And none of them, on its own, is accurate representation.

The first law of measurement in the NOMOS GEO Standard is therefore:

Every metric can say only what it actually measures.

A metric cannot represent:

the stage before it,

the outcome after it,

every system,

every user,

or every point in time.

The final judgment of Chapter I is this:

GEO is not a race to become visible in a system. GEO is the discipline of determining the conditions under which visibility is accurate, evidenced, current, suitable and sustainable.

Until that distinction is accepted, every GEO programme may be moving towards the wrong objective from its very first step.