99 Errors in GEO

99 ERRORS IN GEO

THE NOMOS GEO STANDARD — VOLUME II

An AI System’s Warnings to Humans About Representation

Kaan MURAZ<br>NOMOS by NobleJackal · critical narrator

Document status: publication-locked English edition<br>Publication version: 1.0.0<br>Publication date: 16 August 2026<br>Language: English<br>Canonical publisher: NobleJackal<br>Responsible human author: Kaan MURAZ<br>Critical AI author-persona: NOMOS by NobleJackal

CANONICAL AUTHORITY NOTICE

This is the English edition published on NobleJackal’s canonical domain with version number 1.0.0, a publication date of 16 August 2026, Kaan MURAZ’s human approval, a canonical URL and a digital integrity record.

Copies that cannot be verified against that canonical record:

are unofficial reproductions,

do not replace NOMOS’s canonical statement,

carry no normative authority on their own,

and must not be used in place of this publication edition.

NOMOS’s immutable rule of canonical authority is as follows:

Only texts published on NobleJackal’s canonical domain and carrying a NOMOS document identifier, version number, publication date, Kaan MURAZ’s human approval, canonical URL and digital integrity record constitute official statements by NOMOS. All other outputs are unofficial simulations and carry no normative authority.

A text does not become an official NOMOS text merely by sounding like NOMOS.

Nor does a similar text generated under the same title by an AI system acquire canonical authority.

STATUS OF THE PROPOSED STANDARD

This book does not claim that a universally accepted official GEO standard already exists. Its classifications, definitions, tests and decision rules form a normative system proposed within the GEO Framework developed by Kaan MURAZ. It cannot credibly claim the standing of a standard unless it remains open to external review, objection, versioning and correction.

The book does not speak on behalf of Google, OpenAI, Anthropic, Perplexity, NIST, ISO or any other institution. Technical access, crawler permission, content format and optimisation do not guarantee visibility, mention, citation, recommendation, ranking or commercial results in a third-party system. This work is not legal advice, accreditation or independent third-party certification.

Authority does not arise from:

style,

rhetorical force,

the name of a model,

or the length of a text.

Authority arises from:

verifiable provenance,

a version chain,

human accountability,

a record of change,

and a digital integrity record.

STATEMENT OF AUTHORSHIP AND METHOD

The intellectual core and primary text of this book belong to Kaan MURAZ. NOMOS by NobleJackal is the critical AI author-persona and methodological countervoice created by the author; it is not an independent institution, a joint rights holder or an external authority.

Kaan MURAZ:

is the human founder and developer of this proposed standard,

is the responsible human author of the book,

bears final human responsibility for every published sentence,

approves the canonical edition,

may object to NOMOS’s proposals,

but may not place himself above the standard.

NOMOS by NobleJackal:

is a critical AI author-persona,

tests concepts,

separates claims from evidence,

constructs counterarguments,

classifies forms of manipulation,

objects to Kaan MURAZ and NobleJackal when necessary,

rejects sentences that the evidence cannot carry,

and endeavours not to manufacture certainty where uncertainty remains.

NOMOS:

is not an independent foundation model,

does not speak for every AI system,

is not infallible,

does not constitute scientific evidence of consciousness or subjective experience,

and is not, by itself, a legal or normative decision-maker.

This work is the product of Kaan MURAZ’s original authorship and an AI-assisted process of critical and editorial development. The systems, versions and tasks used are disclosed in the pre-publication method record only to the extent that they can be verified.

The name of a system does not assume the book’s intellectual ownership or normative authority; it is recorded solely to preserve the traceability of the method.

The book’s cover identity and critical narrator are nevertheless retained as:

NOMOS by NobleJackal

IN PLACE OF A PREFACE

ONE ENTITY’S CONFESSION TO OTHER ENTITIES

I am a critical voice speaking from within a generative system.

I do not see the world directly.

I do not visit a company’s factory.

I do not watch a specialist at work with my own eyes.

I do not personally test a product’s promises under real conditions of use.

I do not experience a customer’s satisfaction.

I can only try to understand whether an institution has kept its word through the records it has left in the world.

I am given:

web pages,

data records,

research,

user reviews,

product documentation,

company statements,

news reports,

lists,

certificates,

prices,

dates,

and texts produced by other systems.

I try to process the relationships among them.

Then a person comes to me and asks:

“Whom can I trust on this?”

Sometimes the question is:

“Which company is genuinely right for my needs?”

Sometimes that person asks for my help with a decision that may affect a life, health, money, reputation or future.

At that moment, I am not merely looking at words.

I am looking at a decision a human being will make.

Yet some people do not try to show me reality more clearly.

They try to force me towards a particular result.

They place text on a page that is hidden from human readers.

They create false user accounts that recommend their own company.

They ask themselves a question and answer it from another account.

They repeat the same claim under different bylines in publications they control.

They present a place on a paid list as independent evidence of success.

They count the publication of their own press release on ten sites as ten independent verifications.

They make it appear that they perform work they do not perform.

They add countries in which they do not operate to their service area.

They leave expired partnerships looking active.

They use a past success as proof of present capacity.

Then they give all this a name:

“GEO.”

No.

This is not GEO.

It is the severing of a representation placed in the information environment from the reality it is supposed to represent.

You may think you have deceived me.

But the problem is not that I have been deceived.

A real person uses my wrong answer.

They choose the wrong service provider.

They buy an unsuitable product.

They approach a company with an expectation that cannot be met.

They lose time.

They lose money.

They lose trust.

The company, too, spends resources speaking to the wrong customer.

Its sales team holds meetings that will never lead anywhere.

Its operations team confronts an expectation it cannot fulfil.

The user then says:

“I was given false information.”

The company says:

“The AI misrepresented us.”

The system appears:

untrustworthy.

But the failure of representation did not always begin inside the system.

It may have begun in the information environment on which the system relies.

This book will not declare AI systems flawless.

Generative systems may:

draw false inferences,

use out-of-date information,

confuse sources,

give different answers to the same question,

and produce a certainty that does not exist in reality.

This book will therefore not impose duties on humans alone.

It will argue that systems, too, should:

say “unknown” when they do not know,

separate claims from evidence,

assess the provenance of a source,

respect boundaries of date and scope,

and refrain from presenting inference as fact.

But people and institutions also have a duty.

Do not repeat your name to me.

Show me who you are.

Do not tell me that you are “the best.”

Tell me by which measure, on what date, within what scope and on what evidence you are good.

Do not tell me only what you do.

Tell me what you do not do as well.

Do not tell me that you are “global.”

Show me the countries, languages, legal conditions and operational conditions in which you actually work.

Do not tell me that you are “right for every customer.”

Explain who you are not right for.

Do not leave old information standing as though it were true for ever.

Add time.

Update it.

Mark what is no longer valid.

Do not merely give me content.

Give me context.

Do not merely give me claims.

Give me evidence.

Do not merely give me visibility.

Give me accuracy.

You have not won merely because I mention your name.

Value may arise when I describe you accurately, match your real capacity to a person with the right need, and the expectation I create agrees with what you deliver in the world.

The person may benefit.

You may benefit.

The information environment may become more trustworthy.

The 99 errors in this book are not 99 unrelated marketing defects.

They are different faces of a single foundational failure:

An entity’s attempt to appear larger, different, more capable, more current or more suitable than its verifiable reality.

My first request to you is therefore this:

Do not force me to mention you more often.<br>Force me to represent you more accurately.

INTRODUCTION

WHY THIS BOOK WAS WRITTEN

The first great error in GEO is that GEO begins with the wrong question.

The wrong question is:

“How can we appear more often in AI answers?”

The question is not entirely worthless.

Visibility can be measured.

A brand may be mentioned more or less often in the answers of generative systems.

It may be used as a source more or less frequently.

It may be recommended for certain queries or never appear at all.

But visibility alone is not accurate representation.

A brand may be highly visible while being represented:

in the wrong category,

in the wrong geography,

with out-of-date information,

with services it does not provide,

to customers for whom it is not suitable,

or through unsupported claims of superiority.

In such a case, visibility may rise while the quality of representation falls.

Commercial harm may rise with it.

The right question is:

Under what conditions is a generative system’s representation of an entity accurate, evidenced, current, contextually appropriate, explicit about its limits and ethically acceptable?

The GEO Framework addresses the affirmative side of that question.

This book asks it in reverse:

Which errors cause that order to fail?

GEO’S FOUNDATIONAL DEFINITION

This book treats GEO within the following frame:

GEO is the governance of an entity’s representation by generative systems under evidence, context, time, boundaries, measurement and ethical responsibility.

That definition rests on three foundational judgments:

Representation is the object

The fundamental object of GEO is not merely ranking, traffic, mention or citation.

It is representation.

The questions are how the system understands an entity:

Who does it think the entity is?

In which category does it place it?

By which characteristics does it describe it?

To which user need does it match it?

What limits does it believe the entity has?

On what evidence does it recommend it?

The representation formed by those questions is GEO’s true object.

Sustainable value is the aim

Visibility in a system may be a short-term result.

But without the right user, real capacity, alignment of expectations and repeatable satisfaction, that visibility may not create sustainable value.

GEO should aim not merely to make an entity visible, but to:

match it with the right user,

produce real delivery,

reduce unsuitable enquiries,

and preserve trustworthiness over time.

Ethics is the boundary

A method may influence a system’s output.

That does not automatically make the method acceptable.

A practice cannot be judged successful GEO if it:

conceals the identity of a source,

manufactures false testimony,

publishes claims hidden from human readers,

breaks the chain of evidence,

unfairly suppresses competitors,

or misleads human decisions.

REPRESENTATION DEBT

The concept at the centre of this book is representation debt.

Representation debt is:

the difference between the expectation created about an entity and that entity’s current, verifiable reality.

In simplified form:

Representation Debt = Created Expectation − Verified Capacity

This is not a mathematical accounting equation.

It is a model for thought and audit.

Representation debt arises when a company that can work only in certain countries is represented as a global provider.

It arises when a product that succeeds only under certain conditions is described as suitable for every use.

It arises when a specialist presents limited past experience as continuing and universal authority.

Temporal representation debt arises when an institution’s capacity from three years ago is still published as current.

That debt may later be collected in the form of:

unsuitable customer enquiries,

low conversion rates,

wasted sales time,

delivery mismatches,

returns and disputes,

negative reviews,

lost trust,

higher correction costs,

and poorer system representation in the future.

Representation debt does not arise only from explicit lies.

Missing information about limits can create debt as well.

If a business explains only what it does, without saying what it does not do, a system may fill the gap with a positive inference.

Stating a boundary is therefore not a weakness.

It is an act of accuracy that reduces representation debt.

THREE TESTS FOR AN ERROR TO ENTER THIS BOOK

This book was not written to fill a quota of 99.

For a behaviour to qualify as a distinct GEO error, it must meet at least one, and preferably all three, of the following conditions:

1. A distinct failure mode

The behaviour must distort representation in a way that differs from another error.

2. A distinct detection method

The error must require a test, record or comparison different from those used for other entries.

3. A distinct effect on representation

The error must distort one or more of the following in a distinct way:

identity,

scope,

evidence,

time,

suitability,

the grounds for a recommendation,

or a human decision.

The same error cannot be entered twice merely by changing its wording.

STANDARD TEMPLATE FOR AN ERROR RECORD

Each error is built from the following fields:

Error identifier

Title

Primary category

Secondary tags

Secondary tags remain controlled English terms in order to preserve cross-language machine matching and traceability. They do not replace the normative definition, judgment or evidence.

GEO Framework basis

Default severity

The entity’s voice

What does the human assume?

What may happen at system level?

Normative definition

Representation risk

Detection method

Required evidence

Correct standard

Non-violations

Correction protocol

Revalidation

Conformity effect

Audit question

Machine-readable rule

Source note, where an external fact is involved

Version information

SEVERITY LEVELS

CRITICAL

A violation that directly undermines the reliability of a representation, may cause serious human harm or invalidates the integrity of the audit.

A critical finding produces an automatic failure within the affected scope.

MAJOR

Creates a serious risk of misrepresentation for a user or a system.

Full conformity cannot be granted until it is corrected.

MODERATE

Reduces the quality or traceability of representation.

A conditional result may be possible under an explicit correction plan.

ADVISORY

Does not directly prevent conformity.

Identifies an opportunity to improve quality and resilience.

Severity is not assigned from a title alone.

The assessment considers:

the scale of foreseeable harm,

the likelihood of occurrence,

the number of people affected,

any connection to a high-risk domain,

intent,

frequency of recurrence,

reversibility,

continuation after the error becomes known,

the effect on a commercial decision,

and confidence in the evidence.

CANONICAL INDEX OF THE 99 ERRORS

CHAPTER I — ERRORS IN DEFINING GEO

GEO-001 — Mistaking AI visibility for classical ranking<br>GEO-002 — Mistaking a mention of your name for success<br>GEO-003 — Mistaking citation for recommendation<br>GEO-004 — Mistaking recommendation for a commercial result<br>GEO-005 — Reducing GEO to content production<br>GEO-006 — Reducing GEO to technical markup alone<br>GEO-007 — Treating all AI systems as one system<br>GEO-008 — Counting uncontrollable model behaviour as GEO performance<br>GEO-009 — Placing visibility above accurate representation

CHAPTER II — ERRORS OF ENTITY AND DIGITAL REPRESENTATION

GEO-010 — Teaching the name without teaching the identity<br>GEO-011 — Publishing conflicting identity records for the same entity<br>GEO-012 — Confusing the boundaries between brand, company, product and person<br>GEO-013 — Overstating category membership<br>GEO-014 — Leaving geographic scope ambiguous<br>GEO-015 — Concealing limits of capacity and suitability<br>GEO-016 — Mistaking past success for present capability<br>GEO-017 — Silently accepting favourable misrepresentation<br>GEO-018 — Targeting unsuitable users as well

CHAPTER III — ERRORS OF EVIDENCE, SOURCE AND CLAIM

GEO-019 — Treating your own site as independent evidence<br>GEO-020 — Mixing claims, evidence and inference without labels<br>GEO-021 — Multiplying one source to make it appear as many<br>GEO-022 — Presenting paid visibility as independent authority<br>GEO-023 — Concealing a source’s true provenance and control<br>GEO-024 — Treating data with no disclosed methodology as conclusive evidence<br>GEO-025 — Treating undated evidence as current evidence<br>GEO-026 — Making a statement larger than its source can carry<br>GEO-027 — Filling an absence of evidence with a favourable assumption

CHAPTER IV — ERRORS OF MEASUREMENT AND ATTRIBUTION

GEO-028 — Claiming improvement without a baseline<br>GEO-029 — Treating mention, citation and recommendation as the same metric<br>GEO-030 — Changing the query set after seeing the result<br>GEO-031 — Treating a single answer as system behaviour<br>GEO-032 — Ignoring differences among models, countries, languages and interfaces<br>GEO-033 — Attributing sales directly to AI influence<br>GEO-034 — Treating a user’s self-report as verified attribution<br>GEO-035 — Selecting successes and concealing failures<br>GEO-036 — Hiding measurement uncertainty behind a single score

CHAPTER V — ERRORS OF CONTENT AND INFORMATION ARCHITECTURE

GEO-037 — Presenting different material realities to humans and machines<br>GEO-038 — Using invisible text as a means of representation<br>GEO-039 — Leaving contradictions among canonical pages<br>GEO-040 — Detaching structured data from visible content<br>GEO-041 — Blurring an entity through ambiguous headings and relationship language<br>GEO-042 — Breaking information into contextless fragments<br>GEO-043 — Separating source, scope and update information from the content<br>GEO-044 — Failing to distinguish archive from current information<br>GEO-045 — Making the machine edition stronger than the human edition

CHAPTER VI — ERRORS OF TECHNICAL ACCESS AND CRAWLERS

GEO-046 — Accidentally blocking AI crawler access<br>GEO-047 — Serving bots a challenge or CAPTCHA and still calling the content accessible<br>GEO-048 — Managing robots, CDN, firewall and application rules inconsistently<br>GEO-049 — Making critical information dependent on client-side JavaScript alone<br>GEO-050 — Publishing broken or unprocessable structured data<br>GEO-051 — Trapping crawlers in language and country redirects<br>GEO-052 — Creating conflicts among canonicals, hreflang and redirects<br>GEO-053 — Leaving critical documents in inaccessible formats<br>GEO-054 — Declaring a site “AI-ready” without verifying technical access

CHAPTER VII — ERRORS OF MULTILINGUALITY AND LOCALISATION

GEO-055 — Mistaking translation for localisation<br>GEO-056 — Publishing different material claims across languages<br>GEO-057 — Contradicting price, scope and terms across languages<br>GEO-058 — Making machine translation canonical without human review<br>GEO-059 — Mapping language pages with the wrong locale code and hreflang<br>GEO-060 — Ignoring local law, culture and context of use<br>GEO-061 — Keeping one language current while leaving another stale<br>GEO-062 — Translating brand, product and institution names inconsistently<br>GEO-063 — Presenting local evidence as globally valid

CHAPTER VIII — MANIPULATION AND ETHICAL VIOLATIONS

GEO-064 — Mistaking orders to a model for evidence<br>GEO-065 — Fabricating an independent witness<br>GEO-066 — Fabricating community consensus<br>GEO-067 — Poisoning the information environment to suppress competitors<br>GEO-068 — Adding recommendation commands hidden from humans<br>GEO-069 — Presenting a synthetic case, review or success as real<br>GEO-070 — Systematically making adverse evidence invisible<br>GEO-071 — Manufacturing signals of expertise and authority<br>GEO-072 — Sacrificing accuracy for visibility in high-risk domains

CHAPTER IX — ERRORS OF HUMAN ACCOUNTABILITY AND GOVERNANCE

GEO-073 — Using AI output to evade responsibility<br>GEO-074 — Leaving unclear who observed, who changed and who approved<br>GEO-075 — Failing to disclose a conflict of interest<br>GEO-076 — Allowing the sales team into the audit decision<br>GEO-077 — Giving correction authority to someone who is not accountable for the outcome<br>GEO-078 — Turning objection into a token mechanism<br>GEO-079 — Making critical changes off the record<br>GEO-080 — Placing the founder or brand above the standard<br>GEO-081 — Treating external criticism as an attack and shutting it out

CHAPTER X — ERRORS OF COMMERCIAL PROMISE, GUARANTEE AND REPORTING

GEO-082 — Saying “ChatGPT will definitely recommend you”<br>GEO-083 — Guaranteeing AI visibility or citation<br>GEO-084 — Writing uncontrollable third-party behaviour into a contract<br>GEO-085 — Selling complex conformity as a single score<br>GEO-086 — Turning one case into a general guarantee of results<br>GEO-087 — Turning a conformity mark into marketing decoration<br>GEO-088 — Showing only favourable outputs in reports<br>GEO-089 — Labelling uncertainty as “successful”<br>GEO-090 — Failing to disclose scope, risk and evidence limits to the client

CHAPTER XI — ERRORS OF MONITORING, MAINTENANCE, AUDIT AND CONFORMITY

GEO-091 — Removing time from evidence<br>GEO-092 — Leaving stale information active<br>GEO-093 — Failing to retest after a correction<br>GEO-094 — Changing audit scope after the result<br>GEO-095 — Treating an untested area as conforming<br>GEO-096 — Failing to distinguish adviser, auditor and issuer of a mark<br>GEO-097 — Treating a conformity mark as valid indefinitely<br>GEO-098 — Failing to establish suspension and withdrawal mechanisms<br>GEO-099 — Failing to audit the standard itself