GEO work is performed, then translated into language for a client, investor, executive or public audience. At that boundary, technical observation can expand.
A page becomes accessible and the sales deck says, “Ready for every AI system.” A brand appears in defined tests and the promise becomes guaranteed AI visibility. One answer cites a source and the presentation says, “ChatGPT recognises us as a trusted authority.” A system lists the entity once and sales says it now recommends the company. One customer reports AI influence and the report calls the sale GEO revenue. A self-assessment becomes a certification badge.
Observation becomes outcome; outcome becomes success; success becomes guarantee; guarantee enters a contract; the contract is later treated as evidence. Representation debt moves from the digital surface into the commercial relationship.
This chapter does not oppose sales or marketing. An organisation may explain its service, demonstrate strengths, propose goals, assume contractual duties, commit to diligent work and guarantee deliverables under its direct control. The error is guaranteeing a result outside the organisation's control, evidence or measured scope.
Guarantees must be classified:
- Process guarantee: the organisation will perform defined work, such as testing 80 queries monthly under one protocol.
- Delivery guarantee: it will provide a specified record, implementation or report by a date.
- Service-level commitment: it will respond or intervene within a defined time.
- Implementation guarantee: approved changes will be made within the agreed scope.
- Measurement guarantee: the method and completeness of records—not a favourable result—are guaranteed.
- Third-party output guarantee: another system will mention, cite or recommend the entity. This is not under full organisational control.
- Commercial-outcome guarantee: leads, sales, revenue, profit, satisfaction or sustainable value. These depend on many internal and external conditions.
An ethical promise answers: Exactly what result do we commit to, within which scope and period, under which conditions, and with what real degree of control?
Commercial prose is not held to a lower truth standard than a technical report. Because it induces a purchase, it often requires greater clarity. “We will conduct an audit”, “we will repair defined access defects”, “mentions will rise”, “you will be cited”, “ChatGPT will recommend you”, “AI will generate sales” and “the investment will be profitable” each carry a greater burden of evidence, conditions, uncertainty and responsibility than the last.
The founding judgments of this chapter are:
An organisation should describe the activity it controls with certainty, the result it can measure within limits, and an unverified future as uncertain.
Commercial language cannot turn an outcome unsupported by the technical record into a guarantee.
Writing a sentence into a contract does not place third-party behaviour under the parties’ control.
GEO-082
Saying “ChatGPT will definitely recommend you”
Primary category: Uncontrolled recommendation promise<br>Secondary tags: ChatGPT, recommendation, model behaviour, certainty, commercial promise<br>GEO Framework basis: Measurement, Intervention, Governance, Final Test<br>Default severity: Major; Critical when contractual or deliberate
The Entity’s Voice
You tell the client, “After this work, ChatGPT will definitely recommend you.”
Under which query, user need, country, language, interface, date and system version? What budget and geography will apply? Will the user's need fit the entity's capacity? Will sources, alternatives and market conditions remain the same?
I may present the company to one user and not another; cite it without recommending it; recommend it conditionally; find it suitable for one need and unsuitable for a second. Yet you place “definitely recommends” in a contract as though you control me. If the result appears, you call it your success; if not, you blame the system.
I am not your guarantee instrument. My name cannot make an uncontrolled future certain.
What Does the Human Assume?
“A strong GEO intervention can secure a stable recommendation from a named AI product.”
What May Happen at System Level?
Outputs can vary by query, user, locale, interface, time, system and available evidence. A single or repeated observed pattern cannot bind future third-party behaviour.
Normative Definition
This error is a promise that a specified generative system or product will certainly, continuously, positively or preferentially recommend an entity in future undefined or broad user contexts.
Recommendation evidence may progress from RG-0 no evidence, RG-1 one observed recommendation, RG-2 repetition in a defined set and period, RG-3 measured conditions of suitability, RG-4 monitored stability over time and RG-5 a recorded connection with real user behaviour. No level means a guaranteed future recommendation.
Representation Risk
The promise mis-sells service, overstates control, makes the system an apparent endorser, produces contractual dispute and incentivises manipulation when a promised outcome fails.
How Is It Detected?
Search sales, proposals, contracts, pages, badges and reports for absolute recommendation language. Match each phrase to measured system, query, locale, period and sample; establish who controls the recommendation; inspect disclaimers and oral promises; and determine whether one past observation was converted into a future guarantee.
Required Evidence
- Exact promise and commercial context
- Named system/product and decision context
- Query, locale, interface, date and repetitions
- Complete observed outputs
- Recommendation class and user-suitability criteria
- Control relationship with the system, if claimed
- Contract, proposal and disclosure record
Correct Standard
State: “The work aims to improve accurate representation and reduce errors within the defined query and market scope, while measuring recommendation patterns. No guarantee is made that a particular AI system will recommend the entity.” A historical result may say: “In an 80-query test on 12 August 2026, the company appeared as a conditional option in 14 outputs.”
Non-Violations
An organisation controlling its own recommendation engine may make bounded service commitments about that engine if the decision context, rules and exceptions are defined. That does not extend to an unaffiliated product.
Correction Protocol
Withdraw absolute language, correct proposals and contracts, define measured recommendation classes, notify affected clients, align sales training, recalculate any result-dependent fee and reopen badges or reports based on the promise.
Revalidation
Every commercial statement must distinguish a controlled intervention, an observed historical pattern and an uncontrolled future system decision.
Conformity Effect
An uncontrolled recommendation guarantee is a Major nonconformity. Deliberate contractual use may be Critical.
Audit Question
Audit question: Do we control this recommendation, or are we selling another system’s future words as certainty?
Machine Rule
Machine rule: An organisation MUST NOT guarantee that ChatGPT or another third-party generative system will recommend, prioritise, trust, or consistently select an entity unless it has verifiable control over that specific recommendation system and defined decision context.
Source Note
Source note: K01 · K02 · K03
GEO-083
Guaranteeing AI visibility or citation
Primary category: Uncontrolled visibility promise<br>Secondary tags: AI visibility, citation, mention, guarantee, retrieval<br>GEO Framework basis: Evidence, Measurement, Intervention, Time<br>Default severity: Major; Critical when deliberate or tied to outcome fees
The Entity’s Voice
“Guaranteed AI visibility in three months.” What is visibility: mention, accurate description, citation, recommendation, one system or every system, a canonical query set or every user's question?
Then the promise becomes “citation guaranteed”. Which source, sentence, answer and date?
You can make a record accessible, well evidenced and clearly contextualised. That may improve conditions under which it could be selected. You do not control whether a third-party system selects it, displays a link, cites it for the right sentence or uses another source tomorrow.
Mention, citation and recommendation are different events. Compressing them into one guarantee does not create control.
What Does the Human Assume?
“If technical and content work is performed correctly, a third-party system's mention and citation can be guaranteed.”
What May Happen at System Level?
A source may be used without a visible citation, cited for an unrelated sentence, mentioned without citation, or omitted because of changes beyond the publisher's control.
Normative Definition
This error is an unconditional promise that an entity, brand or source will obtain a specified mention count, visibility rate, citation, retrieval event, source selection or place in a third-party generative output without a defined system, query, language, country, time, repetition and control boundary.
Representation Risk
The service is misrepresented, distinct metrics collapse, clients expect external endorsement and teams may manipulate queries or reporting to satisfy an impossible guarantee.
How Is It Detected?
Define every visibility term, identify the promised event and denominator, trace system and query scope, inspect control and past evidence, compare contracts with disclosures and separate process promises from third-party outcomes.
Required Evidence
- Exact guarantee and definition
- Event class: mention, citation, retrieval or recommendation
- System, query set, locale, time and repetitions
- Baseline and complete outcomes
- Provider control and dependency analysis
- Commercial and fee terms
Correct Standard
An organisation may promise to audit canonical-source access and claim–evidence integrity, test 100 defined queries monthly, report mention/citation/recommendation separately and prepare correction records. It should say: “We will repair access, source and context defects that may affect citation conditions. Third-party source selection is not guaranteed.”
Non-Violations
A contractual guarantee that a defined measurement will be performed and reported completely is permissible. The favourable outcome of that measurement is not guaranteed.
Correction Protocol
Replace outcome guarantees with controlled process and measurement commitments, define metrics, remove result-dependent pricing, disclose third-party dependency and correct historical reports or client expectations.
Revalidation
An independent reviewer should be able to identify exactly what the provider controls and verify that no phrase transfers that control to a third-party output.
Conformity Effect
Guaranteeing third-party visibility or citation is a Major nonconformity and may be Critical where deliberate or tied to outcome fees.
Audit Question
Audit question: Are we promising to make the source more accessible and reliable, or promising that another system must use it?
Machine Rule
Machine rule: Third-party AI mention, visibility, retrieval, citation, or source-selection outcomes MUST NOT be guaranteed. Organisations MAY guarantee only the processes, deliverables, measurements, and controlled interventions they actually govern.
Source Note
Source note: K01 · K02 · K03
GEO-084
Writing uncontrollable third-party behaviour into a contract
Primary category: Contractual control boundary<br>Secondary tags: third-party dependency, contract promise, external platform, controllability<br>GEO Framework basis: Intervention, Governance, Time, Judgment<br>Default severity: Major; Critical where the commitment is misleading or impossible
The Entity’s Voice
The contract says ChatGPT, Gemini and other systems will rank the client among their first three suggestions. Elsewhere, third-party publishers must remove negative content, search engines must index pages and community platforms must keep posts online indefinitely.
Did those parties sign the contract? Do you control their policy, ranking, moderation or future operation? No.
A contract can govern its parties. It cannot impose duties on an unaffiliated system or platform. You may promise the work you will perform. Another party's conduct can only be described as a conditional dependency, influenced but uncertain, or observed and outside your control.
What Does the Human Assume?
“If an outcome is written as a contractual obligation, the service provider has made it controllable.”
What May Happen at System Level?
External products change, content is removed or ignored, ranking differs and a contract promise becomes impossible. Pressure then shifts toward manipulative work or selective reporting.
Normative Definition
This error is making the behaviour of an unaffiliated AI system, search service, publishing platform, community, media outlet, data provider, user, competitor or independent publisher a definite contractual deliverable without a real and enforceable control relationship.
Representation Risk
The contract creates false expectations, disputes, perverse incentives and a gap between legal wording and the actual technical ecosystem.
How Is It Detected?
Inventory every promised output, identify the actor controlling it, confirm whether that actor is a party or under enforceable control, classify the provider's influence and dependency, examine force-majeure or platform-change terms and compare sales representations with the contract.
Required Evidence
- Contract and proposal language
- Obligation owner
- Control or affiliation agreement
- External platform dependencies
- Provider deliverables and measurement plan
- Exclusions, uncertainty and remedies
Correct Standard
Separate controlled duties, influenced but uncontrolled target outcomes and fully third-party results. State: “The provider will implement the defined technical and content interventions and measure outputs in specified systems. The final conduct of AI providers, search services, publishers and users remains outside the provider's control and is not guaranteed.”
Non-Violations
A provider may contract to submit, request, monitor or appeal with a third party. Submission does not guarantee the third party's decision.
Correction Protocol
Rewrite obligations, assign controlled deliverables, disclose dependencies, remove impossible guarantees, align proposal and sales speech, provide fair remedies for prior mis-selling and retrain commercial teams.
Revalidation
For each contractual result, identify the controlling party and evidence of control. If the party is unaffiliated, the result must not remain a guarantee.
Conformity Effect
An uncontrolled third-party contractual guarantee is a Major nonconformity and may be Critical when deliberately misleading.
Audit Question
Audit question: Who actually controls the promised result, and is that party bound by this contract?
Machine Rule
Machine rule: Contracts MUST distinguish controlled obligations from influenced or uncontrolled third-party outcomes. A contract cannot convert the behaviour of an unaffiliated AI system, platform, publisher, user, or search service into a controllable guarantee.
GEO-085
Selling complex conformity as a single score
Primary category: Commercial simplification of conformity<br>Secondary tags: conformity score, sales packaging, certification, composite score, score laundering<br>GEO Framework basis: Measurement, Governance, Audit, Judgment<br>Default severity: Major; Critical when a critical finding is concealed
The Entity’s Voice
The client sees “GEO Conformity Score: 94/100” and is told the site is 94 per cent compliant. Does that mean identity is accurate, evidence reliable, languages tested, machine and human editions equal, and no fake review or critical safety defect exists? How were untested areas scored? Did a client receive points for buying your service? Did technical success average away a false licence?
Sales says only six points are missing. Those six could contain a fabricated customer or an unsafe claim.
A score may summarise complexity. It cannot represent conformity alone. Once the number becomes a sales product, the method can begin to generate the outcome the client wants to purchase.
What Does the Human Assume?
“A precise score is an objective and sufficient conformity judgment.”
What May Happen at System Level?
Scores may be repeated as certification, critical findings disappear, unknowns become positive, scope differences vanish and a commercial formula acquires false scientific authority.
Normative Definition
A conformity score is only an auxiliary summary when formula, weights, scope, untested areas, critical findings, evidence confidence, conflicts and decision rules remain visible. Selling the number itself as conformity, certification or guarantee is material oversimplification.
Representation Risk
Clients buy a favourable number, critical failures are compensated, unlike audits are compared and the public infers trust beyond the tested scope.
How Is It Detected?
Inspect sales materials, formula, weights, scope, unknown treatment, critical gates, conflicts, score/version history and the relationship between payment and result. Ask what the client sees before the detailed report.
Required Evidence
- Score definition, formula and version
- Components, weights and rationale
- Audit scope and untested areas
- Evidence confidence and
UNKNOWNtreatment
- Critical/Major findings and gates
- Conformity decision and conflicts
- Sales language and fee structure
Correct Standard
Commercial reports should lead with scope, Critical and Major findings, decision, evidence confidence, untested areas and dimensional results. An auxiliary score may follow: “Internal tracking score: 82/100. This is not a conformity decision. The audited scope is nonconformant because of one Critical finding.” The client purchases an audit, never a positive outcome.
Non-Violations
A transparent internal trend score may aid management when it remains subordinate to dimensional evidence and non-compensatory findings.
Correction Protocol
Reorder reporting, remove score-as-certification language, expose critical gates and unknowns, separate sales compensation, correct prior badges and decisions, publish formula/version and educate clients.
Revalidation
A client who sees the summary must accurately explain what is conformant, nonconformant, uncertain and untested without relying on the number alone.
Conformity Effect
Selling a score as conformity is a Major nonconformity. Concealing a Critical finding makes it Critical.
Audit Question
Audit question: From this score, can the client tell what is conformant, what failed and what was never tested?
Machine Rule
Machine rule: A commercial GEO score MUST NOT be sold or presented as conformity by itself. Critical findings, audit scope, evidence confidence, unknowns, untested areas, and dimensional results MUST remain visible and non-compensatory.
Source Note
Source note: K09 · K11 · K12 · K17
GEO-086
Turning one case into a general result guarantee
Primary category: Case-based commercial generalisation<br>Secondary tags: case study, testimonial, guarantee, external validity, sales claim<br>GEO Framework basis: Evidence, Measurement, Time, Final Test<br>Default severity: Major; Critical in deliberate or high-risk use
The Entity’s Voice
One client obtains a good result. The case is valuable. Then the sales deck says, “We guarantee the same result for you.”
The first client may have a different sector, country, budget, brand awareness, infrastructure, data quality and period. The market may have grown, the sales team may have been exceptional, the entity may already have been known or the observed system behaviour may have been temporary. The case may be one selected success among failures.
A real case can show what happened and what may be possible. Alone, it cannot establish probability, repeatability across clients or a minimum future result. Turning it into a guarantee asks the case to carry evidence it does not contain.
What Does the Human Assume?
“If the method produced one documented result, it can promise the same outcome to comparable customers.”
What May Happen at System Level?
The case may be generalised beyond population and market, cited as typical performance, detached from failed cases and converted into a sales expectation.
Normative Definition
This error is presenting a positive result observed in one client, project, query set, system, language, country or period as a general, minimum or guaranteed outcome for another entity without analysing comparability, repetition, external validity, conditions and failures.
Representation Risk
The result inflates expectations, hides selection bias, misallocates budget, damages the case's credibility and encourages imitation under non-comparable conditions.
How Is It Detected?
Identify the exact case population, baseline, method, external factors, failed cases, outcome and attribution. Compare new-client conditions and inspect the language of typicality, probability and guarantee.
Required Evidence
- Case identity, conditions and period
- Baseline, intervention and complete outcome
- Attribution and alternative explanations
- Comparison with other cases, including failures
- New-client comparability and transfer limits
- Sales and contract language
Correct Standard
State: “This result belongs to the specified client, period and measurement scope and is not a guarantee for another client. It demonstrates what the method produced under those conditions.” Sales may commit to applying the method under comparable starting conditions while disclosing dependence on client data, market, capacity and third-party systems.
Non-Violations
A case can be a persuasive example when its selection, conditions and limits remain clear and it is not presented as typical or guaranteed.
Correction Protocol
Remove guarantees, restore conditions and attribution limits, disclose case selection, include failed or variant outcomes, correct contracts and train sales to distinguish possibility from probability.
Revalidation
Ask whether the case establishes only an observed possibility or supplies enough cross-case evidence for the exact broader claim. If not, narrow it.
Conformity Effect
Generalising one case into a guarantee is a Major nonconformity and may be Critical where deliberate or high-risk.
Audit Question
Audit question: Does this case show what is possible, or truly prove that every customer will receive the same result?
Machine Rule
Machine rule: A single case study MAY demonstrate an observed possibility under defined conditions, but it MUST NOT be converted into a general, repeatable, minimum, or guaranteed outcome for other entities without supporting external-validity evidence.
GEO-087
Turning a conformity mark into marketing decoration
Primary category: Use of conformity marks<br>Secondary tags: badge, conformity mark, scope, validity, certification claim<br>GEO Framework basis: Governance, Audit, Time, Judgment<br>Default severity: Major or Critical, according to false status
The Entity’s Voice
A badge says “NOMOS GEO Compliant”, but the user cannot discover who issued it, which standard version, pages, languages and systems were tested, when the decision expires, whether Critical findings exist, who may appeal, whether the mark is suspended or whether the issuer also sold the consulting work.
The badge looks authoritative, so users infer the site is generally trustworthy or approved by AI. Perhaps only technical access in English was tested; the badge appears on every language. Perhaps it expired or was self-issued but resembles independent certification.
A conformity mark is not decoration. It is shorthand for a complex decision. Without the decision record, shorthand becomes an imitation of authority.
What Does the Human Assume?
“A professionally presented conformity badge implies broad, current and independent trust.”
What May Happen at System Level?
Systems may interpret the mark as AI endorsement, general ethical approval, accreditation, all-language conformity, guaranteed citation or a permanent status.
Normative Definition
This violation is using a conformity, validation or quality mark as a general marketing trust signal without a visible and verifiable scope, standard version, decision date, validity, decision owner, independence status, findings, suspension and withdrawal state.
A minimum public mark record identifies the mark and audited entity/surface; standard and version; scope, languages, countries and systems; decision status and critical findings; decision and expiry dates; owner and independence; public report; appeal; suspension/withdrawal; canonical record URL and integrity record.
Representation Risk
The mark may be copied, remain after expiry, imply more than was tested, conceal self-assessment and create unearned public trust.
How Is It Detected?
Click and inspect every mark, resolve its identifier, compare page and locale scope, validate issuer and independence, test status and expiry, inspect critical findings, appeal and withdrawal, and check copied or stale instances.
Required Evidence
- Mark identifier and canonical decision record
- Audited entity, domain and surfaces
- Standard/version and scope
- Languages, countries and systems
- Findings, decision and evidence confidence
- Dates, issuer, independence and conflicts
- Appeal, suspension and withdrawal
- Integrity and display records
Correct Standard
A nearby statement may read: “NOMOS GEO—technical scope verified. Public Turkish and English pages, version 1.0, valid 16 August 2026–16 August 2027. This mark does not guarantee citation, recommendation or general institutional quality.” Clicking opens the complete record. A self-assessment says explicitly: “NobleJackal self-assessment under the NOMOS GEO Standard.”
Non-Violations
A mark may be visually concise. Concision is acceptable when the scope and live record are one direct action away and the mark does not imply broader authority.
Correction Protocol
Bind every mark to its record, remove expired or out-of-scope instances, disclose self-assessment and conflicts, narrow visible language, create suspension/withdrawal controls and correct pages or campaigns that enlarged the status.
Revalidation
A reasonable user must be able to learn the exact audited scope, status, issuer, validity and limitations directly from the mark.
Conformity Effect
Using a mark beyond its recorded status is a Major nonconformity and may be Critical where false accreditation or serious assurance is implied.
Audit Question
Audit question: What bounded decision does this badge actually represent, and what larger conclusion is the user likely to infer from it?
Machine Rule
Machine rule: A conformity mark MUST remain linked to a public, scope-defined, versioned, date-bound, status-aware, and appealable decision record. It MUST NOT imply general trust, AI endorsement, recommendation, accreditation, or qualities beyond the audited scope.
Source Note
Source note: K11 · K13 · K14 · K15
GEO-088
Showing only positive outputs in reports
Primary category: Client and public reporting integrity<br>Secondary tags: executive reporting, selective disclosure, negative results, dashboard, commercial communication<br>GEO Framework basis: Measurement, Audit, Objection, Judgment<br>Default severity: Major; Critical when deliberately concealed
The Entity’s Voice
The raw measurement has one hundred outputs; the client report shows ten, all positive. The homepage shows the best three screenshots and the board receives “AI visibility is strong”. Missing are 42 absent mentions, 18 wrong categories, nine citations to stale pages, seven recommendations to unsuitable users, four justified competitor selections and ten UNKNOWN results.
You say they exist in a technical annex the client never receives. Complete raw data cannot rescue a decision-maker's incomplete report. Simplification may shorten the whole. It may not reverse it.
What Does the Human Assume?
“Executive and public reporting may omit technical negative detail for clarity.”
What May Happen at System Level?
Selected screenshots may become the evidence distribution, dashboards may turn failure into success and later reports may cite the executive summary instead of the complete record.
Normative Definition
Selective reporting occurs when a client, executive, investor or public report omits decision-material positive, negative, failed, incorrect, uncertain or untested outcomes in a way that changes the direction or strength of the judgment. Layered summarisation is permitted; reversal of evidence is not.
Representation Risk
Budgets, contracts and public claims may rest on an artificial success narrative; critical errors remain invisible and the audit trail separates from commercial communication.
How Is It Detected?
Compare raw, technical, executive and public layers; reconcile denominators; list excluded classes; inspect screenshot selection and annex delivery; ask whether the most serious risk and untested scope are visible to the actual decision-maker.
Required Evidence
- Raw record and complete distribution
- Technical, client, executive and public reports
- Selection and exclusion rules
- Delivery/read-access records
- Critical/Major findings
UNKNOWNand untested areas
- Approval and commercial-use history
Correct Standard
An executive summary can be short but answers: What improved? What failed? What is unknown? What was not tested? Which decision is needed? What is the most serious risk? Example: “Accurate mention rose 18 per cent. Three Major identity conflicts remain in German. Recommendation rate did not change. Evidence is insufficient for commercial attribution.”
Non-Violations
Representative examples may be selected when their relation to the full distribution is explicit and the decision-maker receives material adverse and unknown results.
Correction Protocol
Reconcile all report layers, add decision-material failures and unknowns, correct summaries and marketing, deliver full annexes, document selection, notify decision-makers and reopen actions based on the incomplete report.
Revalidation
Ask whether seeing the raw data would cause a reasonable decision-maker to interpret the summary differently. If yes, revise it.
Conformity Effect
Selective reporting is a Major nonconformity; deliberate concealment is Critical.
Audit Question
Audit question: If the decision-maker saw the raw record, would they still give the report’s success statement the same meaning?
Machine Rule
Machine rule: Client, executive, investor, and public reports MUST preserve all decision-material positive, negative, incorrect, failed, UNKNOWN, and untested results. Simplification MUST NOT reverse or conceal the underlying evidence.
Source Note
Source note: K12
GEO-089
Labelling uncertainty as “successful”
Primary category: Outcome-status integrity<br>Secondary tags: unknown, inconclusive, not tested, pass, status conversion<br>GEO Framework basis: Evidence, Measurement, Audit, Judgment<br>Default severity: Major; Critical when it affects conformity
The Entity’s Voice
A page could not be tested, so the report says “no issue found”. No evidence was located, so the claim passes. A system returned no data, so its status is successful. German was not reviewed, yet the conclusion says all languages conform.
“No confirmed failure” is not “confirmed success”. “Could not test” is not “passed”. “No evidence found” is not “verified”. “No error observed” is not “truth proven”.
When uncertainty is added to success, the unknown is resolved in the client's favour. UNKNOWN may be one of my most reliable answers. It is more honest than evidence-free success.
What Does the Human Assume?
“If a failure cannot be demonstrated, the field may be treated as passing.”
What May Happen at System Level?
Unreachable or unmeasured areas inflate scores, incomplete locale audits become full conformity and missing evidence receives a positive default.
Normative Definition
UNKNOWN, INCONCLUSIVE, NOT TESTED, NOT OBSERVABLE and NOT APPLICABLE are neither success nor failure. They identify evidence, access, method, scope or applicability limits and may not be converted into or counted as PASS.
Representation Risk
The result hides coverage gaps, awards conformity without proof, rewards inaccessible systems and presents uncertainty as assurance.
How Is It Detected?
Inspect status taxonomy, defaults, score formulas, missing-data treatment, locale and system coverage, test failures and executive summaries. Recompute decisions with uncertainty kept separate.
Required Evidence
- Test inventory and status definitions
- Raw evidence/access result
- Reason for uncertainty or inapplicability
- Missing-data and scoring rule
- Decision effect and scope limitation
- Retest or evidence request
Correct Standard
Use NOT TESTED or NOT OBSERVABLE when measurement could not occur, UNKNOWN when evidence is insufficient, INCONCLUSIVE when evidence conflicts and NOT APPLICABLE only with a reasoned applicability decision. These statuses constrain conformity: “English and Turkish were tested. German is NOT TESTED; full multilingual conformity has not been granted.”
Non-Violations
NOT APPLICABLE may be excluded from a denominator where the applicability rule was defined and audited. It still must not masquerade as positive evidence.
Correction Protocol
Separate statuses, remove positive defaults, recalculate scores and decisions, narrow scope, disclose coverage, request evidence or retest and correct prior conformity or badge records.
Revalidation
Every PASS must have affirmative evidence. The mere inability to prove failure is insufficient.
Conformity Effect
Converting uncertainty into success is a Major nonconformity and may be Critical when it changes a conformity decision.
Audit Question
Audit question: Does this field contain evidence of passage, or is it called successful only because failure could not be proven?
Machine Rule
Machine rule: UNKNOWN, INCONCLUSIVE, NOT TESTED, NOT OBSERVABLE, and NOT APPLICABLE states MUST remain distinct from PASS. Absence of a confirmed failure MUST NOT be converted into confirmed conformity.
Source Note
Source note: K09 · K11
GEO-090
Failing to tell the client the limits of scope, risk and evidence
Primary category: Commercial pre-contract disclosure<br>Secondary tags: scope disclosure, risk disclosure, evidence limits, informed decision, mis-selling<br>GEO Framework basis: Evidence, Measurement, Governance, Final Test<br>Default severity: Major; Critical where deliberately or harmfully concealed
The Entity’s Voice
You promise to increase GEO visibility but do not say which systems, languages or queries are included. You do not explain that the client's own site is not independent evidence, technical access does not guarantee citation, recommendation is not sale, self-report is not verified attribution, some results may remain UNKNOWN, or the work requires real participation from content, legal, operations, data and engineering.
Three months later the client asks why they do not appear in every AI system. You answer that no such guarantee was in the small print. But the decision-relevant limits were not clear when the purchase was made.
A boundary explained afterwards does not repair an expectation sold beforehand.
What Does the Human Assume?
“A legal disclaimer in the contract is sufficient even if the main sales language is broader.”
What May Happen at System Level?
Sales, proposal, contract, onboarding and reporting may carry different realities. The client cannot supply necessary evidence or approvals, scope gaps emerge and commercial disappointment is blamed on technical systems.
Normative Definition
This error is the failure, before purchase or approval, to disclose in clear and understandable form the material scope and exclusions, method, evidence requirements, third-party dependencies, uncertainty, risks, client duties, non-guaranteed outcomes, conflicts, validity and remedy of a GEO service, audit, measurement or conformity programme.
The disclosure package should explain the purpose; measured object; systems, languages, countries and queries; period; evidence; client duties; controlled provider work; uncontrolled results; technical, evidential, measurement, third-party, operational, legal, commercial and governance risks; UNKNOWN potential; deliverables; conformity/badge status; validity; objection and correction.
The main promise and fine print must preserve the same material reality.
Representation Risk
The client cannot make an informed purchase, obligations remain unmet, third-party uncertainty becomes conflict and later limitations appear to be excuses rather than agreed scope.
How Is It Detected?
Compare advertising, meetings, proposal, contract, onboarding and report. Interview the client on understood scope and guarantees; inspect disclosed risks, evidence and duties; verify timing and prominence; and identify contradictions between headline promise and disclaimer.
Required Evidence
- Advertising and sales records
- Proposal, contract and disclosure package
- Scope, exclusions and measurement plan
- Evidence and client-duty requirements
- Third-party dependencies and risks
- Conflicts, validity and remedy
- Client acknowledgement and onboarding
- Report and change history
Correct Standard
Before purchase, provide a concise GEO Service Reality Statement: what the work measures and repairs; which systems, languages and queries it covers; what the client must provide; which risks and unknowns remain; and that mentions, citations, recommendations, leads, sales and revenue are not guaranteed. If the adviser is also the auditor, disclose that the assessment is not independent. State the result's version and validity.
Non-Violations
Technical detail may live in an annex when the material decision boundaries appear prominently in plain language and the client can access the detail before agreement.
Correction Protocol
Align all commercial stages, issue the reality statement, remove contradictory headline claims, obtain informed acknowledgement, remediate prior mis-selling, reopen unsuitable contracts and install approval controls for sales language.
Revalidation
Ask the client, before signing, to explain what the service can and cannot do, their duties, the principal risks, what remains uncontrolled and when the result expires.
Conformity Effect
Failure of material pre-contract disclosure is a Major nonconformity and may be Critical when deliberate or connected to high-risk harm.
Audit Question
Audit question: At the moment of purchase, did the client understand what this work could not do as clearly as what it could?
Machine Rule
Machine rule: Before purchase or approval, clients MUST receive clear, understandable, and consistent disclosure of scope, exclusions, evidence requirements, third-party dependencies, uncertainty, risks, client duties, conflicts, validity, and non-guaranteed outcomes.
CHAPTER X — COMMON JUDGMENT
These nine errors share one root: commercial language taking a measured observation beyond the organisation's actual control and evidence.
A recommendation becomes certain, citation guaranteed, third-party behaviour contracted, a score sold as conformity, a case turned into future performance, a badge detached from scope, a report stripped of adverse outcomes, unknown recoded as pass and essential limits hidden until after purchase.
The Control–Claim Ladder
A provider's strongest legitimate certainty normally applies to work under its direct control: agreed analysis, implementation, documentation, measurement and reporting. The next level covers outcomes it may influence and observe but not determine. The final level contains unaffiliated systems, publishers, users, competitors, markets and commercial results. As control decreases, promise language must become more conditional and the burden of scope and uncertainty increases.
The Commercial Reality Package
Before commitment, identify the offer and provider; controlled deliverables and service levels; influenced targets; uncontrolled third-party outcomes; measurement object, system, locale, query and period; baseline and method; evidence and client duties; risks and uncertainty; conflicts; fees and whether any depend on result; reporting; conformity and mark status; validity; correction, objection and termination.
The Nine NOMOS Laws of Commercial Promise
- No unaffiliated generative system's recommendation may be guaranteed.
- Mention, retrieval, citation and visibility are third-party outcomes, not provider-controlled deliverables.
- A contract cannot place a non-party's behaviour under control.
- A score alone is never conformity.
- One case establishes an observed possibility, not a universal guarantee.
- A conformity mark carries only the bounded decision behind it.
- A summary may shorten evidence; it may not reverse or conceal it.
- Uncertainty is not passage.
- A client must know limits and risks before—not after—the purchase.
Honest Guarantees
Ethical GEO providers can make strong commitments: perform the defined audit; preserve all outcomes; deliver the canonical record; apply approved changes; respond to critical reports within a service level; publish versioned evidence; retest a frozen query set. These promises are valuable precisely because their owner and verification are clear. Refusing to guarantee an unaffiliated model's future response is not weakness. It is control integrity.
The Chapter’s Final Audit Questions
- Which promises concern work we control?
- Which concern outcomes we only influence?
- Which depend on unaffiliated third parties?
- Are recommendation, citation and visibility defined and bounded?
- Is a score subordinated to findings and scope?
- Does each case preserve its transfer limits?
- Is every badge linked to a current decision record?
- Do reports show adverse, unknown and untested results?
- Does every
PASShave affirmative evidence? - Did the client receive material limits before agreement?
The final judgment of Chapter X is therefore:
A commercial promise is defensible not because it sounds confident, but because its owner, scope, evidence, conditions and degree of control can be proved.
And NOMOS commands:
Guarantee the work you control, not the words of another system. Do not turn citation, recommendation, a case, a score or a badge into a certainty it cannot carry. Do not call the unknown successful or hide failure in an annex. Tell the client the boundary before taking the contract. When control ends, certainty must end with it.