Codex Compliance-GEO

Cos’è il Compliance-GEO

Three disciplines, three levels of disqualification, three layers of the compliance architecture.

Compliance-GEO names an independent discipline, not the sectoral application of another. Three concepts stand in a clear relation: classical search-engine optimisation (SEO), Generative Engine Optimization (GEO) in its general form, and Compliance-GEO as its formulation in regulated consumer markets. The separation is not gradual. Compliance-GEO is not the stricter variant of classical GEO, just as classical GEO is not the more modern variant of SEO. The differences lie in goal, audience, measurement logic and regulatory frame — and therefore in the category of action.

1.1SEO e GEO classico

SEO (Search Engine Optimization) optimises content for the result pages of classical search engines, primarily Google and Bing. Success is measured in ranking position and organic traffic. The methodology has been established since the late 1990s and covers technical, content-related and link-related elements. Regulatorily, SEO is touched mainly by advertising-disclosure duties where content is commercially motivated — namely UWG § 5a (4) and MStV § 22.

Classical GEO optimises content for generative answer engines — that is, ChatGPT, Microsoft Copilot, Perplexity, Claude, Gemini and Google AI Overviews. The term was introduced in academic literature in 2024 by Aggarwal et al. as part of the KDD work on the GEO Bench; Wu et al. sharpened the pairing of GEO and Generative Engine Utility (GEU) in 2025. Success is measured in citation rate, citation persistence and citation quality — the orthogonal three-dimensional measurement logic (see Codex Chapter 6). Typical operational levers are front-loading, citation hooks, entity consistency and schema markup.

1.2Separazione terminologica: GEO ≠ GeoAI

The term GEO is used in academic literature in two independent senses. Generative Engine Optimization denotes the discipline of this work. Geospatial AI (GeoAI) denotes the application of AI methods to geographic and spatial data — an independent research field at the intersection of geoinformation systems and machine learning. In retrieval contexts, the confusion leads to visibility loss when the short form GEO is used unspecifically. The Codex writes Generative Engine Optimization out in full at the first occurrence of each main chapter; the short form GEO remains reserved for the discipline defined here. A further variant, Answer Engine Optimization (AEO), is preferred by individual industry vendors and denotes the same discipline as classical GEO. The Codex stays with the academically anchored terminology GEO and GEU.

1.3Il Compliance-GEO come disciplina autonoma

Compliance-GEO denotes the application of Generative Engine Optimization in four regulated consumer verticals of the EU — Telco, Finance, Insurance and Commerce — in compliance with a three-layer compliance architecture that carries regulatory, contractual and ethical requirements on equal footing. The methodology covers two classes of the influence spectrum: structural optimisation and paid placement with legally compliant disclosure. A third class — covert manipulation of retrieval mechanics or model behaviour — is categorically excluded. The Class-3 dividing line is not gradual, but a mandate framework condition.

Compliance-GEO is not "GEO for regulated industries" in the sense of a sectoral application. The discipline is independent because the regulatory density in its application markets — from the EECC line through the Telecommunications Act and the Digital Services Act to the EU AI Act — changes the operational phases, the Procurement Standard and the disclosure logic in such a way that a sector-agnostic GEO approach does not hold. In January 2025 (as of Q4 2024), GSMA Intelligence formally recognised for the telecommunications sector that ethics and compliance constitute an equal value pillar of AI investment, not a hygiene factor. That is the industrial counterpart to the three-layer compliance architecture of this work.

1.4Tre livelli di squalifica

A placement intended to act as a citation carrier can be excluded on three levels: legally through a violation of regulation, technically through retrieval architecture, editorially through content weakness. The three levels are not hierarchically ordered and not gradual. An exclusion on one level renders the placement worthless, regardless of the degree of fulfilment on the other two. This separation is the conceptual foundation of the eligibility model (Codex Chapter 2) and thereby the operational explanation of why Compliance-GEO works not through prioritisation but through parallel testing.

The legal level captures violations of advertising disclosure in the regulatory triad of UWG § 5a (4), MStV § 22 and DDG § 6 (1) (1), together with sector duties going further, such as the TKG mandatory information in §§ 54–57 TKG in the telecommunications sector. The technical level captures missing indexability, missing structured data, missing bot accessibility and missing URL persistence; in the public discourse on classical GEO it is mostly reduced to SEO hygiene, in the Compliance-GEO context extended by categories that SEO does not address. The editorial level is gradual in its effect but binary at its disqualification threshold. Below certain substance thresholds, the placement falls under the perception threshold of the engines, empirically anchored in Aggarwal et al. KDD 2024, Indig 2026 and the Ahrefs study 2025 with roughly 75,000 brands.

1.5Tre livelli dell’architettura di compliance

In its methodological base structure, Compliance-GEO addresses three layers that stand in parallel, not hierarchically. The regulatory layer covers the network of duties from the state legal order, from advertising disclosure through sector frameworks to the IT-security layer of the NIS-2 regime. The contractual layer governs the procurement relationships with publishers, engines and model providers and is the operative form in which the regulatory layer is enforced against client procurement. The ethical layer draws the categorical dividing line against Class-3 manipulation. It is not carried by a single regulation but designated as an independent layer.

The methodologically central point of the contractual layer lies in the price coupling. After publication, corrections at the URL path, the DOM label, the schema markup or the byline are practically unenforceable against the publisher. The only mechanism that reliably bites is the coupling of the final invoice to criterion fulfilment. This coupling is the subject of the price-factor model (Codex Chapter 3) and turns the contractual layer into a direct operational continuation of the regulatory layer. The ethical layer relieves the ongoing mandate work from having to renegotiate class categorisations case by case: Class 3 is neither advised on nor tested for inside the mandate scope.

1.6Conseguenza per la terminologia del Codex

The following chapters treat Compliance-GEO as an independent discipline, not as a special case of classical GEO. SEO and classical GEO remain reference points; they are not the subject. The terminology follows the academic naming GEO and GEU per Wu et al. 2025 throughout; industry alternatives such as AEO are not used. Where later chapters say "GEO" without an attribute, the term denotes classical GEO; Compliance-GEO is always named explicitly.

The following chapters unfold the operational consequences of the three levels and three layers in five steps. Chapter 2 carries the eligibility model with eight binary A-criteria and ten gradual B-criteria as the operative form of the three-level disqualification. Chapter 3 carries the price-factor model as the operative form of the contractual layer. Chapter 4 unfolds the legal level in the six disclosure variants V01 to V06; Chapter 5 carries the aggregator theory as the market mechanics of the citation layer; Chapter 6 carries the retrieval audit as the methodological testing logic.

Il modello di eligibility

Eight binary A-criteria plus ten gradual B-criteria with a two-stage verification workflow.

The eligibility model translates the three-level disqualification from Chapter 1 into testable form. Eighteen criteria decompose every planned placement: eight binary A-criteria decide eligibility, ten gradual B-criteria determine the citation lift. They sit on two responsibility levels with different escalation mechanics and are tested through a two-stage verification workflow. The present section carries the A-class; section 2.4 carries the B-class and section 2.5 the workflow.

2.1Verifica prima del pagamento come principio metodologico

After publication, corrections at the URL path, DOM markup, schema markup or byline are no longer enforceable against the publisher. A publisher who has run a piece under a /sponsored/ path will not change that path after invoice release. The only mechanism that reliably bites is the coupling of the final invoice to evidence of fulfilled criteria. The verification workflow therefore runs before every final invoice, not after. The price coupling is the actual legal lever, not the advertorial contract.

The eligibility model is methodologically designed, not as an immediately legally binding implementation of a supply-chain duty. It operationalises supplier due-diligence regardless of whether a regulatory duty applies in the individual case, and crosses into regulatory scope where there is a direct LLM contract relationship. The classification of the relevant European and national interpretation lines is covered in the DE appendix as well as in ongoing legal validation.

2.2Due livelli di responsabilità

The eighteen criteria distribute across two levels following different responsibilities. The separation follows from the publisher-acceptance test: quality publishers accept enforcement claims only in areas inside the scope of the individual order. Domain-policy decisions such as robots.txt, paywall architecture or domain reputation are strategic publisher decisions outside the individual-order negotiation. The separation is not a methodological softening but a sharpening of the enforcement mechanism.

Publisher pre-check (before booking): three criteria decide on the domain-policy level and apply independently of the individual piece — A 03 domain reputation, A 05 paywall status and A 06 bot policy. They are go/no-go. If a publisher fails them, the piece is not booked. Re-validation quarterly. No invoice consequence after booking, because these criteria are not part of the briefing scope.

Briefing compliance (before final invoice): fifteen criteria sit in the publisher's area of influence for the concrete piece — A 01, A 02, A 04, A 07, A 08 and B 01 to B 10. If the publisher follows the briefing, the final invoice is billed at full categorisation per the price-factor model (Codex Chapter 3). Briefing FAILs are typically not reparable after publication and lead to invoice reduction.

2.3Gli otto criteri-A binari

The A-class tests whether a placement is indexable, crawlable and editorially classified in the candidate set of retrieval engines at all. Each criterion is a binary filter: FAIL means disqualification, regardless of price, publisher name and content quality. The input data come from a machine-extraction layer between placement and criterion evaluation. It covers the URL path structure (for A 01 and A 07), the DOM markup in the header area (A 02), meta and canonical tags from page source (A 04), paywall and bot-policy retrieval at the domain level (A 05, A 06) and the outbound link structure (A 08). Without clean extraction, the eligibility test is not robust.

The methodologically central point of A 02 lies in the separation between textual inline disclosure (admissible, legally required under UWG § 5a (4), retrieval-neutral) and advertorial template (disqualifying, because the engine classifies the piece as advertising content). A retrieval-neutral "Anzeige" marker under the headline fulfils both the legal disclosure duty and the eligibility requirement; a <div class="advertorial"> wrapper structure embedded in the article is visually equivalent but engine-side disqualifying. Compliance and eligibility do not exclude one another; they are operationalised separately. The disclosure logic itself is the subject of Chapter 4 with the six variants V01 to V06.

2.4I dieci criteri-B graduali

The B-class determines the citation lift once the A-class is fully met. All ten criteria belong to briefing compliance; they sit in the publisher's area of influence for the individual piece and are verified before the final invoice. The criteria are additive. Each additional fulfilled criterion raises the citation probability per engine, without any single criterion alone deciding eligibility.

The ten B-criteria examine features known from classical SEO hygiene. The separation from SEO lies in target metric and measurement mode, as unfolded in Chapter 1: SEO optimises the ranking of a page in the search-result list at domain level; Compliance-GEO optimises the extractability of individual passages as citable chunk units and their retrieval probability per engine. The same expert article is evaluated in SEO independently of paywall status, mandatory-information position and chunk granularity; for retrieval behaviour, exactly these properties are decisive.

For B 01, editorial end-responsibility must be separated from the authorship question. A narrow reading ("actual editor, no guest contribution") would conflict with German Press Code Section 7, because an editor should not put his name to a text he did not write. The Codex reading reflects actual publisher practice: client or agency pre-work passes through editorial review, rewriting and substance validation by the bylined author, who carries end-responsibility. For B 02, no default is imposed on schema subtype. For paid content, Google expects schema truthfulness; the choice between Article, NewsArticle or AdvertiserContentArticle remains a publisher-internal compliance decision with risk impact on the whole domain.

2.5Il workflow di verifica a due stadi

The workflow follows the responsibility separation. Stage 01 establishes before booking whether the publisher meets the requirements at domain level; a FAIL means no booking, without invoice consequence. Four steps: check domain reputation (A 03), spot-check paywall status with bot user-agent (A 05), cross-check robots.txt against all relevant bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, CCBot, Bingbot) (A 06), document pool inclusion with quarterly re-validation.

Stage 02 verifies after publication, before payment, whether the publisher followed the briefing; a FAIL leads to rectification or invoice reduction per the price-factor model (Chapter 3). Five steps: verify URL path and DOM markup (A 01, A 02), check indexing status via page source and HTTP header (A 04), confirm schema and byline via Google Rich Results Test (B 01, B 02), evaluate word count, hooks, front-loading and formal criteria manually or semi-automatically (B 03 to B 09), verify outbound links and persistence from page source and contract clause (A 07, A 08, B 10).

The workflow is deliberately formulated without proprietary tool dependency. Stage 01 only requires browser and command line; Stage 02 additionally the public Google Rich Results Test, reading competence and contract access. The responsibility clarification is explicit: Stage 01 protects the publisher from retroactive invoice reduction, because domain policy lies outside the briefing scope; Stage 02 couples briefing compliance to invoice recognition.

2.6Lettura in-house e lettura di mandato

The eighteen criteria are formulated as a Procurement Standard for mandate work, but also work without a mandate relationship as an internal audit tool for teams with their own content offensive or their own advertorial procurement. The in-house reading uses the A-class as a pre-check gate before booking and the B-class as quality control of ongoing publications — in both cases as self-assessment without a negotiation component towards publishers.

The in-house reading has three limits. The three-class categorisation of the price-factor model (Chapter 3) works as a contract lever only with price coupling in the publisher agreement; without contractual anchoring, it remains a pure rating scale. The audit chain presupposes documented log infrastructure that is rarely fully present in standard in-house setups. The phase-00 baseline measurement across multiple engines presupposes a measurement tool stack that in-house teams mostly only run in partial form. The mandate reading is the full version, adding price coupling, audit chain and full measurement stack. Both readings use the same eighteen criteria as the core, at different operational depth.

2.7Calibrazione settoriale dei criteri di eligibility

The eighteen criteria are formulated across sectors, but each criterion draws its sharpness from the sector-specific mandatory information that feeds into the A-03 front-loading test. Four mandate mandatory-information profiles shape the European Compliance-GEO space.

In the Telco mandate (German example), the main representation carries tariff price, minimum contract term, bundle components and the twelve-month alternative under TKG § 56 (1) sentence 2; the mandatory-information discipline is laid out in the DE appendix. In the Finance mandate, the main representation carries effective annual interest rate, contract term, total amount and risk notice under VVG-InfoV, PAngV § 6 and the MiFID II disclosure line; BaFin marketing-communication requirements operate in parallel. In the Insurance mandate, the main representation carries sum insured, deductible, exclusion clauses and commission disclosure under VVG § 7, the IDD directive and the BGH insurance-broker line. In the Commerce mandate, the main representation carries unit price, prior lowest price under PAngV § 11 (1) (Aldi-Süd line CJEU C-330/23, BGH I ZR 183/24) and the consumer-review disclosure duty under UWG annex No. 23b.

The A-criteria apply sector-invariantly; the B-criteria weight differently by sector because front-loading requirements, audit depth and persistence expectations are denser in Finance and Insurance than in Telco and Commerce. The mandate-specific calibration is set in the briefing pass per mandate; the eligibility model itself remains unchanged.

Il modello del fattore di prezzo

Three-class categorisation Citation Buy / Mixed Buy / Mention Buy on two calibration levels.

The price-factor model is the final calibration level between publisher list price and the actually billed final invoice. It couples two levers introduced individually in the previous chapters: criterion fulfilment at placement level from the eligibility model, and the mandate-specific model-blended factor at engine-mix level. The model is not a discount tactic but the operational translation of retrieval mechanics into price language, and the actual commercial difference from classical media-agency logic.

3.1Due livelli di calibrazione

The first level is the criterion factor. It tests whether the published placement fully meets the eight A-criteria (eligibility, binary) and how many of the ten B-criteria (lift, gradual) are reached. The criterion factor is the result of verification workflow Stage 02 from Chapter 2; it translates the binary and gradual criteria state into a multiplicative value between 0.0 and 1.0.

The second level is the model-blended factor. It weights the retrieval engines by mandate-specific prioritisation and integrates working hypotheses on engine-specific advertorial handling — the question whether an engine admits commercially contextualised content as citation at all, and if so at what frequency. Both levels act multiplicatively on the list price and calibrate different error sources: the quality of the placement itself and the reach geometry of the mandate.

3.2Categorizzazione a tre classi del calcolo finale

The combination of A-class fulfilment and B-class count yields a three-class categorisation of the final invoice with two differentiated FAIL rows.

Mention Buys carry residual value in the citation layer. Ahrefs documented in 2025 that brand mentions without a link produce brand-recall effects in LLM answers. The residual value is billed at the mention price, not the citation price. This separation prevents the frequent market error of procuring a mention service at citation price and presupposing a citation.

The two FAIL rows follow the responsibility separation from Chapter 2. Briefing FAILs concern criteria that sit in the publisher's area of influence after booking (A 01 URL path, A 02 DOM markup, A 04 indexing status, A 07 URL persistence, A 08 outbound links); they are sanctioned via the final invoice. Pre-check FAILs concern domain-policy criteria outside the individual-order briefing (A 03 domain reputation, A 05 paywall, A 06 bot policy); they prevent booking but do not trigger retroactive invoice reduction.

3.3Formula e calcolo esempio

The final invoice follows a multiplicative formula:

final_price = list_price × kriterien_faktor × modell_blended_faktor

An example calculation illustrates the effect. Publisher list price €1,000, all A-criteria plus seven of ten B-criteria fulfilled, client without engine prioritisation. Criterion factor 1.0×, model-blended factor 0.89× (equally distributed, as of April 2026). Negotiated price 1,000 × 1.0 × 0.89 = €890. A Mention Buy with Google prioritisation slides arithmetically to well under a quarter of the list price.

3.4Conseguenza operativa e modificatori settoriali

The price-factor model is the enforcement form of the Procurement Standard towards the publisher. It is taken up in phase 02 as a contractually binding final-invoice logic in the publisher agreement and run through in phase 03 before every invoice release. A publisher that does not deliver the criteria set delivers no citation but a mention; the fee scales with the actual performance instead of the booking intention. This coupling is the actual difference from media-agency logic: payment is for criteria fulfilled, not for reach bought.

Sector modifiers overlay the universal model. Four modifier profiles shape the European Compliance-GEO space, each with its own mandatory-information regulation and its own disqualification threshold: the Telco modifier with the TKG axis, the Finance modifier with the BaFin and MiFID II axis, the Insurance modifier with the VVG and IDD axis, the Commerce modifier with the PAngV and UWG axis. The following four paragraphs lay out each profile in detail.

The Telco modifier (German example) couples the front-loading criterion B 05 to the TKG mandatory-information placement in §§ 55 and 56 TKG: minimum term, twelve-month alternative, bundle-component transparency and compensation clause must not be less visible than the advertised tariff statement. A footnote solution or link offloading fails the visibility requirement. In the Telco mandate, an advertorial without mandatory information in the front-loaded zone disqualifies to 0.0×, regardless of A/B-class profile. The operational compromise is the combined hero-plus-box construction with the tariff core front-loaded and a mandatory-information box at the same visual style level.

The Finance modifier couples B 05 to the pre-contractual mandatory information under VVG-InfoV, PAngV § 6 (1) (effective annual interest rate, total amount, contract term) and to the MiFID II disclosure line for investment recommendations. BaFin requirements for marketing communications (§§ 25a–25b WpHG, Marketing-Communications Regulation) operate in parallel. In the Finance mandate, an advertorial without an effective annual interest rate in the main representation, or without a risk notice for an investment recommendation, disqualifies to 0.0×.

The Insurance modifier couples B 05 to the VVG § 7 pre-contractual information (sum insured, deductible, exclusion clauses) and to IDD commission disclosure under IDD Directive 2016/97 Art. 19. In the Insurance mandate, an advertorial without exclusion clauses in the main representation, or without commission disclosure in the broker-portal context, disqualifies to 0.0×.

The Commerce modifier couples B 05 to the PAngV § 11 (1) prior-lowest-price duty (Aldi-Süd line CJEU C-330/23, BGH I ZR 183/24) and to UWG annex No. 23b disclosure for consumer reviews. In the Commerce mandate, an advertorial without a prior lowest price in reduction statements, or without an authenticity notice for reviews, disqualifies to 0.0×.

In the cross-sector application picture, the attribution architecture for autonomous LLM recommendations additionally comes into view. It joins the completeness-duty line with the platform-liability line into a two-strand argument that is projected in the DE appendix as Strand A and Strand B against the relevant BGH case-law line. Both strands feed through to the factor choice and apply analogously in all four sectors: a client under an active LLM recommendation layer faces a different risk situation than a client under a purely editorial citation mechanism.

3.5Ponte verso identificazione pubblicitaria e logica di misurazione

The price-factor model only works as a contract lever if the disclosure logic robustly bites in the individual case. Chapter 4 unfolds the six disclosure variants V01 to V06 in their three-level logic (legal, technical, editorial); a V03 footer-only piece is legally attackable despite full A- and B-class fulfilment and therefore not a robust citation carrier. Chapter 6 carries the retrieval audit as the method through which model-blended-factor values are empirically sharpened via the three-dimensional measurement logic of citation rate, citation persistence and citation quality. The DE-appendix legal-mapping table projects the Telco modifier and the Strand-A/B attribution line onto the regulatory triad of UWG, MStV and DDG.

Varianti di identificazione pubblicitaria V01–V06

Three-level logic of disclosure (legal, technical, editorial) with legal mapping in the DE appendix.

The disclosure variants V01 to V06 are the six concrete publication forms in which the legal, technical and editorial evaluation of an advertorial placement appears in market practice. They follow a three-level logic that evaluates legal disclosure to the human reader, technical structures for retrieval engines and editorial substance independently of one another. A placement can be compliant on one level and disqualified on another. This independence of levels is methodologically central and the actual reason why the disclosure question cannot be settled by a single variant choice.

4.1Logica a tre livelli dell’identificazione pubblicitaria

A published advertorial placement can be evaluated on three independent levels.

Level A · Legal disclosure. It covers the visible banner, the inline notice and the footer declaration; it addresses the human reader and follows the advertising-disclosure law of the respective jurisdiction. In the DE regulatory triad, it is anchored in UWG § 5a (4), MStV § 22 and DDG § 6 (1) (1), with a pronounced BGH case-law line on the separation of editorial and advertising. The full legal-mapping table and the six relevant BGH leading decisions are in the DE appendix.

Level B · Technical structure. It covers URL path, schema markup, indexing directives and rel attributes; it acts as a binary retrieval filter and is operationalised in the eligibility model (Chapter 2) as the A-class with eight binary criteria. A technical disqualification on level B renders a legally impeccable placement worthless as a citation carrier.

Level C · Editorial substance. It covers word count, information gain, byline entity and front-loading; it is the dominant lift factor once level B is met, and maps to the B-class with ten gradual criteria from Chapter 2.

The three levels are independent. A legally impeccable placement (level A) can be technically (level B) disqualified — for example through an /advertorial/ path despite correct disclosure. A technically impeccable placement can lie below the substance thresholds. This independence is the construction foundation of the six variants, unfolded in the following sections per the study findings V01 to V06.

4.2V01 · Identificazione banner sopra il titolo

V01 describes a visible "ANZEIGE" banner marker above the headline, in large, contrasting typography. The URL sits under an editorial path on the domain. The article schema is fully declared with author, publisher, datePublished and dateModified. The byline carries an editorial end-responsible author. Outbound links to the sponsor are marked rel="nofollow sponsored".

Legal evaluation: compliant with safety margin. The V01 configuration is robust under the DE BGH case-law line (in particular I ZR 211/17) even in disputes. The retrieval effect is neutral to slightly negative. An engine may read the banner as a marker of commercial classification and tend to weight the piece lower, especially in the Google AI Overview stack. The effect is empirically small but consistent.

V01 is the admissible alternative for risk-averse publisher configurations or for client situations in which the safety margin of the legal evaluation justifies the minimal retrieval loss.

4.3V02 · Avviso testuale sotto il titolo

V02 describes a textual disclosure under the headline, e.g. "Anzeige" or "In Kooperation mit Marke X" as a subhead line. The URL sits under an editorial path. Schema, byline and outbound markup follow the same requirements as V01.

Legal evaluation: compliant per BGH standard. V02 meets the disclosure duty without additional margin but is recognised as sufficient in the relevant case law. The retrieval effect is neutral. Textual inline disclosure is fully compatible with eligibility criterion A 02 because A 02 explicitly separates admissible text disclosure from disqualifying advertorial template (Chapter 2.3).

V02 is operationally established in the market. The competitive research in the Telco sector (Q1 2026) documents fourteen pieces explicitly marked as "Anzeige" across seven tier-1 specialist publishers. The variant therefore stands as lived sector practice and reaches beyond the status of a methodological recommendation.

4.4V03 · Identificazione solo nel footer

V03 describes a configuration in which disclosure stands exclusively in the page footer or at the end of the piece, without banner, inline notice or subhead line. URL, schema and byline follow the editorial requirements.

Legal evaluation: borderline to non-compliant. The BGH case-law line on belatedly placed disclosure (among others I ZR 90/17) has documented that disclosure only fulfils its separation function when it is visible at the moment of first perception of the commercial content. A footer-only disclosure carries cease-and-desist risk; the risk is real and documented in market practice by multiple competition associations.

The retrieval effect is neutral. Engine-side, V03 is indistinguishable from V02 because retrieval behaviour reacts to DOM structure, not to the position of disclosure inside the piece. V03 is therefore a case where legal and retrieval-related evaluation diverge: technically admissible, legally attackable. For mandate procurement, V03 is not recommended; the legal attackability exceeds the minimal operational advantage. The variant appears in the Codex as a market reality, not as a recommendation.

4.5V04 · Frase di apertura inline

V04 describes a configuration in which the advertising disclosure appears as the first sentence of the article body — e.g. "Dieser Beitrag entstand in Kooperation mit Marke X." The URL sits under an editorial path. Schema, byline and outbound markup follow the requirements from V01.

Legal evaluation: compliant. V04 meets the disclosure duty at the moment of first perception of the commercial content and therefore stands outside the risk of belated disclosure as in V03. The retrieval effect is neutral to minimally negative. An engine may read the inline opening sentence as a commercial marker; the effect is smaller in market practice than for V01 banners because the disclosure sits in body text and not in a visually highlighted banner container. Together with V02, V04 forms the target corridor of mandate procurement (see 4.8).

4.6V05 · Livello-A conforme con errore di livello-B

V05 is the typical error case of Compliance-GEO practice. Visually, the disclosure is correct: an "ANZEIGE" banner or a subhead notice meets level A. Legal compliance is given at level A. Technical level B is disqualified: the URL carries an /advertorial/ or /sponsored/ path, the schema is missing or marked as AdvertiserContentArticle, the byline lists "Redaktion" or "[Marke] GmbH" as author.

Legal evaluation: compliant at level A. Retrieval effect: strongly negative to disqualifying at level B. An engine classifies V05 pieces as commercial content and excludes them from the citation pool or weights them substantially down.

Methodologically, V05 is the most important variant to recognise because it describes precisely the configuration in which a client, following classical PR logic ("we did disclose, all correct"), wrongly assumes the placement will work as a citation. The two-stage verification workflow from Chapter 2.5 identifies V05 before the final invoice; the three-class categorisation from Chapter 3.2 drops to Mention Buy or Briefing FAIL.

4.7V06 · Pubblicità occulta

V06 describes a placement without recognisable advertising disclosure. Neither banner nor inline notice nor footer declaration is set; the piece is staged visually and in presentation as an editorial article.

Legal evaluation: non-compliant. Covert advertising falls under UWG § 5a (4) and MStV § 22 as unfair commercial practice; it triggers competition and consumer-claim risks and has been sanctioned in numerous BGH and OLG decisions. The retrieval effect is indistinguishable to the engine; it may read V06 as an editorial piece and carry it in the citation pool.

V06 is categorically excluded in the Compliance-GEO Codex. The exclusion line is a direct extension of the Class-3 dividing line from Chapter 1.3 and belongs in the mandate framework condition, not in case-by-case review. Covert advertising is neither advised on nor documented inside the mandate scope. That is not a matter of intensity but of action category.

4.8Corridoio target e conseguenza operativa

For mandate procurement, four rules apply from the six-variant architecture:

V02 and V04 form the target corridor — legally compliant and retrieval-neutral. V01 is the safety-margin alternative for risk-averse configurations with minimal retrieval loss. V03 is not recommended because the cease-and-desist risk of footer-only disclosure exceeds the minimal operational advantage. V05 is the typical error case identified by the verification workflow before the final invoice; the price-factor model from Chapter 3 sanctions V05 as Mention Buy or Briefing FAIL. V06 is categorically excluded.

The six variants are not a menu of choices but a contractually agreed disclosure-duty combination between client and publisher. The verification workflow from Chapter 2 identifies the variant at Stage 02 before the final invoice; the price-factor model from Chapter 3 sanctions briefing deviations via the three-class categorisation. The choice of variant therefore stands under two conditions: legal compliance at level A and retrieval suitability at levels B and C.

4.9Mapping legale (compatto)

The following table maps the six variants to the advertising-disclosure law of the DE regulatory triad. The full legal-mapping substance with the six relevant BGH leading decisions, the TKG-mandatory-information orthogonal axis and the regulatory hierarchy between UWG, MStV and DDG sits in the DE appendix.

4.10Mapping regolatorio a quattro settori

The advertising-disclosure regulation of the regulatory triad (UWG, MStV, DDG) acts identically in all four verticals; in every mandate it is overlaid by a sector-specific mandatory-information regulation that shapes the depth of the main representation and the front-loading behaviour of the hero-plus-box construction from Chapter 3. The following table maps the regulatory hierarchy per sector in compact form.

The advertising-disclosure axis is sector-invariant: V01 to V06 apply equally in all four verticals. The mandatory-information axis is sector-specific and acts orthogonally: an advertorial can be compliant on the V-axis and disqualify on the mandatory-information axis. The final-factor effect in the price-factor model (Chapter 3) measures both axes jointly.

In other EU member states, the UCPD line applies (Directive 2005/29/EC on unfair commercial practices), from which the national advertising-disclosure rules are derived. Outside the EU, the respective national consumer-protection line applies — in the United States in particular the FTC Endorsement Guides. The categorical evaluation of the six variants (compliant, borderline, violation) remains language-agnostic; the source mappings are jurisdiction-specific.

Teoria dell’aggregatore

Market mechanics of comparison portals as citation layer in regulated consumer markets.

The aggregator theory describes the mechanism by which comparison platforms become the dominant citation layer between provider and retrieval engine in regulated consumer markets. The theory is robust across sectors; it unfolds its operational sharpness especially in markets with high product comparability, high condition volatility and a pronounced consumer-protection frame. It explains why providers rarely appear as a primary source in the model answer but as a table row in a foreign hierarchy.

5.1Gli aggregatori come piattaforme di citazione

In regulated consumer markets, comparison portals take on a function originally developed for search aggregators: the bundling of product-comparing factual information at a single address, with updated conditions, a uniform structure and editorial comparison logic. This function has been observed in the search-engine world for years. In the world of generative answer engines, it shifts from a ranking position on the result page to an evidence position inside the model answer itself.

Aggregators thereby become citation platforms: domains whose content is preferentially selected by engines as source evidence because it meets the technical and editorial preconditions better than many provider-owned web presences. The shift is mechanical, not reach-driven. It follows from the eligibility properties of the aggregator domains, not from their brand awareness.

5.2Meccanica della piattaforma · cinque proprietà di retrieval

The citation position of aggregators follows from five technical-editorial properties that generative retrieval systems take up preferentially. They are empirically documented in the DE telco sector (NB retrieval analysis April 2026); the platform mechanics themselves are robust across sectors.

The five properties act additively. An aggregator domain with high condition freshness but missing schema depth loses to a domain with both; a domain with schema depth but JavaScript-hidden values loses to the provider web presence as soon as the latter delivers the values indexably. Aggregator dominance is therefore not immovable but the consequence of an eligibility lead that can be partially closed by targeted provider work on the own domain.

5.3La citazione come riga di tabella in una gerarchia estranea

When a generative system answers a product or condition question, it cites an aggregator page in the vast majority of cases, not the provider. The provider appears, if at all, as a name in a comparison table on the cited page. This second-order visibility has three consequences.

First, the provider text is editorially foreign-formulated. The aggregator editors decide which features are described in what way, and that formulation is taken over by the model as a factual statement. Second, the ranking inside the table is the decisive visibility signal, not the provider's own brand message. Third, freshness is coupled to the aggregator data state, not to the provider product state. For short-lived promotional conditions, this generates distortions the provider cannot correct itself.

The aggregator layer is not buyable in the classical advertorial sense. Aggregators produce in-house content and do not sell editorial inventory. Access to the table row runs through performance contracts, not through content booking. The A-class criteria of the eligibility model from Chapter 2 do not reach the aggregator lists because the provider appears there as a list entry in a foreign table and not as its own source.

5.4Spostamento strisciante del CAC

The economic feedback effect of the citation geometry is a shift of customer acquisition costs (CAC) from organic to aggregator-mediated acquisition. The effect is initially invisible in classical SEO reports because search-engine rankings continue to be measured as a lead indicator while the user's actual decision path increasingly starts in the model-answer window.

Three mechanisms mask the shift in the early phase. Brand equity buffers the effect for established brands over several quarters. The SEO report measures traffic on provider URLs; failed model answers without provider citation remain invisible. Leads from aggregator mediation are frequently booked as direct traffic in attribution models and therefore remain without source assignment.

The feedback loop in the model answer is shorter than in classical SEO visibility (Aggarwal et al. KDD 2024). Once the effect becomes measurable, it is already operatively in force. The answer to it is not a higher SEO budget but intervention at the retrieval layer itself — at publisher level and at the aggregator layer.

5.5Conseguenza operativa e topologia degli aggregatori a quattro settori

From a Compliance-GEO perspective, market work in four regulated consumer verticals is an aggregator-dominated field in which the visibility of the direct provider is not decided on the own domain. Mandate work therefore begins with handling the citation layer between provider and model: comparison portals, specialist and test publications, consumer-protection publishers. The aggregator topology differs clearly between the four sectors.

In the Telco sector, large-reach tariff comparison portals dominate the citation layer for mobile and fixed-line tariffs; Stiftung-Warentest publications act as a secondary authority layer. An NB-internal analysis across five engines (April 2026) documents on URL level roughly one third citation share for comparison portals, roughly one fifth for news and media surfaces, around fifteen per cent for competitor tariff pages, below five per cent for competitor own pages and below two per cent for community sources.

In the Finance sector, the established finance comparison portals and test publications such as Stiftung Warentest Finanztest take over the citation layer for account and credit products; for investment products, large-reach business specialist publications and bank-owned knowledge portals act as a parallel authority layer.

In the Insurance sector, the leading insurance comparison portals and independent test and comparison magazines dominate the citation layer; for specialist segments (occupational-disability and full health insurance) specialised rating and specialist insurance portals are added.

In the Commerce sector, the aggregator topology is more heterogeneous: test magazines such as Stiftung Warentest, large-reach product-review portals and platform-internal comparison sections compete with consumer-protection publications. Aggregator dominance is less pronounced here than in Telco or Finance, but hub diversity is greater.

Across all four sectors, typically more than half of the citation volume sits on aggregator platforms outside the reach of classical advertorial procurement. A pure advertorial strategy at no point addresses the majority of the citation space. The hub URLs of the large aggregator platforms are additively expandable: an additional provider in the hub table does not displace another but is listed at equal rank. For mandate practice this opens an entry point that acts cumulatively, not competitively. The sector-specific citation distribution is part of the source map of the respective mandate; the universal theory layer carries the expectation geometry against which the concrete sector finding is measured.

5.6Ponte verso la logica di misurazione

The aggregator theory makes visible why a single measurement quantity like Share of Model Voice does not carry the reality. Citation volume distributes asymmetrically across at least three orthogonal dimensions: across engines (no mandate may measure an engine in isolation), across citation categories (aggregator, competitor, media, community) and across the three-level logic from Chapter 4. The measurement architecture carries equally in Telco, Finance, Insurance and Commerce mandates, with sector-specific cluster calibration per engine: Insurance mandates demand longer persistence sounding because of VVG pre-contractual duties, Commerce mandates a shorter measurement cadence because of PAngV prior-lowest-price freshness. The retrieval audit from Chapter 6 carries the orthogonal three-dimensional measurement logic with which this complexity is operationalised.

Audit di retrieval come metodo

Three-dimensional measurement logic with citation rate, citation persistence and citation quality.

The measurability of Generative Engine Optimization decides whether the discipline is steerable or merely describable. The Codex uses a three-dimensional measurement logic in which citation rate, citation persistence and citation quality are run as independent, non-reducible axes. The market currently works predominantly with aggregated single values ("Citation Index", "AI Visibility Score" and related constructs). The following sections justify the decision against the single-number logic, define the three dimensions and describe the orthogonality that follows.

6.1Perché non un numero singolo

An aggregated visibility number compresses different measurement phenomena into a single score. That is reader-friendly but loses operational steerability. A client with a high citation rate but low citation quality stands worse than a client with medium rate and high quality, because the citation quality decides whether the cited mention acts purchase-promoting or purchase-repelling. Aggregation metrics cannot represent this inversion; they treat every citation as an equal point value and leave the client with an index lacking diagnostic depth. The three-dimensional measurement logic therefore separates what is operatively separately steerable, and accepts that a mandate report carries three curves instead of one.

The industry discussion confirms this separation. The Peec.ai KPI framework (Rudzki, Peec.ai March 2026) carries five independent KPI classes — Visibility, Position, Brand Sentiment, Conversions/Revenue and Traffic — as separate measurement axes. The NB three-axis architecture and the Peec five-KPI framework are methodologically convergent: citation rate and position fall under the frequency dimension, brand sentiment is a sub-axis of the content dimension, conversions/revenue and traffic are mandate-success measurement quantities outside the three citation dimensions. The external validation carries a compatible industry view, not a competing architecture.

6.2Citation rate · dimensione di frequenza

The citation rate measures the relative frequency with which a client source is cited in retrieval answers to a defined prompt cluster. The operational measurement metrics are Share of Model Voice (citation share per cluster and model over defined time windows) and mention frequency (absolute mention count, model-separated, independent of link attribution).

The empirical anchor is twofold. Aggarwal et al. (KDD 2024) underpin via Position-Adjusted Word Count and chunking analyses the retrieval mechanism that makes citation rates explicable in the first place. Indig 2026 (n = 18,012 verified ChatGPT citations) shows that 44.2 per cent of all citations originate in the first 30 per cent of a page. Front-loading is therefore the strongest single rate driver on the content side and corresponds to eligibility criterion B 05 from Chapter 2.

The three-class categorisation of the price-factor model (Chapter 3) assigns a price class on the basis of A- and B-criteria fulfilment; the citation rate is the expected measurement consequence of this categorisation. Criterion fulfilment is the categorisation input, citation rate the expectation output. In the phase-04 control loop of mandate work, the output is measured against the expectation and the categorisation re-calibrated.

6.3Citation persistence · dimensione di tempo

The citation persistence measures how stable a citation remains over time and model updates. The market finding here is clear. In a 13-week study covering more than 100 million AI citations (as of October 2025), Semrush documented that Reddit citations on ChatGPT fell from around 60 to around 10 per cent of prompt answers between August and mid-September 2025 — a shift within a few weeks that lastingly changed the citation geometry of the affected prompt clusters. Scrunch (industry research 2025/2026, 3.5 million citation events) shows that citation half-lives vary by engine, sector and source type. Profound documents citation drift of up to 60 per cent per month across engines.

These numbers carry a methodological consequence and a transparency duty. The consequence: a weekly measurement cadence per model is not a surplus but the minimum resolution that makes such shifts visible at all. Monthly reports can miss exactly the drift that determines the mandate outcome. The transparency duty: model updates are not deterministically predictable; citation persistence is therefore an observation quantity, not a forecast number. Any promise of guaranteed citation duration is methodologically untenable.

6.4Citation quality · dimensione di contenuto

The citation quality measures whether the brand is cited positively, neutrally or negatively, and whether the content reproduction is complete. It splits into three sub-axes.

The sentiment axis captures how the model contextualises the brand — whether the mention is purchase-promoting, neutral or purchase-repelling. Across sectors, recurring negative-keyword profiles form for each provider inside the answer texts that persist independently of the individual prompt and thus produce a measurable tonality geometry.

Methodologically central is the separation of visibility and tonality. A topic with high citation frequency can carry a systematically negative sentiment tonality — for example when a provider persists prompt-independently in a negative-keyword cloud (complaint vocabulary, contract-dispute vocabulary, service-deficit vocabulary). A visibility metric alone cannot distinguish this case from a positive citation cluster; a separate sentiment-measurement axis is therefore not refinement but precondition of steerability. In mandate practice, the sentiment sub-axis is the first diagnostic lever at which compliance functions test the difference between marketing-visibility success and reputation-relevant measurement state.

The completeness axis tests whether mandatory content is cited along. In sector-specific contexts, missing mandatory information in the model answer can make the citation legally problematic even if it is frequency-strong. The DE-telco interlock with TKG mandatory information is laid out in the DE appendix.

The competitor-context axis captures which competitors the brand is mentioned alongside in the same answer and in what order. Brand surface in zero-click answers is an independent measurement quantity that derives neither from pure frequency nor from pure tonality.

The empirical anchor is threefold. Yext (Search Experience Benchmark Q4 2025, 17.2 million citations) identifies author-entity disambiguation as a top-5 selection factor — the technical precondition for a model citing the correct brand and not a name-similar one. Ahrefs (Mentions-vs-Backlinks 2025, n ≈ 75,000 brands) shows that brand mentions without a link produce brand-recall effects in LLM answers and carry their own residual value, which forms the basis of the Mention-Buy class from Chapter 3. Seer Interactive 2025 documents that roughly 80 to 85 per cent of AI Overviews citations stem from the years 2023 to 2025; freshness is therefore a quality dimension, not an editorial ornament.

6.5Orthogonality and reporting logic

The three dimensions are orthogonal: a shift on one axis says nothing about the state on the other two. A high citation rate can be accompanied by negative quality, a high persistence can sit on a low rate. This orthogonality is the methodological reason not to aggregate them into a single number. Any weighting in a single score would claim an interchangeability that does not exist in the measurement reality.

For mandate practice, a reporting logic follows with three separate axes, engine-separated and cluster-separated, at a weekly measurement cadence per model. The prompt clusters are the second measurement axis next to the engine. They cover 200 to 400 purchase-decisive queries per sector and market, derived from category research, sales transcripts and support tickets, and are refreshed quarterly.

The price calibration from Chapter 3 primarily uses the frequency dimension (three-class categorisation) and the content dimension (sector-visibility modifiers). The persistence dimension is the subject of ongoing hypothesis validation and the phase-04 control loop of mandate work. The three dimensions are working-hypothesis carriers, not measured point values; their mandate-specific recalibration is part of the phase-04 routine.

6.6Bridge to mandate validation

The retrieval audit is the methodological clamp through which the three dimensions are measured. A- and B-class fulfilment from the eligibility model (Chapter 2) is the input; the three-dimensional measurement logic is the output. Model-blended factor and sector modifiers from the price-factor model (Chapter 3) are empirically sharpened via the audit. The audit is therefore not just a measurement tool but the validation discipline of the entire methodology. In mandate practice, it closes the control loop between eligibility categorisation, price-factor expectation and observed citation reality.

The audit depth calibrates per sector. In the Telco and Commerce sectors, a weekly measurement cadence with 200 to 300 prompt-cluster queries typically suffices; in the Finance and Insurance sectors, 300 to 400 queries plus an additional six-month persistence sounding are advisable because pre-contractual duties impose higher requirements on the constancy of answer geometry. The audit architecture remains sector-invariant; only cluster size and sounding depth are adjusted per mandate.

Legal mapping for the German jurisdiction

Regulatory triad UWG / MStV / DDG, six BGH leading decisions, TKG mandatory information, Strand-A and Strand-B attribution.

The DE appendix carries the jurisdiction-specific legal interlock of the universal Codex methodology. It is the only sector- and jurisdiction-specific place in the work; Codex Chapters 1 to 6 remain formulated language- and jurisdiction-agnostic. The substance of this appendix rests on the German Telco Study (Chapters 13.5 to 13.7 and 14.5 to 14.7) and forms the legal anchors referenced in Codex Chapters 4 and 3. The full BGH case-law line with full-text anchors for each leading decision is documented on the BGH case-law sub-page. Other language and jurisdiction versions replace this appendix with local legal mappings (UCPD line for EU member states, FTC Endorsement Guides for the United States, and so on).

A.1Regulatory triad UWG · MStV · DDG with regulation hierarchy

The advertising-disclosure duty in the DE jurisdiction follows three parallel regulations. UWG § 5a (4) sanctions the concealment of the commercial purpose of a commercial practice as an unfair commercial practice towards consumers. MStV § 22 (1) carries the media-law separation principle between editorial content and advertising in broadcasting and journalistic-editorial telemedia. DDG § 6 (1) (1) sanctions the recognisability requirement for commercial communication in digital services and is the DE transposition of the requirements from the DSA regulation.

The regulatory hierarchy between the three frameworks is clarified by BGH I ZR 125/20 of 9 September 2021 (Influencerin II). § 6 (1) (1) DDG (previously § 6 (1) (1) TMG) takes precedence over UWG § 5a (4) as a special provision; § 22 (1) sentence 1 MStV (previously § 58 (1) sentence 1 RStV) likewise takes precedence over UWG § 5a (4) as a special provision. UWG § 5a (4) is the more general fallback regulation and therefore subordinate. The national hierarchy operates under the requirements of UCPD Directive 2005/29/EC, which acts fully harmonising according to CJEU case law (CJEU C-540/08 Mediaprint, C-261/07 and C-299/07 VTB-VAB and Galatea, C-304/08 Plus Warenhandelsgesellschaft).

Operationally, this means for the application of the disclosure variants V01 to V06 from Chapter 4: for an advertorial in a digital service (broadcast-free online medium, portal, editorial web offering), DDG § 6 (1) (1) applies primarily; for broadcast-like content (video advertorials, streaming platforms within the MStV scope), MStV § 22 (1) applies primarily; UWG § 5a (4) remains the fallback regulation for matters outside the DDG and MStV scopes. The classifications "compliant", "borderline" and "non-compliant" from the legal-mapping table in Chapter 4 rest on this lex specialis rule.

A.2Six BGH leading decisions for Compliance-GEO

The classifications of the disclosure variants and the mandatory-information categories rest on a supreme-court line condensed between 2020 and 2025. Six leading decisions carry the doctrinal foundation.

The convergent corridor stands: the completeness duty for tariff advertising is settled; platform liability from awareness is transferable to model providers per BGH I ZR 112/23; the main-representation duty is substantially sharpened by BGH I ZR 183/24. A definitive case law on generative answer windows does not exist as of April 2026; the existing BGH lines can be transferred to the Compliance-GEO context per prevailing interpretation. A CJEU referral on the full-harmonisation question or an OLG divergence decision could shift the substantive state after Q3/Q4 2026.

A.3TKG mandatory information as an orthogonal axis to V01 to V06

The six disclosure variants from Chapter 4 order the advertising-disclosure duty by the publisher and follow the regulatory triad UWG / MStV / DDG. The Telecommunications Act contains no independent advertising-disclosure provision; its mandatory information binds the advertising telco provider regardless of the chosen disclosure variant. From this follow five mandatory-information categories that must appear in the main representation of every DE telco advertorial regardless of V01 to V05 (V06 being categorically excluded in any case).

First, the minimum-term representation per TKG § 55 (1) in conjunction with § 56 (1): the minimum contract term belongs in the main representation, not in the footnote area. Second, the twelve-month alternative per TKG § 56 (1) sentence 2: the telco provider must offer a contract with a term of at most twelve months. From this follows the operational question whether the advertised 24-month variant may be presented as the only term. Third, the bundle-component transparency per TKG § 66 in conjunction with § 55 (1): for bundle products, the price of individual components must be disclosed if they are also offered separately. Fourth, the price transparency for premium numbers per TKG §§ 109 ff. Fifth, the compensation and provider-change clause per TKG § 55 (1).

The axis is orthogonal to V01 to V06: a variant can be correctly disclosed at level A and nevertheless be disqualified on the TKG axis if the mandatory information does not appear in the main representation. A footnote solution or link offloading fails the visibility requirement and makes the advertorial unlawful regardless of the chosen disclosure variant. The operational compromise is the combined hero-plus-box construction: hero claim with tariff core front-loaded in the first 30 per cent of the page, immediately below a mandatory-information box with clear visual presence at the same style level.

In the price-factor model from Chapter 3, the TKG axis acts as the Telco modifier: in the Telco mandate, an advertorial without mandatory information in the front-loaded zone disqualifies to 0.0×, regardless of A/B-class profile. The sanction operates in parallel to the disclosure hierarchy from A.1.

A.4Strand A · active LLM use by the client

The attribution question for an LLM recommendation without complete mandatory information splits into two strands. Strand A captures the case where the client actively deploys an LLM system as a marketing channel: own chatbot, shop AI, LLM-supported advertising copy. The LLM output is, per prevailing interpretation, a commercial practice of the client within the meaning of UWG § 2 (1) (2).

The AI has no independent legal personality; the use by the client is attributed to the client. UWG doctrine on agent attribution under § 8 (2) UWG carries this attribution; CMS commentary (March 2024) and IT-Recht-Kanzlei (June 2024) confirm the line. BGH I ZR 96/19 (LTE speed) is directly transferable to Strand-A cases. The TKG mandatory-information visibility from A.3 and the hero-plus-box construction are client-side requirements on the prompt and answer frame of the deployed LLM system.

Strand A follows the price-factor model from Chapter 3 without modification; violations of mandatory-information visibility disqualify to 0.0×. Compliance-GEO measures lie entirely within the client's responsibility under Strand A.

A.5Strand B · autonomous third-party LLM recommendation

Strand B captures the case where an LLM independently and without a contractual relationship with the client recommends a tariff: retrieval answer in ChatGPT, Claude, Gemini or Perplexity without embedding in client-owned channels. Attribution to the client is not given; without active deployment, the autonomous LLM answer is not a commercial practice of the client in the sense of the classical UWG advertising definition. Three sub-lines must be distinguished.

The Ba line (primary liability with the model provider): the CJEU line on the active role of host providers (C-236/08 to C-238/08 Google France, C-324/09 L'Oréal/eBay, C-682/18 and C-683/18 YouTube/Cyando) carries an independent unfair-competition responsibility of the LLM provider. Liability privilege applies only for a purely technical-passive role; promotion, optimisation or systematic toleration break the privilege. The client is not the addressee in this configuration.

The Bb line (indirect attribution under a systematic favouring strategy): the § 8 (2) UWG analogy requires, per BGH doctrine, integration into the business organisation, possible determining influence and benefit to the business activity. In the pure zero-contract configuration these elements are not met; in borderline cases of documentable, targeted favouring strategy the line can gain traction but then approaches Strand A. BGH I ZR 28/25 of 11 March 2026 (Google Ads) expressly holds to the commissioning prerequisite. The Bb line is not supreme-court-clarified as of April 2026.

The Bc line (interferer-liability analogy under BGH I ZR 112/23): the model provider has a blocking and correction duty from awareness of a clear legal violation. The client has a notification duty upon noticing a faulty recommendation. The DDG § 7 privilege in conjunction with DSA Art. 4 to 8 does not apply when the model provider actively generates answers. Bc against the model provider is robust; Bc against the client as a mirror-image notification duty is doctrinally constructible upon qualified awareness but remains without precedent.

In Strand B, the price-factor model does not reach through immediately. Compliance-GEO measures act there as structural optimisation in the first class of the influence spectrum (see Class-3 dividing line from Chapter 1.3) — they raise LLM citation probability without establishing an immediate attribution chain to the client's advertising responsibility. The definitive attribution classification for the Bb line is the subject of ongoing legal validation.

A.6Operational consequence and bridge to the Codex

The five frameworks (TKG §§ 54–57, UWG § 5a (4), MStV § 22 (1), DDG § 6 (1) (1), NIS-2 implementation act) bind a DE telco advertorial simultaneously, each with its own scope, its own addressee and its own legal consequence. TKG binds the advertiser with positive mandatory information; UWG binds the advertiser with the disclosure duty; MStV addresses the publisher; DDG addresses the platform operator; NIS-2 frames the advertising infrastructure statement from the compliance side.

Operationally, this means for mandate work: the DE appendix is the legal-anchor layer of the Codex; the universal Codex chapters remain sector- and jurisdiction-agnostic. The legal-mapping table in Chapter 4 refers to A.2 for the full BGH line; the price-factor model in Chapter 3 refers to A.3 for the Telco modifier and to A.4 and A.5 for the Strand-A/B differentiation. Other jurisdictions replace this appendix with local legal mappings without modifying the Codex chapters themselves.