If you are not cited there, you are not in the market. Classical Google search is losing volume; generative answer engines are taking over the research and decision step. More than 1 billion people worldwide regularly research in ChatGPT, Gemini, Perplexity, Claude or Google AI Overviews.
In regulated consumer markets — in Telco, Finance, Insurance and Commerce — the citability of these answers increasingly decides the buying decision of the end customer before the brand website is even opened.
Citation = a named source reference inside an AI answer. Retrieval engine = an AI system that pulls sources when answering a question (ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, Google AI Mode). Aggregator = a comparison or test platform that bundles third-party offers.
Three findings work together. First: classical search traffic is projected to fall by half by 2028. Second: the research stage moves into the answer stream of the AI engines. Third: in the German telco sector roughly one third of the citation volume sits on comparison portals without advertorial inventory, and three engines together carry roughly 5/6 of the volume. Anyone steering without this distribution invests in the wrong places.
The four columns of the board answer four different mandate questions. Drift answers whether now is the right moment for a citation mandate. Citation shows which share of the volume is reachable for classical advertising procurement and which is not. Engine answers where the mandate budget is allocated. Outcome delivers the economic statement for CFO and Procurement. The five sections that follow each deepen one column.
The board shown here was empirically gathered in the German telco market, because the citation mechanics around tariff aggregators are particularly clean to measure there. The mechanics themselves are, however, cross-sectoral: In the finance sector, citation volume shifts analogously to BaFin-compliant advisory sources and comparison platforms. In the insurance sector, to providers under VVG advisory duty and insurance brokers with IDD transparency. In the commerce sector, to UWG-compliant D2C platforms and influencer-disclosure carriers. The columns and procurement classes hold in all four sectors. The concrete volume shares are calibrated sector-specifically at the start of the mandate.
The drift discussion becomes a steering question, not an alarmist headline. The board is the anchor for every subsequent mandate decision, regardless of the sector.
→ Schritt 01 nach 02 · Wer die Drift erkannt hat, muss anders einkaufen
Classical media bookings buy reach. Answer visibility buys citation probability. That probability collapses when a placement violates one of eight binary eligibility criteria: UWG disclosure, main-representation duty, completeness duty, sector specifics. A single mandatory criterion fails — and the placement becomes citation-worthless, regardless of the publisher's reach.
~1/3 comparison portals without advertorial inventory (NB survey)
BaFin-supervised comparison platforms and independent finance portals — no classical sponsored content
VVG-compliant advisory providers and broker portals under IDD transparency duty
UWG-compliant test magazines and independent D2C review platforms
| Klasse | Was wird gekauft | Factor on list price |
|---|---|---|
| Citation Buy | Direct quote in the answer corpus with source reference | 1,0 × |
| Mixed-Buy | Brand mention plus half-citation, on equal footing in the answer text | 0,5 bis 0,9 × |
| Mention-Buy | Mere brand mention without a citation anchor | 0,2 bis 0,5 × |
| briefing-Fail | mandatory criterion gerissen, Placement citation-worthless | 0.0× |
In regulated sectors, a second multiplier stage applies. In the telco sector, a breach of TKG mandatory-information visibility disqualifies to 0.0×; in the finance sector, a breach of the BaFin risk-disclosure duty; in the insurance sector, a breach of the VVG advisory duty; in the commerce sector, a breach of the UWG disclosure duty or of the BGH Influencerin-II line. In all four cases regardless of the A- and B-profile. Worked out in the Compliance-GEO Codex, Chapter 3, testable in the Price-Factor Calculator.
When roughly 1/3 of the citation volume falls on non-buyable comparison portals and about 4 of 10 remaining bookings end up in the loss zone, classical reach-based procurement in regulated sectors consistently misses the citation target. This is rarely measured today, because no binary class logic exists.
Ahead of every media plan, the proposed placement runs through an eight-point eligibility check. The factor decides at what weight the placement enters the media-plan valuation. Citation Buy counts in full (1.0×), Mixed Buy weighted (0.5–0.9×), Mention Buy only partially (0.2–0.5×), briefing Fail is excluded entirely. Per sector, a second disqualification stage applies: TKG mandatory information (Telco), BaFin risk disclosure (Finance), VVG advisory duty (Insurance) or UWG disclosure (Commerce).
Reach and citation impact are made separately visible inside the media plan, instead of being lost in the quarterly reporting.
→ Schritt 02 nach 03 · Wer eingekauft hat, muss messen
The biggest measurement error sits in the frequency. Profound documents citation drift of up to 60% per month between engines. A monthly measurement cadence consistently misses the movements that determine the mandate outcome. The three-dimensional measurement logic is captured weekly, separately for the six central retrieval engines.
Share of cited URLs per prompt set, measured per engine. A citation-rate value without engine context is not methodologically meaningful.
How long a citation persists across measurement waves. Drift quantification as a half-life measure in mandate reporting.
Position in the answer window, context, trust score, aggregator-vs-original differentiation. Listicle share per engine.
Three engines each carry more than 20% citation share, and citations lose roughly half of their effect over twelve months. A monthly single-engine measurement misses the movement at three points simultaneously: it misses when an engine suddenly ranks differently, when citations disappear faster than assumed, and when non-English sessions frequently tip into English sub-searches. mandatee steering runs on a picture from the past.
The measurement programme is moved to a weekly cadence, all six engines separately, with three-dimensional analysis of citation rate, persistence and quality. Re-measurement points are set well before the half-life point. Follow-up waves are staggered by sector complexity, not distributed uniformly.
Measurement frequency mathematically justified in the audit, instead of having to defend it on gut feeling.
→ Schritt 03 nach 04 · Wer misst, sieht ein Outcome
In the Study Pyramid model, applied to a mid-sized German telco operator, the channel mix shifts substantially over eighteen months. Classical Google Search SEO has lost five points, classical Paid has lost two, and a new channel — LLM Direct Citation — adds twenty index points. The ChatGPT conversion in the model is 7% versus Google at 5% (Similarweb GenAI Landscape 2025, 1.1 bn visits), which arithmetically corresponds to a +40% advantage per citation. The same model logic is calibrated sector-specifically to Finance, Insurance or Commerce in the mandate.
Three sector examples of how citation substance plays out inside the answer text. Stylised answers, not real client output.
Model table on the German Telco example · Pyramid-model logic applies analogously to Finance, Insurance, Commerce with industry-specific calibration of conversion values and channel-mix shares
| Kanal | Heute | In 18 months | Delta |
|---|---|---|---|
| Organic Search (SEO) | 30 | 25 | −5 |
| Paid (SEA, Display) | 20 | 18 | −2 |
| Word-of-Mouth, Referral | 10 | 10 | 0 |
| Retail, Direct Sales | 5 | 5 | 0 |
| LLM Direct Citation (neuer Kanal) | neu | 20 | +20 |
| Summe sales index | 100 | 108 | +8 |
Model assumptions: Sistrix Prompt Research DACH 2025 (62 m questions), Scrunch half-life study, Similarweb GenAI Landscape 2025 (1.1 bn visits, ChatGPT 7% vs. Google 5% conversion), Indig 2026 retrieval mechanics. Channel-mix note: The annual reports of the five major German telco operators (Deutsche Telekom, Vodafone, Telefónica/O2, 1&1, Freenet) do not publish a sales-channel breakdown at the granularity SEO, SEA, WoM, Retail, Direct. The distribution shown here is a plausibility-checked industry assumption from public secondary sources (HDE, BDD, IBI 2024) and internal NB estimates, normalised to a sales index of 100. Sector transfer: The Pyramid-model logic (classical search and paid channels lose share, a new citation channel is added, net plus from citation substance without an increase in media budget) applies analogously to Finance, Insurance, Commerce. The concrete conversion values and channel shares are calibrated sector-specifically at mandate start. A model, not a mandate promise. Full derivation in the Study Pyramid.
Two measurement points carry the reporting. After 12 weeks, the citation rate in the leading engine has risen by roughly 85%. That is visible in the weekly steering. After 18 months, the sales index moves by plus eight index points, from citation substance in the three leading engines. Both emerge without an increase in media budget. The classical search and paid volume partially shifts into the new citation channel.
Reporting runs on two clearly separated levels. On the operational weekly level, citation rate is reported per engine. Steering of ongoing investment is based on this movement. On the quarterly board level, the sales index carries the before-after comparison, with sector calibration. The engine geometry from Section 03 steers the re-investment decision.
In the CFO conversation, a demonstrable efficiency gain is reported, rather than an outcome claim defended.
Model assumptions: Three-class factor 0.7 as the mean value. Citation-channel share 20% of total new customers in steady state, derived from the Pyramid model. Linear build-up of citation impact over 18 months from 0 to full. SEA equivalent: at the same monthly spend, how many new customers SEA delivers. Break-even at the intersection of the cumulative new-customer curves. Type in your own values — the calculator updates live. A model, not a mandate promise.
→ Step 04 to 05 · The outcome only carries if the methodology is sector-sharp
Beyond the ongoing measurement routine, the academic oversight pursues three research lines that are being prepared for peer-reviewed publication. This makes mandate statements audit-defensible. The certainty is documented, not merely claimed.
Half-life metrics (half-life measures from the natural sciences, transferred to citation impact) for citation persistence. The measured half-life yields the methodologically justified re-measurement frequency.
Mixed-effects modelling (a statistical method that separates fixed and random effects) of citation variance across engines, sectors and clusters. Yields sampling recommendations for follow-up waves.
Cohen's Kappa and Krippendorff's Alpha (statistical measures of agreement between independent coders) for the aggregator-versus-original classification. Closes the validation gap against industry tools.
The measurement and class model behind Compliance-GEO is filed for patent and utility-model protection in Germany. The subject matter of protection is a computer system for valuing digital publication placements inside generative AI systems — namely the deterministic aggregation logic across multiple independent inference sources, the three-class scoring, the override mechanic for sector-specific mandatory information, and the revision-safe hash-chain logging.
→ Science trifft sector-Recht · vier regulierte Verticals
Compliance-GEO is built sector-sharp. The discipline operates in four regulated consumer verticals of the EU, where advertising disclosure, main-representation duties and sector-specific regulation interact. Measurement logic and Procurement Standard are applicable in every vertical; the legal context determines the focus areas.
Aggregator dominance in tariff queries through comparison portals. High regulatory density at the intersection of TKG, TTDSG, BNetzA and NIS-2. Tier-1 competition for top-3 positions in the answer engines.
Credit and investment recommendations require risk disclosures directly in the answer corpus. BaFin supervisory practice as an additional layer on top of UWG disclosure in digital channels.
Advisory-duty architecture under VVG. Product-specific information sheets and IDD specifications on remuneration transparency apply directly inside the answer path of the engines.
Influencer cooperation under the BGH Influencerin-II line. Subscription-contract transparency and UWG disclosure duties on D2C platforms as operative anchors.
A university cooperation accompanies the measurement architecture across seven focus areas. For each focus area, what is examined, what follows from it and the value behind it is made visible.
| What we do | What we find with it | Value for you |
|---|---|---|
| Power analysis | How large the sample must be so that engine and sector effects do not disappear into statistical uncertainty. | Measurement reports stand up to audit questions from BaFin or BNetzA contexts. |
| Cohen's Kappa | How strong the agreement is between independent coders on the classification of aggregator vs. original vs. hybrid. | Citation classification is quantifiably reliable, not merely claimed. |
| Half-life measure | After how many months a typical citation has lost 50% of its initial strength, differentiated by engine and sector. | A methodologically justified re-measurement frequency — not a gut feeling. |
| Mixed-effects modelling | Which share of the citation variance is explained by engine, sector and cluster level respectively, with a robustness variant. | Stratification architecture for follow-up waves data-founded. |
| Pre-registration | Hypotheses, measurement designs and analysis procedures are frozen and documented ahead of every wave. | Protection against retrospective hypothesis adjustment and p-hacking allegations. |
| Anonymised data with DOI | Research data are published on Zenodo with a DOI (Digital Object Identifier), client identifier anonymised. | Study reproducibility for external audits and for follow-up research. |
| Multi-engine comparison | Cross-engine citation patterns between ChatGPT, Gemini, Perplexity and Google AI Overviews. | An engine-mix recommendation instead of single-engine optimisation, with data-driven spend allocation. |
In an escalation case, a supervisory authority asks three concrete questions. First question: Which power analysis underlies the measurement logic — how large is the sample and how reliable the measured effect? Second question: What is Cohen's Kappa for the citation classification — how reliable is the aggregator-original-hybrid distinction? Third question: How is the sector distribution documented? industry tools answer none of these questions. A client without answers carries the audit risk alone.
Die Methodology wird vor jedem mandate start sektor-scharf angepasst und in einer Compliance-Deklaration documented. Drei Mechanismen tragen die Audit-Belastbarkeit: Pre-Registration der Mess-hypotheses ahead of every Welle, anonymisede Daten mit DOI auf Zenodo zur externen Reproduzierbarkeit, Co-Autorschaft peer-reviewter Publikationen. Im escalation case liegt die Antwort vor, bevor die Frage gestellt wird.
Audit defence with documented methodology, instead of with an advisory report and gut feeling.
→ Vor dem Initial conversation · vier Tools zur self-assessment
A great deal can be checked yourself before the initial conversation. The four Northbridge tools are usable directly in the browser, free of charge, without registration. Each tool starts from a realistic scenario; each output points to the Codex passage that stands behind it.
Eight binary A-criteria checked against a planned placement, with three starting scenarios. The tool shows which link the chain breaks at and whether the cause sits with the publisher (pre-check) or the client (briefing).
Drei-Klassen-Einstufung Citation Buy, Mixed-Buy, Mention-Buy mit Modell-Blended-Faktor. Schiebt sich der B-Regler, springt der Faktor, der Balken wandert, die Einstufung kippt.
Vier sector-Beispiele aus den Northbridge-Sector-Dossiers. Pro Vertical werden drei sektorale Kern-Risiken benannt, mit Verlinkung in das passende Codex-Chapter.
Decision matrix with six disclosure variants from full advertising disclosure through hybrid forms to legally inadmissible covert advertising. Three-axis logic under UWG, MStV, DDG plus TKG mandatory information in telco mode.
Four interactive tools cover the core mechanics of the discipline: Eligibility Check tests eight mandatory criteria against a specific placement. Price-Factor Calculator computes class weighting under three-class logic. Sector Explorer shows sector-specific legal anchors for Telco, Finance, Insurance, Commerce. Disclosure Matrix orders advertising disclosure from fully compliant to covert advertising. Today, comparable checks often run only after the mandate starts, by which point the room for action is smaller.
The tools are free of charge and usable without registration. Before the initial conversation, two or three of your own placements are run through. Typical gaps become visible before anyone replies. Inside the mandate, the tools are integrated into the steering workflow: Eligibility Check before every booking, Price-Factor Calculator inside the media plan, Sector Explorer for legal positioning, Disclosure Matrix as a briefing template for publishers.
Initial conversation with concrete gap questions, instead of general themes. Preparation effort on your side, faster mandate clarification on ours.
Northbridge runs an orchestrated stack of fourteen tools, grouped in three operative categories. Deliberate overlap at single points, where cross-validation carries methodologically. No duplicate licences out of convenience.
Peec AI · Rankscale · Sistrix Plus · Scrunch AI · Profound
Weekly citation capture across the six retrieval engines with a DE deep-pipeline and an EU four-country snapshot, sub-search visibility, AIO layer (AI Overviews) for DE search, accuracy validation of the cited content.
Screaming Frog · Ahrefs · Surfer SEO · Google Search Console + IndexNow · Claude Pro · ChatGPT Plus · Perplexity Pro
Site crawl with eligibility check, backlinks and international visibility, engine-pro accounts for cross-engine validation of the own measurement programme.
Looker Studio · Brandwatch oder Talkwalker
Dashboard layer for mandate steering, social listening for citation echo in non-LLM channels, preparation of the quarterly board reports.
Full comparison matrix with all sixteen dimensions per tool: view the Tool-Stack Supermatrix.
Forty minutes, no obligation. You bring the concrete mandate question, we bring the eighteen-criteria audit. Three steps:
We listen to your mandate question and place it inside the four-vertical logic. Sector specifics and regulatory anchors are named in the first ten minutes.
We test against the eighteen-criteria standard which eligibility anchors are already in place and where the mandatory criteria show gaps.
You take a methodology proposal with you, even if no engagement follows. Compliance-GEO as a discipline carries independently of the mandate.
