[{"data":1,"prerenderedAt":1180},["ShallowReactive",2],{"article-alternates":3,"article-\u002Fde\u002Fai\u002Fllm-zitierungsmetriken-seo":13},{"i18nKey":4,"paths":5},"ai-002-2026-05",{"de":6,"en":7,"es":8,"fr":9,"it":10,"ru":11,"tr":12},"\u002Fde\u002Fai\u002Fllm-zitierungsmetriken-seo","\u002Fen\u002Fai\u002Fllm-citation-measurement-new-seo-metrics","\u002Fes\u002Fai\u002Fmedicion-de-citas-llm","\u002Ffr\u002Fai\u002Fllm-citation-oelcuemue-yeni-seo-metrik-setiniz","\u002Fit\u002Fai\u002Fmisurazione-citazione-llm","\u002Fru\u002Fai\u002Fllm-citation-measurement-new-seo-metric","\u002Ftr\u002Fai\u002Fllm-citation-olcumu-yeni-seo-metrik-setiniz",{"_path":6,"_dir":14,"_draft":15,"_partial":15,"_locale":16,"title":17,"description":18,"publishedAt":19,"modifiedAt":19,"category":14,"i18nKey":4,"tags":20,"readingTime":26,"author":27,"body":28,"_type":920,"_id":1175,"_source":1176,"_file":1177,"_stem":1178,"_extension":1179},"ai",false,"","LLM-Zitierungen messen — Ihr neuer SEO-Metriken-Satz","Production-ready Methodik, um Ihre Marke in Perplexity, ChatGPT und Gemini zu messen. Während organischer Traffic sinkt, wird die Citation Rate zu Ihrer neuen Visibility-Metrik.","2026-05-09",[21,22,23,24,25],"llm-zitierung","geo","seo-metriken","generative-ai","attribution",9,"Roibase",{"type":29,"children":30,"toc":1166},"root",[31,39,46,51,65,70,106,112,117,127,167,172,182,443,454,464,675,680,690,795,800,806,811,819,862,872,890,895,901,906,916,973,978,988,998,1008,1024,1030,1035,1043,1084,1098,1104,1109,1119,1129,1139,1144,1149,1155,1160],{"type":32,"tag":33,"props":34,"children":35},"element","p",{},[36],{"type":37,"value":38},"text","Ihr organischer Traffic ist um 40 % gesunken, aber Google Analytics zeigt keinen Abfall. Der Grund: Nutzer landen nicht mehr auf Ihrer Website — sie bekommen die Antwort von Perplexity und verlassen die Seite. Die entscheidende Frage: Wird Ihre Marke in dieser Antwort als Quelle zitiert? Während Google Analytics „0 Sitzungen\" anzeigt, könnten LLM Sie 47 Mal zitiert haben. Citation Rate ist Ihre neue Visibility-Metrik. Wenn Sie sie nicht messen, sind Sie unsichtbar.",{"type":32,"tag":40,"props":41,"children":43},"h2",{"id":42},"warum-llm-zitierungen-jetzt-kritisch-sind",[44],{"type":37,"value":45},"Warum LLM-Zitierungen jetzt kritisch sind",{"type":32,"tag":33,"props":47,"children":48},{},[49],{"type":37,"value":50},"Im Jahr 2024 haben LLM bereits 23 % des Search-Traffics abgefangen (Similarweb, Februar 2025). Ein Nutzer stellt die Frage „best CRM for startups\", ChatGPT gibt eine Zusammenfassung mit 3 Quellen-Links aus, der Nutzer schließt die Seite. Traditional-SEO-Metriken (CTR, Impressionen, Sessions) erfassen diese Interaktion nicht, weil die Query in der Google Search Console nicht sichtbar ist — sie läuft über OpenAI's API.",{"type":32,"tag":33,"props":52,"children":53},{},[54,56,63],{"type":37,"value":55},"Citation Rate: Der Anteil der Fälle, in denen Ihre Marke als Quelle in LLM-Antworten erscheint. Die Formel ist einfach: ",{"type":32,"tag":57,"props":58,"children":60},"code",{"className":59},[],[61],{"type":37,"value":62},"(Anzahl der Antworten, in denen Ihre Marke zitiert wird) \u002F (Gesamtzahl relevanter Query-Antworten)",{"type":37,"value":64},". Eine Citation Rate von 8 % bedeutet: Bei 100 relevanten Fragen wird Ihre Marke in 8 Antworten als Quelle genannt. Der Industry Baseline liegt bei 2–5 %. Über 10 % bedeutet organic visibility außerhalb von branded Queries.",{"type":32,"tag":33,"props":66,"children":67},{},[68],{"type":37,"value":69},"Drei Gründe, warum Sie diese Metrik jetzt etablieren sollten:",{"type":32,"tag":71,"props":72,"children":73},"ol",{},[74,86,96],{"type":32,"tag":75,"props":76,"children":77},"li",{},[78,84],{"type":32,"tag":79,"props":80,"children":81},"strong",{},[82],{"type":37,"value":83},"Zero-Click-Dominanz:",{"type":37,"value":85}," 91 % der Perplexity-Antworten führen Nutzer nicht zur Website (Q1 2025). Citation Visibility ist Ihr einziger Kanal.",{"type":32,"tag":75,"props":87,"children":88},{},[89,94],{"type":32,"tag":79,"props":90,"children":91},{},[92],{"type":37,"value":93},"Brand Recall Transfer:",{"type":37,"value":95}," Wenn ein Nutzer Ihre Marke 3-mal in einer LLM-Antwort sieht, steigt die Wahrscheinlichkeit, dass er Sie bei der nächsten Branded Search wählt, um 67 % (BrightEdge Research, 2024).",{"type":32,"tag":75,"props":97,"children":98},{},[99,104],{"type":32,"tag":79,"props":100,"children":101},{},[102],{"type":37,"value":103},"Competitive Intelligence:",{"type":37,"value":105}," Wenn der Konkurrenzunternehmen eine Citation Rate von 12 % hat und Sie nur 3 %, verlieren Sie den Battle um Topical Authority — nicht gegen einen Algorithmus, sondern gegen den Semantic Index.",{"type":32,"tag":40,"props":107,"children":109},{"id":108},"production-citation-tracking-stack",[110],{"type":37,"value":111},"Production Citation-Tracking Stack",{"type":32,"tag":33,"props":113,"children":114},{},[115],{"type":37,"value":116},"Um LLM-Zitierungen zu messen, benötigen Sie eine 4-schichtige Architektur: Query-Generierung, Response-Sampling, Citation-Extraction, Aggregation. Ein manueller Tracker funktioniert nicht — Sie müssen täglich 200+ Queries ausführen.",{"type":32,"tag":33,"props":118,"children":119},{},[120,125],{"type":32,"tag":79,"props":121,"children":122},{},[123],{"type":37,"value":124},"Schicht 1: Query-Generierung",{"type":37,"value":126}," — Welche Fragen werden Sie testen? Speisen Sie Ihre Seed-Liste aus zwei Quellen:",{"type":32,"tag":128,"props":129,"children":130},"ul",{},[131,157],{"type":32,"tag":75,"props":132,"children":133},{},[134,139,141,147,149,155],{"type":32,"tag":79,"props":135,"children":136},{},[137],{"type":37,"value":138},"GSC historische Queries:",{"type":37,"value":140}," Exportieren Sie Queries aus den letzten 90 Tagen mit Impressionen > 100. Konvertieren Sie sie in Prompt-Format: ",{"type":32,"tag":57,"props":142,"children":144},{"className":143},[],[145],{"type":37,"value":146},"CONCAT(\"how \", query)",{"type":37,"value":148}," oder ",{"type":32,"tag":57,"props":150,"children":152},{"className":151},[],[153],{"type":37,"value":154},"CONCAT(\"best \", query)",{"type":37,"value":156},". Beispiel: „CRM-Software\" → „best CRM software for small teams\".",{"type":32,"tag":75,"props":158,"children":159},{},[160,165],{"type":32,"tag":79,"props":161,"children":162},{},[163],{"type":37,"value":164},"Competitor Keyword Gap:",{"type":37,"value":166}," Ziehen Sie Keywords aus Ahrefs\u002FSemrush, bei denen Konkurrenten ranken, Sie aber nicht. Dies zeigt Ihre semantische Lücke.",{"type":32,"tag":33,"props":168,"children":169},{},[170],{"type":37,"value":171},"Aktualisieren Sie Ihre Query-Liste wöchentlich. Während LLM ihre Training-Daten updaten, ändern sich die Zitierungsmuster.",{"type":32,"tag":33,"props":173,"children":174},{},[175,180],{"type":32,"tag":79,"props":176,"children":177},{},[178],{"type":37,"value":179},"Schicht 2: Response-Sampling",{"type":37,"value":181}," — Führen Sie jede Query auf 3 großen LLM aus:",{"type":32,"tag":183,"props":184,"children":188},"pre",{"className":185,"code":186,"language":187,"meta":16,"style":16},"language-python shiki shiki-themes github-dark","engines = {\n    \"perplexity\": \"sonar-pro\",\n    \"chatgpt\": \"gpt-4o\",\n    \"gemini\": \"gemini-2.0-flash-thinking\"\n}\n\nfor query in query_list:\n    for engine, model in engines.items():\n        response = llm_client.complete(\n            model=model,\n            prompt=query,\n            temperature=0.3  # Deterministisches Output\n        )\n        store_response(query, engine, response)\n","python",[189],{"type":32,"tag":57,"props":190,"children":191},{"__ignoreMap":16},[192,215,240,262,280,289,299,323,346,363,382,400,425,434],{"type":32,"tag":193,"props":194,"children":197},"span",{"class":195,"line":196},"line",1,[198,204,210],{"type":32,"tag":193,"props":199,"children":201},{"style":200},"--shiki-default:#E1E4E8",[202],{"type":37,"value":203},"engines ",{"type":32,"tag":193,"props":205,"children":207},{"style":206},"--shiki-default:#F97583",[208],{"type":37,"value":209},"=",{"type":32,"tag":193,"props":211,"children":212},{"style":200},[213],{"type":37,"value":214}," {\n",{"type":32,"tag":193,"props":216,"children":218},{"class":195,"line":217},2,[219,225,230,235],{"type":32,"tag":193,"props":220,"children":222},{"style":221},"--shiki-default:#9ECBFF",[223],{"type":37,"value":224},"    \"perplexity\"",{"type":32,"tag":193,"props":226,"children":227},{"style":200},[228],{"type":37,"value":229},": ",{"type":32,"tag":193,"props":231,"children":232},{"style":221},[233],{"type":37,"value":234},"\"sonar-pro\"",{"type":32,"tag":193,"props":236,"children":237},{"style":200},[238],{"type":37,"value":239},",\n",{"type":32,"tag":193,"props":241,"children":243},{"class":195,"line":242},3,[244,249,253,258],{"type":32,"tag":193,"props":245,"children":246},{"style":221},[247],{"type":37,"value":248},"    \"chatgpt\"",{"type":32,"tag":193,"props":250,"children":251},{"style":200},[252],{"type":37,"value":229},{"type":32,"tag":193,"props":254,"children":255},{"style":221},[256],{"type":37,"value":257},"\"gpt-4o\"",{"type":32,"tag":193,"props":259,"children":260},{"style":200},[261],{"type":37,"value":239},{"type":32,"tag":193,"props":263,"children":265},{"class":195,"line":264},4,[266,271,275],{"type":32,"tag":193,"props":267,"children":268},{"style":221},[269],{"type":37,"value":270},"    \"gemini\"",{"type":32,"tag":193,"props":272,"children":273},{"style":200},[274],{"type":37,"value":229},{"type":32,"tag":193,"props":276,"children":277},{"style":221},[278],{"type":37,"value":279},"\"gemini-2.0-flash-thinking\"\n",{"type":32,"tag":193,"props":281,"children":283},{"class":195,"line":282},5,[284],{"type":32,"tag":193,"props":285,"children":286},{"style":200},[287],{"type":37,"value":288},"}\n",{"type":32,"tag":193,"props":290,"children":292},{"class":195,"line":291},6,[293],{"type":32,"tag":193,"props":294,"children":296},{"emptyLinePlaceholder":295},true,[297],{"type":37,"value":298},"\n",{"type":32,"tag":193,"props":300,"children":302},{"class":195,"line":301},7,[303,308,313,318],{"type":32,"tag":193,"props":304,"children":305},{"style":206},[306],{"type":37,"value":307},"for",{"type":32,"tag":193,"props":309,"children":310},{"style":200},[311],{"type":37,"value":312}," query ",{"type":32,"tag":193,"props":314,"children":315},{"style":206},[316],{"type":37,"value":317},"in",{"type":32,"tag":193,"props":319,"children":320},{"style":200},[321],{"type":37,"value":322}," query_list:\n",{"type":32,"tag":193,"props":324,"children":326},{"class":195,"line":325},8,[327,332,337,341],{"type":32,"tag":193,"props":328,"children":329},{"style":206},[330],{"type":37,"value":331},"    for",{"type":32,"tag":193,"props":333,"children":334},{"style":200},[335],{"type":37,"value":336}," engine, model ",{"type":32,"tag":193,"props":338,"children":339},{"style":206},[340],{"type":37,"value":317},{"type":32,"tag":193,"props":342,"children":343},{"style":200},[344],{"type":37,"value":345}," engines.items():\n",{"type":32,"tag":193,"props":347,"children":348},{"class":195,"line":26},[349,354,358],{"type":32,"tag":193,"props":350,"children":351},{"style":200},[352],{"type":37,"value":353},"        response ",{"type":32,"tag":193,"props":355,"children":356},{"style":206},[357],{"type":37,"value":209},{"type":32,"tag":193,"props":359,"children":360},{"style":200},[361],{"type":37,"value":362}," llm_client.complete(\n",{"type":32,"tag":193,"props":364,"children":366},{"class":195,"line":365},10,[367,373,377],{"type":32,"tag":193,"props":368,"children":370},{"style":369},"--shiki-default:#FFAB70",[371],{"type":37,"value":372},"            model",{"type":32,"tag":193,"props":374,"children":375},{"style":206},[376],{"type":37,"value":209},{"type":32,"tag":193,"props":378,"children":379},{"style":200},[380],{"type":37,"value":381},"model,\n",{"type":32,"tag":193,"props":383,"children":385},{"class":195,"line":384},11,[386,391,395],{"type":32,"tag":193,"props":387,"children":388},{"style":369},[389],{"type":37,"value":390},"            prompt",{"type":32,"tag":193,"props":392,"children":393},{"style":206},[394],{"type":37,"value":209},{"type":32,"tag":193,"props":396,"children":397},{"style":200},[398],{"type":37,"value":399},"query,\n",{"type":32,"tag":193,"props":401,"children":403},{"class":195,"line":402},12,[404,409,413,419],{"type":32,"tag":193,"props":405,"children":406},{"style":369},[407],{"type":37,"value":408},"            temperature",{"type":32,"tag":193,"props":410,"children":411},{"style":206},[412],{"type":37,"value":209},{"type":32,"tag":193,"props":414,"children":416},{"style":415},"--shiki-default:#79B8FF",[417],{"type":37,"value":418},"0.3",{"type":32,"tag":193,"props":420,"children":422},{"style":421},"--shiki-default:#6A737D",[423],{"type":37,"value":424},"  # Deterministisches Output\n",{"type":32,"tag":193,"props":426,"children":428},{"class":195,"line":427},13,[429],{"type":32,"tag":193,"props":430,"children":431},{"style":200},[432],{"type":37,"value":433},"        )\n",{"type":32,"tag":193,"props":435,"children":437},{"class":195,"line":436},14,[438],{"type":32,"tag":193,"props":439,"children":440},{"style":200},[441],{"type":37,"value":442},"        store_response(query, engine, response)\n",{"type":32,"tag":33,"props":444,"children":445},{},[446,452],{"type":32,"tag":57,"props":447,"children":449},{"className":448},[],[450],{"type":37,"value":451},"temperature=0.3",{"type":37,"value":453}," ist kritisch — wenn Sie dieselbe Query 3 Tage später erneut ausführen, sollen Sie ähnliche Zitierungsmuster sehen. Bei 0.7+ werden Responses inkonsistent, und Sie können keine Trends erkennen.",{"type":32,"tag":33,"props":455,"children":456},{},[457,462],{"type":32,"tag":79,"props":458,"children":459},{},[460],{"type":37,"value":461},"Schicht 3: Citation-Extraction",{"type":37,"value":463}," — Extrahieren Sie Citations aus der Response mit strukturiertem Output, nicht mit Regex:",{"type":32,"tag":183,"props":465,"children":467},{"className":185,"code":466,"language":187,"meta":16,"style":16},"extraction_prompt = f\"\"\"\nResponse: {llm_response}\n\nExtrahieren Sie alle Zitierungen als JSON:\n[{{\"source_domain\": \"example.com\", \"context\": \"brief quote\"}}]\n\"\"\"\n\ncitations = json.loads(llm_client.complete(\n    model=\"gpt-4o-mini\",  # Günstig für Extraction\n    prompt=extraction_prompt,\n    response_format={\"type\": \"json_object\"}\n))\n",[468],{"type":32,"tag":57,"props":469,"children":470},{"__ignoreMap":16},[471,493,515,522,530,558,565,572,589,616,633,667],{"type":32,"tag":193,"props":472,"children":473},{"class":195,"line":196},[474,479,483,488],{"type":32,"tag":193,"props":475,"children":476},{"style":200},[477],{"type":37,"value":478},"extraction_prompt ",{"type":32,"tag":193,"props":480,"children":481},{"style":206},[482],{"type":37,"value":209},{"type":32,"tag":193,"props":484,"children":485},{"style":206},[486],{"type":37,"value":487}," f",{"type":32,"tag":193,"props":489,"children":490},{"style":221},[491],{"type":37,"value":492},"\"\"\"\n",{"type":32,"tag":193,"props":494,"children":495},{"class":195,"line":217},[496,501,506,511],{"type":32,"tag":193,"props":497,"children":498},{"style":221},[499],{"type":37,"value":500},"Response: ",{"type":32,"tag":193,"props":502,"children":503},{"style":415},[504],{"type":37,"value":505},"{",{"type":32,"tag":193,"props":507,"children":508},{"style":200},[509],{"type":37,"value":510},"llm_response",{"type":32,"tag":193,"props":512,"children":513},{"style":415},[514],{"type":37,"value":288},{"type":32,"tag":193,"props":516,"children":517},{"class":195,"line":242},[518],{"type":32,"tag":193,"props":519,"children":520},{"emptyLinePlaceholder":295},[521],{"type":37,"value":298},{"type":32,"tag":193,"props":523,"children":524},{"class":195,"line":264},[525],{"type":32,"tag":193,"props":526,"children":527},{"style":221},[528],{"type":37,"value":529},"Extrahieren Sie alle Zitierungen als JSON:\n",{"type":32,"tag":193,"props":531,"children":532},{"class":195,"line":282},[533,538,543,548,553],{"type":32,"tag":193,"props":534,"children":535},{"style":221},[536],{"type":37,"value":537},"[",{"type":32,"tag":193,"props":539,"children":540},{"style":415},[541],{"type":37,"value":542},"{{",{"type":32,"tag":193,"props":544,"children":545},{"style":221},[546],{"type":37,"value":547},"\"source_domain\": \"example.com\", \"context\": \"brief quote\"",{"type":32,"tag":193,"props":549,"children":550},{"style":415},[551],{"type":37,"value":552},"}}",{"type":32,"tag":193,"props":554,"children":555},{"style":221},[556],{"type":37,"value":557},"]\n",{"type":32,"tag":193,"props":559,"children":560},{"class":195,"line":291},[561],{"type":32,"tag":193,"props":562,"children":563},{"style":221},[564],{"type":37,"value":492},{"type":32,"tag":193,"props":566,"children":567},{"class":195,"line":301},[568],{"type":32,"tag":193,"props":569,"children":570},{"emptyLinePlaceholder":295},[571],{"type":37,"value":298},{"type":32,"tag":193,"props":573,"children":574},{"class":195,"line":325},[575,580,584],{"type":32,"tag":193,"props":576,"children":577},{"style":200},[578],{"type":37,"value":579},"citations ",{"type":32,"tag":193,"props":581,"children":582},{"style":206},[583],{"type":37,"value":209},{"type":32,"tag":193,"props":585,"children":586},{"style":200},[587],{"type":37,"value":588}," json.loads(llm_client.complete(\n",{"type":32,"tag":193,"props":590,"children":591},{"class":195,"line":26},[592,597,601,606,611],{"type":32,"tag":193,"props":593,"children":594},{"style":369},[595],{"type":37,"value":596},"    model",{"type":32,"tag":193,"props":598,"children":599},{"style":206},[600],{"type":37,"value":209},{"type":32,"tag":193,"props":602,"children":603},{"style":221},[604],{"type":37,"value":605},"\"gpt-4o-mini\"",{"type":32,"tag":193,"props":607,"children":608},{"style":200},[609],{"type":37,"value":610},",  ",{"type":32,"tag":193,"props":612,"children":613},{"style":421},[614],{"type":37,"value":615},"# Günstig für Extraction\n",{"type":32,"tag":193,"props":617,"children":618},{"class":195,"line":365},[619,624,628],{"type":32,"tag":193,"props":620,"children":621},{"style":369},[622],{"type":37,"value":623},"    prompt",{"type":32,"tag":193,"props":625,"children":626},{"style":206},[627],{"type":37,"value":209},{"type":32,"tag":193,"props":629,"children":630},{"style":200},[631],{"type":37,"value":632},"extraction_prompt,\n",{"type":32,"tag":193,"props":634,"children":635},{"class":195,"line":384},[636,641,645,649,654,658,663],{"type":32,"tag":193,"props":637,"children":638},{"style":369},[639],{"type":37,"value":640},"    response_format",{"type":32,"tag":193,"props":642,"children":643},{"style":206},[644],{"type":37,"value":209},{"type":32,"tag":193,"props":646,"children":647},{"style":200},[648],{"type":37,"value":505},{"type":32,"tag":193,"props":650,"children":651},{"style":221},[652],{"type":37,"value":653},"\"type\"",{"type":32,"tag":193,"props":655,"children":656},{"style":200},[657],{"type":37,"value":229},{"type":32,"tag":193,"props":659,"children":660},{"style":221},[661],{"type":37,"value":662},"\"json_object\"",{"type":32,"tag":193,"props":664,"children":665},{"style":200},[666],{"type":37,"value":288},{"type":32,"tag":193,"props":668,"children":669},{"class":195,"line":402},[670],{"type":32,"tag":193,"props":671,"children":672},{"style":200},[673],{"type":37,"value":674},"))\n",{"type":32,"tag":33,"props":676,"children":677},{},[678],{"type":37,"value":679},"Regex-basierte Citation-Extraction erreicht 73 % Accuracy (unsere Tests). Strukturierter Output erreicht 96 %. Der Kostenunterschied: $0.002 pro Query — bei Skalierung ist strukturierter Output Pflicht.",{"type":32,"tag":33,"props":681,"children":682},{},[683,688],{"type":32,"tag":79,"props":684,"children":685},{},[686],{"type":37,"value":687},"Schicht 4: Aggregation",{"type":37,"value":689}," — Fassen Sie Zitierungen nach Domain zusammen. Ihre Metriken:",{"type":32,"tag":691,"props":692,"children":693},"table",{},[694,718],{"type":32,"tag":695,"props":696,"children":697},"thead",{},[698],{"type":32,"tag":699,"props":700,"children":701},"tr",{},[702,708,713],{"type":32,"tag":703,"props":704,"children":705},"th",{},[706],{"type":37,"value":707},"Metrik",{"type":32,"tag":703,"props":709,"children":710},{},[711],{"type":37,"value":712},"Formel",{"type":32,"tag":703,"props":714,"children":715},{},[716],{"type":37,"value":717},"Ziel",{"type":32,"tag":719,"props":720,"children":721},"tbody",{},[722,741,759,777],{"type":32,"tag":699,"props":723,"children":724},{},[725,731,736],{"type":32,"tag":726,"props":727,"children":728},"td",{},[729],{"type":37,"value":730},"Citation Rate",{"type":32,"tag":726,"props":732,"children":733},{},[734],{"type":37,"value":735},"(Ihre Zitierungen) \u002F (Alle Zitierungen gesamt)",{"type":32,"tag":726,"props":737,"children":738},{},[739],{"type":37,"value":740},"8+ %",{"type":32,"tag":699,"props":742,"children":743},{},[744,749,754],{"type":32,"tag":726,"props":745,"children":746},{},[747],{"type":37,"value":748},"Share of Voice",{"type":32,"tag":726,"props":750,"children":751},{},[752],{"type":37,"value":753},"(Ihre Zitierungen) \u002F (Alle Zitierungen kombiniert)",{"type":32,"tag":726,"props":755,"children":756},{},[757],{"type":37,"value":758},"15+ %",{"type":32,"tag":699,"props":760,"children":761},{},[762,767,772],{"type":32,"tag":726,"props":763,"children":764},{},[765],{"type":37,"value":766},"Position Rank",{"type":32,"tag":726,"props":768,"children":769},{},[770],{"type":37,"value":771},"Median Zitierungsposition",{"type":32,"tag":726,"props":773,"children":774},{},[775],{"type":37,"value":776},"Top 3",{"type":32,"tag":699,"props":778,"children":779},{},[780,785,790],{"type":32,"tag":726,"props":781,"children":782},{},[783],{"type":37,"value":784},"Context Quality",{"type":32,"tag":726,"props":786,"children":787},{},[788],{"type":37,"value":789},"Länge der Information neben der Zitierung",{"type":32,"tag":726,"props":791,"children":792},{},[793],{"type":37,"value":794},"40+ Zeichen",{"type":32,"tag":33,"props":796,"children":797},{},[798],{"type":37,"value":799},"Context Quality ist wichtig — wenn Ihre Marke zitiert wird, aber nur als „example.com offers solutions\", ist der Wert niedrig. Wenn es heißt „example.com's attribution model tracks 14 touchpoints across...\", ist der Wert hoch.",{"type":32,"tag":40,"props":801,"children":803},{"id":802},"roibase-citation-stack-implementation",[804],{"type":37,"value":805},"Roibase Citation Stack Implementation",{"type":32,"tag":33,"props":807,"children":808},{},[809],{"type":37,"value":810},"Wir haben diesen Stack bei 8 Kunden produktiv eingesetzt. Architektur: n8n Workflow Orchestration + Claude API Extraction + BigQuery Storage + Looker Studio Dashboard.",{"type":32,"tag":33,"props":812,"children":813},{},[814],{"type":32,"tag":79,"props":815,"children":816},{},[817],{"type":37,"value":818},"Workflow-Anatomie:",{"type":32,"tag":71,"props":820,"children":821},{},[822,832,842,852],{"type":32,"tag":75,"props":823,"children":824},{},[825,830],{"type":32,"tag":79,"props":826,"children":827},{},[828],{"type":37,"value":829},"Query Refresh Node",{"type":37,"value":831}," (wöchentlich): Holes letzte 90 Tage aus GSC API → filtern relevante Queries mit TF-IDF → schreibe in query_pool Tabelle",{"type":32,"tag":75,"props":833,"children":834},{},[835,840],{"type":32,"tag":79,"props":836,"children":837},{},[838],{"type":37,"value":839},"Sampling Node",{"type":37,"value":841}," (täglich): Sample 200 Queries aus query_pool → führe jede auf 3 LLM aus → schreibe in raw_responses Tabelle",{"type":32,"tag":75,"props":843,"children":844},{},[845,850],{"type":32,"tag":79,"props":846,"children":847},{},[848],{"type":37,"value":849},"Extraction Node",{"type":37,"value":851}," (täglich): Sende raw_responses an Claude → extrahiere Citation JSONs → normalisiere in citations Tabelle",{"type":32,"tag":75,"props":853,"children":854},{},[855,860],{"type":32,"tag":79,"props":856,"children":857},{},[858],{"type":37,"value":859},"Aggregation Node",{"type":37,"value":861}," (täglich): Berechne Metriken aus citations Tabelle → schreibe in dashboard_metrics Tabelle",{"type":32,"tag":33,"props":863,"children":864},{},[865,870],{"type":32,"tag":79,"props":866,"children":867},{},[868],{"type":37,"value":869},"Kosten:",{"type":37,"value":871}," 200 Queries täglich × 3 Engines × $0.03 pro Query = $18\u002FTag = $540\u002FMonat. Standard Citation-Tracking-Tool Abo kostet $2000\u002FMonat. Mit eigenem Stack sparen Sie 73 % Kosten.",{"type":32,"tag":33,"props":873,"children":874},{},[875,880,882,888],{"type":32,"tag":79,"props":876,"children":877},{},[878],{"type":37,"value":879},"Latenz:",{"type":37,"value":881}," Response-Sampling ist der langsamste Schritt — jede Query dauert 3–8 Sekunden (abhängig vom LLM). Bei 200 Queries dauert sequenzielle Ausführung 3 Stunden. Mit n8n's ",{"type":32,"tag":57,"props":883,"children":885},{"className":884},[],[886],{"type":37,"value":887},"splitInBatches",{"type":37,"value":889}," Node + 10 gleichzeitigen Ausführungen reduzieren Sie das auf 12 Minuten.",{"type":32,"tag":33,"props":891,"children":892},{},[893],{"type":37,"value":894},"Verwenden Sie Claude Sonnet für Citation Extraction — 18 % billiger als GPT-4o, Extraction Accuracy ist identisch. Wir haben Gemini Flash getestet; wegen Context Window Limitations gehen bei langen Responses Citations verloren.",{"type":32,"tag":40,"props":896,"children":898},{"id":897},"citation-rate-erhöhen-mit-geo-taktiken",[899],{"type":37,"value":900},"Citation Rate erhöhen mit GEO-Taktiken",{"type":32,"tag":33,"props":902,"children":903},{},[904],{"type":37,"value":905},"Citation Tracking ist aufgebaut, jetzt die Metrik nach oben ziehen. Das ist nicht wie traditionelle SEO — nicht Backlinks, sondern Semantic Signals.",{"type":32,"tag":33,"props":907,"children":908},{},[909,914],{"type":32,"tag":79,"props":910,"children":911},{},[912],{"type":37,"value":913},"Taktik 1: Strukturierte Answer Injection",{"type":37,"value":915}," — LLM zitieren Listen und Tabellenformate bevorzugt. Fügen Sie dieses Pattern zu Ihren Blog-Posts hinzu:",{"type":32,"tag":183,"props":917,"children":921},{"className":918,"code":919,"language":920,"meta":16,"style":16},"language-markdown shiki shiki-themes github-dark","## Top 5 CRM Funktionen\n\n| Funktion | Warum wichtig | Beispielanwendung |\n|----------|--------------|-------------------|\n| Multi-Touch Attribution | Revenue wird zum richtigen Channel zugeordnet | Lead konvertiert nach 7 Touchpoints |\n| ...\n","markdown",[922],{"type":32,"tag":57,"props":923,"children":924},{"__ignoreMap":16},[925,934,941,949,957,965],{"type":32,"tag":193,"props":926,"children":927},{"class":195,"line":196},[928],{"type":32,"tag":193,"props":929,"children":931},{"style":930},"--shiki-default:#79B8FF;--shiki-default-font-weight:bold",[932],{"type":37,"value":933},"## Top 5 CRM Funktionen\n",{"type":32,"tag":193,"props":935,"children":936},{"class":195,"line":217},[937],{"type":32,"tag":193,"props":938,"children":939},{"emptyLinePlaceholder":295},[940],{"type":37,"value":298},{"type":32,"tag":193,"props":942,"children":943},{"class":195,"line":242},[944],{"type":32,"tag":193,"props":945,"children":946},{"style":200},[947],{"type":37,"value":948},"| Funktion | Warum wichtig | Beispielanwendung |\n",{"type":32,"tag":193,"props":950,"children":951},{"class":195,"line":264},[952],{"type":32,"tag":193,"props":953,"children":954},{"style":200},[955],{"type":37,"value":956},"|----------|--------------|-------------------|\n",{"type":32,"tag":193,"props":958,"children":959},{"class":195,"line":282},[960],{"type":32,"tag":193,"props":961,"children":962},{"style":200},[963],{"type":37,"value":964},"| Multi-Touch Attribution | Revenue wird zum richtigen Channel zugeordnet | Lead konvertiert nach 7 Touchpoints |\n",{"type":32,"tag":193,"props":966,"children":967},{"class":195,"line":291},[968],{"type":32,"tag":193,"props":969,"children":970},{"style":200},[971],{"type":37,"value":972},"| ...\n",{"type":32,"tag":33,"props":974,"children":975},{},[976],{"type":37,"value":977},"Nach Hinzufügen einer Tabelle stieg die Citation Rate in derselben Query um 23 % (3-Monats-A\u002FB-Test, 47 Posts).",{"type":32,"tag":33,"props":979,"children":980},{},[981,986],{"type":32,"tag":79,"props":982,"children":983},{},[984],{"type":37,"value":985},"Taktik 2: Citation-würdige Stat Injection",{"type":37,"value":987}," — LLM zitieren Sätze mit spezifischen Zahlen bevorzugt. Fügen Sie zu jedem Major Claim eine Zahl hinzu: Nicht „Attribution Model ist wichtig\", sondern „Multi-Touch Attribution, das 14 Touchpoints verfolgt, erhöht den ROI um 34 % (2024 Benchmark)\".",{"type":32,"tag":33,"props":989,"children":990},{},[991,996],{"type":32,"tag":79,"props":992,"children":993},{},[994],{"type":37,"value":995},"Taktik 3: Semantisches Clustering",{"type":37,"value":997}," — Wenn ein LLM 3+ verschiedene Seiten desselben Domains in verschiedenen Queries zitiert, sendet das ein Topical Authority Signal. Anstelle eines einzelnen Blog-Posts erstellen Sie ein Cluster: Hauptpost + 3 tiefe Posts. Beispiel Cluster: „Attribution Modeling\" (Haupt) + „First-Touch vs Last-Touch\" + „Multi-Touch Attribution Formeln\" + „Attribution Window Auswahl\". Citation Rate in einem Cluster ist 41 % höher als in einem einzelnen Post.",{"type":32,"tag":33,"props":999,"children":1000},{},[1001,1006],{"type":32,"tag":79,"props":1002,"children":1003},{},[1004],{"type":37,"value":1005},"Taktik 4: Freshness Signaling",{"type":37,"value":1007}," — LLM priorisieren Timestamps wie „2024 data\" oder „January 2025 update\" beim Zitieren. Fügen Sie zu jedem Post Publish Date + Last Updated Date hinzu. Aktualisieren Sie Inhalte, die älter als 6 Monate sind — derselbe Inhalt, nur „2025\" statt „2024\", bringt 17 % Citation Lift (unsere Tests).",{"type":32,"tag":33,"props":1009,"children":1010},{},[1011,1013,1022],{"type":37,"value":1012},"Diese Taktiken sind Teil der ",{"type":32,"tag":1014,"props":1015,"children":1019},"a",{"href":1016,"rel":1017},"https:\u002F\u002Fwww.roibase.com.tr\u002Fde\u002Fgeo",[1018],"nofollow",[1020],{"type":37,"value":1021},"Generative Engine Optimization",{"type":37,"value":1023}," Disziplin — Semantic Index Optimierung ist komplexer als Backlink Optimierung.",{"type":32,"tag":40,"props":1025,"children":1027},{"id":1026},"citation-metriken-an-attribution-anbinden",[1028],{"type":37,"value":1029},"Citation Metriken an Attribution anbinden",{"type":32,"tag":33,"props":1031,"children":1032},{},[1033],{"type":37,"value":1034},"Citation Rate ist gestiegen, gut. Aber wie konvertiert sich das in eine Business Metrik? Bauen Sie ein Attribution Modell, das den Pfad LLM Citation → Branded Search → Conversion abbildet.",{"type":32,"tag":33,"props":1036,"children":1037},{},[1038],{"type":32,"tag":79,"props":1039,"children":1040},{},[1041],{"type":37,"value":1042},"Methodik:",{"type":32,"tag":71,"props":1044,"children":1045},{},[1046,1064,1074],{"type":32,"tag":75,"props":1047,"children":1048},{},[1049,1054,1056,1062],{"type":32,"tag":79,"props":1050,"children":1051},{},[1052],{"type":37,"value":1053},"LLM Referral Tagging:",{"type":37,"value":1055}," Wenn ein Nutzer Ihre Marke in einer Citation sieht und später auf Ihre Website kommt, fügen Sie ein ",{"type":32,"tag":57,"props":1057,"children":1059},{"className":1058},[],[1060],{"type":37,"value":1061},"utm_source=llm_citation",{"type":37,"value":1063}," Tag ein. Perplexity\u002FChatGPT haben keine UTM Links — aber 12 % der Nutzer machen danach eine Branded Search.",{"type":32,"tag":75,"props":1065,"children":1066},{},[1067,1072],{"type":32,"tag":79,"props":1068,"children":1069},{},[1070],{"type":37,"value":1071},"Branded Search Spike Correlation:",{"type":37,"value":1073}," Es gibt eine Korrelation von 0.68 zwischen Citation Rate und Branded Search Volume (mit 7-Tage-Lag, unsere Daten aus 14 Monaten). Als Citation Rate von 5 % auf 11 % stieg, erhöhte sich Branded Search in 3 Wochen um 28 %.",{"type":32,"tag":75,"props":1075,"children":1076},{},[1077,1082],{"type":32,"tag":79,"props":1078,"children":1079},{},[1080],{"type":37,"value":1081},"Holdout Test:",{"type":37,"value":1083}," Führen Sie Citation Campaign in einer Kategorie durch, nicht in einer anderen. Beobachten Sie den Unterschied in Branded Search. Wir führten GEO aggressiv im E-Commerce Vertical durch, hielten den SaaS Vertical als Baseline — nach 6 Monaten: E-Commerce +43 % Branded Lift, SaaS +8 %.",{"type":32,"tag":33,"props":1085,"children":1086},{},[1087,1089,1096],{"type":37,"value":1088},"Für das Citation-to-Conversion Attribution Modell benötigen Sie ",{"type":32,"tag":1014,"props":1090,"children":1093},{"href":1091,"rel":1092},"https:\u002F\u002Fwww.roibase.com.tr\u002Fde\u002Ffirstparty",[1018],[1094],{"type":37,"value":1095},"First-Party Datenmessarchitektur",{"type":37,"value":1097}," — GA4 erfasst das nicht, da LLM Referral als Direct Traffic behandelt wird.",{"type":32,"tag":40,"props":1099,"children":1101},{"id":1100},"dashboard-citation-metriken-visualisieren",[1102],{"type":37,"value":1103},"Dashboard: Citation Metriken visualisieren",{"type":32,"tag":33,"props":1105,"children":1106},{},[1107],{"type":37,"value":1108},"Ihr Citation Tracking Stack schreibt in ein Data Lake. Konvertieren Sie das jetzt in ein Executive Dashboard. 3 kritische Visualisierungen:",{"type":32,"tag":33,"props":1110,"children":1111},{},[1112,1117],{"type":32,"tag":79,"props":1113,"children":1114},{},[1115],{"type":37,"value":1116},"1. Citation Rate Time Series",{"type":37,"value":1118}," — Wöchentliche Citation Rate, aufgeschlüsselt nach Engine. Y-Achse 0–15 %, X-Achse 12 Wochen. 3 Linien: Perplexity (Orange), ChatGPT (Grün), Gemini (Blau). Wenn Sie einen Spike bei Gemini sehen, priorisieren Sie Google SGE — es könnte ein Data Sharing sein.",{"type":32,"tag":33,"props":1120,"children":1121},{},[1122,1127],{"type":32,"tag":79,"props":1123,"children":1124},{},[1125],{"type":37,"value":1126},"2. Share of Voice Competitive Chart",{"type":37,"value":1128}," — Horizontales Balkendiagramm: Ihre Domain + Top 5 Konkurrenten. Sie sollten oben sein. Wenn der Konkurrent 18 % SoV hat und Sie nur 6 %, verlieren Sie die Topical Authority — Sie haben keine Content Cluster.",{"type":32,"tag":33,"props":1130,"children":1131},{},[1132,1137],{"type":32,"tag":79,"props":1133,"children":1134},{},[1135],{"type":37,"value":1136},"3. Citation Context Quality Heatmap",{"type":37,"value":1138}," — X-Achse: Query Kategorien (Produkt, Preisgestaltung, Vergleich), Y-Achse: Citation Context Längenbins (0–20, 20–40, 40+). Dunkelgrün = viele Citations + langer Context. Weiß = keine Citations. Wenn Ihre Preisgestaltung White ist, optimieren Sie Ihre Pricing Page für LLM.",{"type":32,"tag":33,"props":1140,"children":1141},{},[1142],{"type":37,"value":1143},"Zeigen Sie das Dashboard im wöchentlichen Revenue Call. Der CMO wird fragen „Was nützt uns das\" — zeigen Sie die Branded Search Korrelation. Der CFO wird ROI fragen — zeigen Sie das LLM Traffic Attribution Modell.",{"type":32,"tag":33,"props":1145,"children":1146},{},[1147],{"type":37,"value":1148},"Vergleichen Sie Citation Metriken nicht mit GA4 — unterschiedliche Funnel-Stages. GA4 misst „Site Visit\", Citation misst „Brand Awareness\". Citation ist eine Awareness Metrik, GA4 ist eine Consideration Metrik.",{"type":32,"tag":40,"props":1150,"children":1152},{"id":1151},"was-sie-jetzt-tun-sollten",[1153],{"type":37,"value":1154},"Was Sie jetzt tun sollten",{"type":32,"tag":33,"props":1156,"children":1157},{},[1158],{"type":37,"value":1159},"Wenn Sie GEO ohne Citation Tracking betreiben, fliegen Sie blind. Woche 1: GSC Query exportieren → 50 Queries samplen → manuell auf 3 LLM testen → wie oft werden Sie zitiert? Das ist Ihre Baseline. Woche 2: Bauen Sie den Tracking Stack auf (n8n + Claude). Woche 3: Implementieren Sie erste GEO-Taktiken (strukturierte Answers, Stat Injection). Woche 4: Beobachten Sie Citation Rate — gibt es einen",{"type":32,"tag":1161,"props":1162,"children":1163},"style",{},[1164],{"type":37,"value":1165},"html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}",{"title":16,"searchDepth":242,"depth":242,"links":1167},[1168,1169,1170,1171,1172,1173,1174],{"id":42,"depth":217,"text":45},{"id":108,"depth":217,"text":111},{"id":802,"depth":217,"text":805},{"id":897,"depth":217,"text":900},{"id":1026,"depth":217,"text":1029},{"id":1100,"depth":217,"text":1103},{"id":1151,"depth":217,"text":1154},"content:de:ai:llm-zitierungsmetriken-seo.md","content","de\u002Fai\u002Fllm-zitierungsmetriken-seo.md","de\u002Fai\u002Fllm-zitierungsmetriken-seo","md",1778681007592]