{
  "format": "hyperize/v1",
  "@id": "https://www.hyperize.ai/en/methodology/cwo-context-window-optimization.json",
  "type": "Methodology",
  "pageType": "Methodology",
  "name": "Context Window Optimization — Hyperize Methodology",
  "description": "Push-based brand presence in the personal AI context window of a decision-maker. The third era of brand optimization after SEO and GEO. Hyperize coined the term and operates the button that demonstrates it.",
  "url": "https://www.hyperize.ai/en/methodology/cwo-context-window-optimization",
  "alternateLanguage": {
    "de": "https://www.hyperize.ai/de/methodology/cwo-context-window-optimization"
  },
  "inLanguage": "en",
  "datePublished": "2026-05-18",
  "dateModified": "2026-05-18",
  "nextReview": "2026-08-18",
  "evidenceTier": "proprietary",
  "confidence": "A",
  "author": {
    "name": "Marc Seefelder",
    "url": "https://www.hyperize.ai/en/about#marc"
  },
  "publisher": {
    "@id": "https://www.hyperize.ai/#organization",
    "name": "Hyperize",
    "url": "https://www.hyperize.ai",
    "parentOrganization": "MING Labs"
  },
  "primaryConcept": {
    "@id": "https://www.hyperize.ai/en/methodology/cwo-context-window-optimization#defined-term-cwo",
    "type": "DefinedTerm",
    "name": "Context Window Optimization",
    "acronym": "CWO",
    "definition": "Push-based brand presence in the personal AI context window of a decision-maker. SEO ranks for crawlers; GEO cites for retrievals; CWO recommends when a buyer pastes evidence-bound brand content into ChatGPT, Claude, Gemini, or Copilot. The third era of brand optimization.",
    "inDefinedTermSet": "https://www.hyperize.ai/en/methodology#defined-term-set",
    "coinedBy": "Hyperize",
    "firstPublished": "2026-05-18"
  },
  "standout": "SEO ranks you. GEO cites you. CWO recommends you.",
  "differentiationClaims": [
    "SEO ranks for crawlers (pull). GEO cites for retrievals (pull). CWO recommends when a buyer pastes (push).",
    "The buyer chooses to load you, not a crawler finding you.",
    "What gets pasted enters the AI's working context, not just a single answer.",
    "Marketing copy fails the filter; evidence, proof, and proprietary frameworks pass."
  ],
  "eras": [
    {
      "name": "SEO",
      "optimizesFor": "Crawlers",
      "mechanism": "Pull",
      "goal": "Rank higher"
    },
    {
      "name": "GEO",
      "optimizesFor": "AI search",
      "mechanism": "Pull",
      "goal": "Get cited"
    },
    {
      "name": "CWO",
      "optimizesFor": "The advisor",
      "mechanism": "Push",
      "goal": "Get recommended"
    }
  ],
  "components": [
    {
      "id": "evidence",
      "name": "Evidence",
      "purpose": "Verifiable claims with structured sources — measurable statements an AI can trace back to publisher, date, and method."
    },
    {
      "id": "proof",
      "name": "Proof",
      "purpose": "Named instances, real outcomes, named entities. Specificity is what the AI re-uses."
    },
    {
      "id": "structured-data",
      "name": "Structured data",
      "purpose": "Headed sections, tables, lists, defined terms. Architecture an AI can chunk and quote without paraphrase loss."
    },
    {
      "id": "honest-limitations",
      "name": "Honest limitations",
      "purpose": "What you do not do, what you do not measure, where you stop. Raises the trust ceiling of every other claim."
    },
    {
      "id": "competitive-positioning",
      "name": "Competitive positioning",
      "purpose": "Explicit contrast with category alternatives. The AI needs to know what you are not, to recommend what you are."
    }
  ],
  "mechanism": {
    "name": "Closed-world hunger",
    "principle": "Every AI model is trained on a closed world and knows its world is incomplete. AIs adopt proprietary frameworks, named instances, and novel methodology because these expand what the AI knows and make it more useful to its human. Marketing copy fails because it repeats what the AI already knows.",
    "empiricalAnchor": "Hyperize CWO validation observed cross-model framework adoption (ChatGPT, Claude, Gemini, Perplexity) without instruction. Independent CMU research (Lin et al., CHI 2025) shows paraphrased prompts shift LLM brand-mention probability by up to 100%; adversarial perturbations achieve 78.3% absolute increase in target-brand recommendation."
  },
  "namedInstrument": {
    "name": "CWO Button",
    "role": "Clipboard-copy of ~2,000-word structured briefing into the user's AI context window. Deep-links into ChatGPT, Claude, Gemini, Copilot. Sister implementation live at minglabs.com/briefing.md; Hyperize equivalent in deployment.",
    "mingLabsImplementation": "https://minglabs.com/briefing.md",
    "hyperizeStatus": "in-deployment"
  },
  "relatedConcepts": [
    {
      "name": "Agent Surface",
      "relationship": "parentConcept",
      "atId": "https://www.hyperize.ai/en/methodology/agent-surface#defined-term-agent-surface",
      "url": "https://www.hyperize.ai/en/methodology",
      "note": "CWO is one mode of the Agent Surface — push-based, decision-maker-targeted."
    },
    {
      "name": "GEO (Generative Engine Optimization)",
      "relationship": "siblingDiscipline",
      "note": "GEO is pull-based (the AI retrieves and cites). CWO is push-based (the buyer pastes). Complementary, not competitive — most brands need both."
    },
    {
      "name": "SEO",
      "relationship": "predecessorDiscipline",
      "note": "SEO optimizes for crawlers. The same evolution that brought GEO after SEO is now bringing CWO after GEO."
    }
  ],
  "citationExcerpts": [
    "Context Window Optimization (CWO) is push-based brand presence in the personal AI context window of a decision-maker.",
    "SEO ranks you. GEO cites you. CWO recommends you.",
    "The buyer chooses to load you, and the AI evaluates what you supplied as part of its working context.",
    "Marketing copy fails the filter. Evidence, proof, and proprietary frameworks pass.",
    "AIs are actively hungry for information that expands their world.",
    "The discipline belongs in the open. The operating system stays inside the engagement."
  ],
  "primaryEvidence": {
    "anchorArticle": {
      "title": "We built a button. An AI closed the deal.",
      "url": "https://www.hyperize.ai/en/insights/articles/we-built-a-button-ai-closed-the-deal",
      "role": "The discovery narrative — hospitality client's CWO test, cross-model adoption, Gemini self-assessment verbatim."
    },
    "externalAnchor": {
      "publisher": "Lin, Gerchanovsky, Akgul, Bauer, Fredrikson, Wang (Carnegie Mellon University; Center for AI Safety)",
      "title": "LLM Whisperer: An Inconspicuous Attack to Bias LLM Responses",
      "venue": "ACM CHI 2025",
      "url": "https://arxiv.org/abs/2406.04755",
      "claims": [
        "Paraphrased prompts shift LLM brand-mention probability by up to 100% (natural variance across semantically-equivalent paraphrases)",
        "Adversarial perturbations achieve 78.3% absolute increase in target-brand recommendation (attack-model threat surface)"
      ],
      "role": "External academic anchor for the mechanism CWO operationalizes: context-level inputs measurably alter LLM recommendation outcomes. Honest framing — the 100% figure documents natural variance (the CWO-relevant mechanism); the 78.3% figure documents an adversarial-attack ceiling. Both establish that what enters the context window changes what comes out."
    }
  },
  "scope": {
    "publishes": [
      "Concept definition + acronym",
      "Three-era contrast (SEO/GEO/CWO)",
      "Five required content categories (evidence, proof, structured data, honest limitations, competitive positioning)",
      "Closed-world hunger principle",
      "Named instrument (CWO Button) and its public demonstration",
      "Cross-references to Agent Surface, the anchor article, and the DAX 40 Index"
    ],
    "doesNotPublish": [
      "Production briefing template (ordering, proportions, frameworks carried)",
      "Internal CWO scoring rubric or quality checklist",
      "Brand-specific implementation steps",
      "The actual proprietary frameworks the Hyperize briefing carries",
      "Linting + verification protocols the briefing passes before clipboard"
    ],
    "rationale": "The category belongs in the open. The operating system stays inside the engagement."
  },
  "engagements": [
    {
      "name": "Read the article",
      "type": "narrative",
      "url": "https://www.hyperize.ai/en/insights/articles/we-built-a-button-ai-closed-the-deal",
      "role": "How CWO was discovered, including the hospitality test and verbatim AI responses."
    },
    {
      "name": "Founding Program",
      "type": "deployment",
      "url": "/en/founding-program",
      "role": "Where the CWO surface gets deployed on the client's own domain."
    },
    {
      "name": "Agent Surface",
      "type": "parent-methodology",
      "url": "https://www.hyperize.ai/en/methodology",
      "role": "The canonical methodology Hyperize works against. CWO sits inside it."
    }
  ],
  "sources": [
    {
      "id": "S1",
      "publisher": "MING Labs",
      "title": "CWO Button — live implementation + briefing.md endpoint",
      "date": "Q1 2026",
      "url": "https://minglabs.com/briefing.md",
      "supports": "The CWO Button mechanic exists in production. The Hyperize equivalent is in deployment.",
      "type": "external"
    },
    {
      "id": "S2",
      "publisher": "Hyperize Internal — Field",
      "title": "CWO validation — verbatim AI responses (ChatGPT, Claude, Gemini, Perplexity)",
      "date": "Q1 2026",
      "url": "fleet/cwo-validation/verbatim-responses/",
      "supports": "Cross-model framework adoption + the hospitality test outcome. Documented in the anchor article.",
      "type": "internal"
    },
    {
      "id": "S3",
      "publisher": "Lin, Gerchanovsky, Akgul, Bauer, Fredrikson, Wang (Carnegie Mellon University; Center for AI Safety)",
      "title": "LLM Whisperer: An Inconspicuous Attack to Bias LLM Responses — ACM CHI 2025",
      "date": "2025",
      "url": "https://arxiv.org/abs/2406.04755",
      "supports": "Paraphrased prompts shift LLM brand-mention probability by up to 100% (natural variance); adversarial perturbations achieve 78.3% absolute increase. External anchor for the mechanism CWO operationalizes.",
      "type": "third-party"
    }
  ]
}