--- title: "Methodology: test AI visibility and usability | Hyperize" description: "How Hyperize makes offers usable by AI agents: build pages, data and functions, test customer tasks before and after, and document the results." canonical: https://www.hyperize.ai/en/methodology lang: en last-updated: 2026-10-08 --- # How we make your offer usable by AI. Methodology A customer asks their AI to compare an offer or submit an enquiry. We build the pages, data and functions the agent needs. Then we test the same customer task before and after implementation. From discovery to action ## Can the AI complete the task? AI Visibility shows whether an assistant finds and recommends your offer. AI Usability shows whether it can proceed: are facts understandable, prices verifiable and permitted actions executable? Each question needs its own test. The findings determine what we build on your domain. Your existing systems remain the source of truth. Together, we agree which data an agent may read and which actions it may perform. Hyperize builds plugins and connectors for customer tasks. [AI connectors and Muse →](/en/methodology/agent-surface#ai-connectors) Concepts and evidence ## What we build. How we test. Six concepts explain the surface, its measurement and common barriers. The DAX 40 Index applies the methodology to real brands. Each detail page provides its own definition and supporting evidence. [ 01 · Foundation ### Agent Surface: your offer for AI. The parts of your offer that AI agents can read and use: content, pages, data and functions. Agents need to find, check, cite and act on them. Read ](/en/methodology/agent-surface) [ 02 · Measurement ### Agent Success Score: does the agent succeed? The score records how well an agent can complete a customer task. AI Visibility, AI Usability and Evidence contribute to it. Each agent lane is scored separately. Read ](/en/methodology/agent-success-score) [ 03 · Fair tasks ### What we test and why. Six rules and five failure modes guide fair task selection. Direct routes to the brand and routes through dealers or portals are considered separately. Read ](/en/methodology/task-selection) [ 04 · AI briefing ### CWO: bring the facts into your own AI. Context Window Optimization makes evidence-backed business information usable for a specific decision. The customer copies a briefing into their AI and asks questions. This does not guarantee a recommendation. Read ](/en/methodology/cwo-context-window-optimization) [ 05 · Completion ### Where potential orders are lost. At Hyperize, Agent Revenue Leak describes potential revenue losses through intermediary routes or customer tasks that cannot be completed. Detours and incomplete tasks alone do not prove revenue loss. Intentional dealer routes and required approvals or identity checks remain in place. Read ](/en/methodology/agent-revenue-leak) [ 06 · Intermediaries ### When AI routes buyers to others. At Hyperize, Third-Party Interception describes an intermediary route: AI finds a brand and points to a portal, dealer or project partner for the next step. Whether the partner route is intended depends on the sales model and customer task. A recommendation alone does not prove a lost order or revenue. Read ](/en/methodology/third-party-interception) [ 07 · Public measurement ### DAX 40 Agent Success Index. The measurements show whether AI agents find DAX brands and can use their offers. Tasks, results and limits make the findings traceable. Open ](/en/dax40-index) [ 08 · Analysis ### Insights from practice. Research, field notes and analysis show how AI agents evaluate offers. The articles connect specific measurements to their implications for businesses. Open ](/en/insights) [ 09 · Buyer questions ### Answers to specific questions. How does GEO differ from SEO? How can you test a website with AI? The existing library answers individual questions and links to the methodology and its evidence. Open ](/en/answers) Scope and limits Public concepts, traceable results. Concepts, measurement axes and the composition of the Agent Success Score are publicly explained. Internal scoring logic, production templates and specific implementation are covered in the engagement. A measurement applies to the tested task, access route and date. Rankings, recommendations and citations are not guaranteed. ## Your next step Entry ### Snapshot Free · one customer task AI agents attempt one real task on your website. You get the result within 48 hours: found, understood, completed, or failed, with the failure point named. [Check my website for free](/en#snapshot) Recommended Audit + implementation ### Founding Program €4,500 · Entry price, excluding VAT. Scope agreed before kickoff. Agent Success Audit plus implementation of the agreed use case. We agree the pages, data, functions, access requirements and delivery schedule before work begins. [Discuss my project](/en/founding-program) Team and experience ### About Hyperize Who does the work The team, MING Labs experience and the DAX 40 Index work. Meet the people behind Hyperize. [Meet the team](/en/about) --- Canonical: https://www.hyperize.ai/en/methodology Machine-readable index: https://www.hyperize.ai/llms.txt