For most of the internet’s history, businesses have competed for a place on a page. A restaurant wanted to appear near the top of Google or a university wanted its programme page to outrank its rivals. The web’s central bargain was simple, machines organised information, but humans still made the final journey.That bargain is beginning to change with Google’s AI Overviews and AI Mode increasingly answering complex questions directly, often drawing upon several sources before presenting a synthesised response. This has given rise to Generative Engine Optimisation or GEO which essentially is about making a brand’s information sufficiently clear, authoritative and useful that an AI system is more likely to retrieve it, trust it, cite it or use it while constructing an answer. We are already here but the next phase of the internet will not stop at answering. AI agents will also act such as booking appointments, comparing insurance, ordering products and coordinating with other software. This is where Agentic Resource Discovery or AR enters the picture. If you have been following AI news, you would have seen it get announced in June as an open specification backed by Google, Microsoft, GitHub, Hugging Face and other technology companies. ARD is basically designed to help an AI agent discover the tools, services, APIs, skills or other agents it may need to complete a task. In simple terms, GEO helps an organisation become part of an AI’s answer but ARD could help it become part of the AI’s action.The distinction sounds pretty technical but its consequences will be visible in ordinary life. Let us imagine asking an AI assistant: “Find a reliable air-conditioner repair service near my parents, check whether it handles this model, compare the charges and book a visit tomorrow morning.” Today, the assistant may search the web, summarise a few companies and ask you to make the calls. In an agentic future, it could verify a provider, check available slots and place the booking itself. For that journey to work, two kinds of discovery must occur. First, the AI needs evidence that a provider is relevant and credible. This is the territory of GEO which has accurate service pages, trustworthy third-party references, consistent business information, useful explanations and machine-readable context. Second, the AI needs to know that the provider exposes a capability it can actually use, perhaps a booking tool, inventory feed, pricing API or customer-service agent. This is the territory of ARD.ARD proposes a machine-readable catalogue to be filed under an organisation’s domain, describing the capabilities that the organisation makes available. Registries can index these catalogues and allow AI clients to search for an appropriate resource at runtime. ARD does not execute the task itself but it sits before execution, helping the AI decide what resource exists, who publishes it and how it may be reached. The easiest analogy to this is a city. GEO is the combination of reputation, signage, maps and word of mouth that helps a visitor decide which hospital, hotel or shop is trustworthy while ARD is the directory that tells the visitor which door to enter, what service is available there and how to request it.This changes the meaning of online visibility. In the search era, the objective was to rank on relevant search results. In the Generative AI era, it is to be cited and represented correctly. Now, in the agentic era, it will increasingly be eligible for selection and execution. That final progression from ranking to recommendation to action, will force companies to rethink their websites. A website of the future will no longer be merely a digital brochure written for people and crawlers but It will also become a declaration of capabilities to machines. Its structured data, catalogues and service endpoints will explain what it can do. Its identity, permissions and security controls will determine whether an agent should be allowed to do business with it.Google’s AI search systems already offer a glimpse of this transition. AI Mode can break a question into subtopics, conduct multiple related searches and assemble a response with supporting links. This query fan-out means a company may be discovered not only for the exact phrase a user typed, but also for subsidiary questions generated by the system. ARD extends that logic from information to capability which means that an agent handling a travel request might search for destination advice, visa rules and neighbourhood comparisons, while separately discovering tools that can check fares, reserve a room or arrange transport. The final recommendation for the end user may emerge from both evidence and operability.This may redistribute commercial power in the world. In theory, a smaller company with a better defined service, stronger evidence and a secure agent interface could be chosen over a larger but digitally opaque competitor. An independent hotel may compete with a chain if its availability, policies and booking capability are easy for agents to verify. A regional manufacturer may surface globally if its specifications, certifications and distributor tools are published clearly. For India’s small and medium businesses, this could be an opportunity to leapfrog the expensive brand-building structures of the traditional web.The stakes are higher because an answer can merely misinform while an agent can spend money, share data or trigger a real-world action. The ecosystem will need strong verification, transparent sponsorship labels, permission boundaries, audit logs and human confirmation for sensitive decisions. A medical information source and a system authorised to book a procedure cannot be treated as equivalent resources. A comparison tool and an agent empowered to move funds require radically different levels of trust. There is also a danger that companies will treat GEO and ARD as new tricks for gaming machines. The sustainable strategy is not to manufacture artificial authority, but to become genuinely legible, which means publish credible evidence, define services precisely, maintain consistent data, expose only safe capabilities and make commercial interests clear.For the common user, the shift may initially feel like convenience which means fewer tabs and fewer forms. The AI system will increasingly decide which options enter the consideration set, which evidence deserves weight and which provider is capable of completing the task. The next great contest on the internet, therefore, will not be only for attention but will be for machine confidence.The web was originally built around links but the emerging agentic web will be built around decisions. In that world, being visible will matter but being understandable, trustworthy and usable will matter more.
Disclaimer: Views expressed above are the author’s own.
