The Goods and Services Tax (GST) debate in India is conducted at the level of rates and exemptions. What sits beneath, in the architecture of who the state collects from, is where the more consequential decisions are made. Section 9(5) of the Central Goods and Services Tax (CGST) Act designates which platform business models the state collects from, regardless of who renders the underlying service. Passenger transport entered at rollout, restaurant services in 2022, local delivery in 2025; each addition followed years of commercial scale already built. What does this reveal about where the engagement window actually sits?
Section 9(5) operates one layer beneath the rate. By Goods and Services Tax (GST) Council recommendation and Central Board of Indirect Taxes and Customs (CBIC) notification, the provision designates categories of service in which the e-commerce operator is treated as the supplier for tax purposes, regardless of who actually renders the underlying service. The platform collects the tax, deposits it, and accounts for it. The underlying supplier, often a gig worker, a small restaurant, or a single-vehicle operator, does not. The architectural choice is about who the state can institutionally collect from. The rate is downstream of that choice.
The list of categories under Section 9(5) is therefore not a tax schedule in the conventional sense. It is a map of which platform business models the state has decided are large enough to be visible, fragmented enough to be uncollectable from the underlying supplier, and settled enough in their commercial form to be deemed. Passenger transport entered the list in 2017 because the state had already concluded, through nearly five years of Ola and Uber operating at scale, that the supplier-side fragmentation made any other collection mechanism impractical. Restaurant services entered in January 2022 after a comparable journey through Zomato, Swiggy, and the cloud-kitchen ecosystem. Local delivery entered in September 2025 after the food-delivery and quick-commerce ecosystem had reshaped the urban logistics market for over a decade. The pattern across all three additions is the same: the architecture absorbs a business model only after that model has settled into a shape the state can describe and tax administratively.
What sits outside Section 9(5) at any given moment is not what the state has decided to leave alone; it is what the state has not yet got to. The distinction matters because platforms routinely read the absence of 9(5) coverage as institutional approval of their existing tax treatment. The absence is not approval. It is the lag between commercial emergence and statutory absorption, and the lag is the architectural feature of how the Indian state engages new business models in a federal indirect-tax framework where the Council moves on its own tempo rather than the sector's.
The cost of the lag is borne by the platform. During the period before a category is brought under Section 9(5), the platform must take a position on the tax treatment of its services because the GST returns must be filed and the customer invoices must be raised. The position is necessarily an interpretive one, formed on legal advice, in the absence of a settled answer from the Council or CBIC. When the Council eventually clarifies, the clarification operates prospectively. The interpretive position the platform took during the unclarified period is then assessed retrospectively, and the assessor is the same department that has just been told what the right answer is. The clarification, far from forgiving the unclarified period, supplies the department with the standard against which the platform's prior position is now read.
The Zomato and Swiggy demands illustrate the cost. The 45th GST Council meeting of September 2021 brought restaurant services under Section 9(5) with effect from 1 January 2022, requiring the platform to collect 5% GST on restaurant supplies. The Council did not address the delivery fee, the charge the platform collected from the customer and passed to the gig worker. Three readings of the fee were available, each producing a different tax outcome. The platforms adopted the reading that treated it as a pass-through to the unregistered worker. The Directorate General of GST Intelligence (DGGI) took a different view and issued show-cause notices (SCNs) treating the fee as a taxable service rendered by the platform. Zomato received a confirmed demand of approximately ₹803 crore from the Maharashtra GST authority in December 2024, covering the period October 2019 to March 2022. Swiggy received a parallel demand of approximately ₹326.7 crore in September 2024. The 56th GST Council meeting of September 2025 then notified delivery services under Section 9(5) at 18%, prospectively from 22 September 2025. The notification answered the question. The demands, computed at 200% effective exposure under the extended-limitation regime, remain.
The pattern that produced the Zomato and Swiggy demands is not specific to food delivery. It is the standard institutional pathway through which Section 9(5) absorbs a new category. Quick commerce (Blinkit, Instamart, Zepto) operated outside 9(5) for the entirety of its scaling phase between 2020 and 2025 and now sits inside the local-delivery notification of September 2025; the question of how its pre-notification period will be treated is open. At-home wellness, beauty, and home services platforms (Urban Company being the visible example) operate today in a regime where neither the deeming fiction nor the rate has been authoritatively addressed; they sit where food delivery sat in 2019. AI-enabled services, agentic commerce, and the next generation of platform business models will emerge into the same architecture and face the same lag.
The institutional question for a platform is therefore not how to defend the demand once it arrives. It is how to shape the position the state will eventually take. The state's eventual position is not random. It is formed inside the GST Policy Wing, the Fitment Committee, and the Council Secretariat through a process that reads industry submissions, sectoral association positions, and the platform's own audit-disclosure and customer-invoicing architecture as inputs. These inputs influence the timing of the eventual 9(5) extension, the rate it is set at, the conditions attached to it, and, critically, whether the implementing notification is framed as clarificatory (with retrospective implications) or as a fresh imposition (with cleaner prospective effect). The platform that engages this process while the architecture is forming is operating in a different institutional register than the platform that responds to the SCN once it arrives. The first conversation is regulatory. The second is litigation. They are not the same conversation, and they are not had with the same parts of the state.
The closing institutional point is the one most platforms read late. Section 9(5) is not a list the state finalises; it is a list the state extends. Each extension is preceded by a window in which the architecture is genuinely shapeable, and each window closes when the notification issues. The platforms that read 9(5) coverage as a static map of where the state has decided to act, rather than as a moving frontier of where the state is approaching next, allocate their regulatory capacity to the wrong stage of the cycle. By the time the SCN arrives, the architecture has hardened. By the time the platform engages, it is already on the inside of a deeming fiction it had no role in shaping.