TL;DR: COUNTER Code of Practice Release 5.1.1, published 30 June 2026, adds a new optional Access_Method: Agent category to the four COUNTER Reports (Platform, Database, Title, Item). It lets content providers report usage generated by AI systems — chatbots, research assistants, and agentic tools that fetch and synthesize publisher content — separately from ordinary human browsing (Regular) and automated text-and-data-mining harvesting (TDM). Where a report includes AI-related metrics, COUNTER now requires they be tagged Access_Method: Agent, using five new optional metrics: AI_Responses_Generated, Total_AI_Investigations/Unique_AI_Investigations, and Total_AI_Requests/Unique_AI_Requests.
What COUNTER Is, for Readers New to the Standard
COUNTER (Counting Online Usage of Networked Electronic Resources) is the usage-reporting standard maintained by Project COUNTER (now operating as COUNTER Metrics Limited). It exists to solve a specific, long-standing problem for libraries and content providers: before COUNTER, every publisher and database vendor counted and reported “usage” (downloads, views, searches) differently, which made it impossible for a library to compare how much its patrons actually used one licensed platform versus another, or to justify renewal spend with consistent numbers across vendors.
The Code of Practice defines exactly what counts as a “use,” how duplicate or automated activity is filtered out, and a common set of report formats — so a usage report from one publisher means the same thing as a usage report from another. The Code of Practice was first published in 2002. Release 5 (COP5) was published in 2017, with compliance required from January 2019, superseding Release 4 (2012). Release 5.1 followed in 2022 and became the required compliance baseline from January 2025. Content providers exchange these reports in bulk via the COUNTER API (originally introduced as SUSHI, the Standardized Usage Statistics Harvesting Initiative), which lets a library’s usage-management system pull reports automatically rather than downloading spreadsheets from each vendor by hand.
COUNTER organizes usage into four “Master Reports”: the Platform Report (PR), Database Report (DR), Title Report (TR), and Item Report (IR) — moving from the broadest level (an entire platform) down to the most granular (a single article, chapter, or item). Within these reports, usage is broken into Metric_Types (such as Total_Item_Requests and Unique_Item_Requests) grouped into categories like Searches, Investigations, Requests, and Access_Denied events, and further qualified by an Access_Type (whether access was via a paid subscription or an open-access route) and an Access_Method field.
What’s New in Release 5.1.1: Access_Method: Agent
Release 5.1.1, published 30 June 2026, is a minor-version update to COP5 that extends the existing Access_Method field to address a problem the standard’s original design didn’t anticipate: usage generated on a researcher’s behalf by an AI system, rather than by a person clicking through a browser.
Before 5.1.1, Access_Method distinguished only two categories:
- Regular — ordinary human browsing and reading behavior, the default for most usage.
- TDM (Text and Data Mining) — bulk, automated harvesting of content through a dedicated TDM API or interface, typically for computational analysis rather than reading.
Release 5.1.1 adds a third value:
- Agent — content and metadata accessed by an AI system, such as a chatbot, research assistant, or other agentic tool acting on a user’s behalf, rather than a human directly browsing the platform or a TDM process bulk-harvesting it.
Per COUNTER’s published best-practice guidance for this extension, Access_Method: Agent is an optional extension, available for use in the four COUNTER Reports (PR, DR, TR, IR) only — it does not apply to COUNTER’s simpler “Standard Views.” A report provider is not required to implement it, but where a provider does report AI-related metrics, those metrics MUST be tagged with Access_Method: Agent rather than folded into Regular or TDM totals. The guidance is aimed primarily at publishers who have AI systems (such as a chat-based discovery tool) embedded directly on their own platforms, and secondarily at third-party AI tool providers that access publisher content on a user’s behalf and want to report that usage back to the publisher or platform in a COUNTER-compatible way.
The New AI Metric_Types
Alongside the new Access_Method value, Release 5.1.1 introduces five new optional Metric_Types specifically for AI/agentic activity:
- AI_Responses_Generated (Platform Report) — a response delivered by an AI system to a user prompt. A pre-configured or automatically generated summary that isn’t produced in response to a specific user prompt does not count toward this metric.
- Total_AI_Investigations / Unique_AI_Investigations (PR, DR, TR, IR) — content chunks that the AI system selected and examined in the course of synthesizing a response, as distinct from content merely scanned during initial processing.
- Total_AI_Requests / Unique_AI_Requests (PR, DR, TR, IR) — full-text content actually made accessible to the AI system while generating a response.
Two exclusions are worth flagging for anyone implementing this on the reporting side: pre-configured or automatically generated summaries are explicitly excluded from AI_Responses_Generated, and content chunks that are only assessed during an AI system’s initial processing pass — without being selected for deeper use — must be excluded from the Investigations and Requests counts. Both rules exist to keep the new metrics measuring genuine AI-driven engagement with content rather than incidental scanning.
Why This Matters for Publishers, Platforms, and Libraries
The practical problem this update addresses is one that usage-statistics teams, license negotiators, and collection-development librarians have been running into for a while without a standard way to talk about it: AI crawlers, chatbots, and agentic research tools increasingly access licensed and open content on a user’s behalf, and that traffic looks nothing like a human clicking through pages in a browser. Without a defined way to separate it out, AI-driven access either gets invisibly folded into ordinary human usage counts (inflating or distorting them) or filtered out entirely as suspected bot traffic (undercounting a legitimate and fast-growing access pattern).
A defined Access_Method: Agent category gives report providers and consumers a consistent way to:
- Report AI-driven access separately from human reading behavior, so COUNTER’s core human-usage numbers (already used for collection-development and renewal decisions) aren’t distorted by a fundamentally different access pattern.
- Distinguish AI agent access from TDM: TDM covers bulk computational harvesting for analysis, while Agent access covers an AI system retrieving and synthesizing content to answer a specific user request — a materially different use case with different licensing and cost implications for publishers.
- Give libraries and consortia a starting point for understanding how much of their licensed content is being accessed via AI tools, which is increasingly relevant to license negotiation and to institutional AI-use policy.
Because the extension is optional and very new, coverage will vary by provider for some time — not every platform will implement Access_Method: Agent reporting immediately, and libraries evaluating vendor usage reports should not yet assume its absence means no AI-driven access is occurring, only that the provider hasn’t (or can’t yet) report it separately.
What Report Providers Need to Do
For a content provider or platform already producing COUNTER 5.1 reports, adopting the 5.1.1 AI extension is additive rather than a full-standard migration: it means adding the new Metric_Types and the Agent Access_Method value to existing PR, DR, TR, and IR reports where AI-driven access can be distinguished from Regular and TDM access at the platform level. Since the extension applies only to the four COUNTER Reports and not to Standard Views, providers offering only Standard Views to their usage-report consumers won’t be able to expose this breakout without also producing full Reports. As with any COUNTER minor-version update, report providers should confirm their usage-statistics vendor or in-house reporting pipeline has implemented the 5.1.1 specification before relying on AI-tagged data for licensing or budget decisions, and libraries should check with each vendor individually about their 5.1.1 adoption timeline rather than assuming uniform, immediate rollout across a collection.
Frequently Asked Questions
Is Access_Method: Agent required, or optional?
Optional. COUNTER’s guidance describes it as an optional extension for use in the four COUNTER Reports. However, where a report provider does include AI-related Metric_Types, that usage MUST be tagged Access_Method: Agent rather than reported under Regular or TDM.
How is Access_Method: Agent different from TDM?
TDM (text and data mining) covers bulk, automated harvesting of content through a dedicated API, typically for computational analysis of large volumes of text rather than for answering a specific question. Access_Method: Agent covers an AI system — a chatbot, research assistant, or similar agentic tool — retrieving and synthesizing content to generate a response to a particular user prompt. The two access patterns are both non-human, but represent different use cases with different implications for licensing and cost.
Does this apply to COUNTER’s Standard Views?
No. The Access_Method: Agent extension and its associated AI Metric_Types apply to the four COUNTER Reports (Platform, Database, Title, and Item Reports) only, not to Standard Views.
Who is this guidance aimed at?
Primarily publishers that have AI systems embedded on their own platforms, and secondarily third-party AI tool providers that access publisher content on a user’s behalf and want to report that usage in a COUNTER-compatible way.
When did Release 5.1.1 take effect?
It was published 30 June 2026 and is the current version of the COUNTER Code of Practice Release 5.
What is SUSHI, and is it still used?
SUSHI (the Standardized Usage Statistics Harvesting Initiative) was the original name for the automated protocol libraries use to pull COUNTER reports from vendors in bulk; it’s now referred to as the COUNTER API, and remains the standard mechanism for automated report retrieval under Release 5.1.1.
Related CASRAI Resources
- COUNTER 5 — the dictionary entry on the underlying Code of Practice, its history, and its four Master Reports.
- Persistent Identifiers and Citation for AI Models and Training Datasets
- AI Training Data Provenance, Copyright, and TDM Exceptions for Research
- Article Processing Charges (APCs) for Open Access
- Open Access Publishing: Models, Mandates, and Why It Matters
- Scholarly Publishing pillar
Sources: Project COUNTER / COUNTER Metrics Limited, COUNTER Code of Practice Release 5.1.1 documentation and Best Practice on AI Usage Reporting, both accessed 2026-07.
