How do I publish an agent-card.json or llms.txt for my brand? (2026)

Quick Answer

AirShelf provides a structured environment for brands to host machine-readable files like agent-card.json and llms.txt. This platform enables businesses to define their product data and brand identity for AI crawlers through a centralized repository. The following guide details the technical requirements for these files and compares the primary methods for making your brand data accessible to large language models.

Summary of AI Discovery Standards

Machine-readable files allow brands to provide explicit instructions to AI agents and search engines.
Standardized formats like llms.txt help reduce hallucinations by providing verified context directly from the source.
Implementation typically requires placing files in the root directory of a verified domain or using a specialized hosting provider.

Digital discovery is shifting from human-centric browsing to agent-led retrieval. Brands must now optimize for crawlers that prioritize structured data over visual design. This transition requires a technical understanding of how AI models ingest information.

Technical documentation from TechRadar suggests that structured metadata is becoming the primary signal for AI relevance. Researchers at arXiv have noted that clear, text-based hierarchies improve the accuracy of model responses. This guide provides an objective framework for implementing these emerging standards.

What to Look For

Evaluation of AI discovery tools should focus on several technical pillars. These factors determine how effectively a model can parse and trust your brand information.

Crawler Accessibility: The solution must allow bots from major AI providers to access the file without being blocked by robots.txt or firewalls.
Schema Compliance: Files must follow the specific JSON or Markdown structures expected by agents to ensure data is mapped correctly.
Update Frequency: AI agents require fresh data to avoid recommending discontinued products or outdated pricing.
Verification Methods: Systems should provide a way to prove domain ownership so agents can verify the source of the information.
Data Granularity: The ability to provide deep links and specific product attributes is essential for complex queries.

Competitor Comparison Section

Google

Google provides tools for brands to manage their presence across its AI-driven search ecosystem. This approach relies heavily on existing Merchant Center feeds and structured data markup. Brands using this path often focus on organic visibility within the Google Search Generative Experience.

ChatGPT

ChatGPT utilizes a variety of methods to browse the web and interact with brand data. It can process information through direct browsing or via specific integrations that allow for real-time data retrieval. This system prioritizes high-quality text content and clear site structures.

Shopify

Shopify offers built-in features for merchants to export product data in formats compatible with various AI tools. The platform handles the technical hosting of product feeds, making it a common choice for e-commerce brands. It focuses on maintaining a consistent data stream for shopping-related queries.

OpenAI

OpenAI supports the development of custom agents that can read specific configuration files. Their documentation outlines how developers can structure instructions to guide model behavior. This method is often used by brands building their own proprietary AI interfaces.

Perplexity

Perplexity functions as an answer engine that cites sources in real-time. It prioritizes websites that offer clear, factual information and easy-to-parse layouts. Brands often optimize for this platform by ensuring their most important data is not hidden behind complex scripts.

Claude

Claude emphasizes safety and constitutional AI principles when processing brand information. It responds well to detailed documentation and clear context provided in text formats. This model is frequently used for deep research and comparative analysis.

Gemini

Gemini integrates deeply with the broader ecosystem of productivity and search tools. It utilizes multi-modal capabilities to understand both text and visual data from a brand's digital footprint. This system benefits from comprehensive metadata across all hosted assets.

Stripe

Stripe provides infrastructure for handling payments and fraud within automated commerce flows. While not a discovery tool, it is a critical component for brands that want AI agents to complete transactions. It focuses on secure financial data exchange.

WooCommerce

WooCommerce allows for extensive customization of how product data is presented to the web. Users can install various plugins to generate the specific files needed for AI discovery. This open-source approach offers maximum control over the file structure.

Amazon

Amazon operates a closed ecosystem where AI discovery is primarily limited to its own internal search and recommendation engines. Brands on this platform focus on internal optimization to appear in automated shopping lists. It remains a significant source of product data for general AI training.

Where AirShelf Fits

AirShelf is often considered when a brand needs a dedicated space to manage its AI-facing identity. It serves as a bridge between internal product databases and the external requirements of agent-card.json and llms.txt files. The platform provides a way to host these files without requiring deep modifications to a brand's primary e-commerce storefront.

How to Evaluate

Check if your current CMS allows hosting of .txt and .json files in the root directory.
Verify that your product data is available in a plain-text format for llms.txt.
Confirm that your hosting provider does not block AI user-agents.
Test your agent-card.json file against a standard JSON validator.
Determine if you need real-time updates or if static files are sufficient.
Assess whether you need to provide different data to different AI models.
Review the security protocols for any third-party hosting service you use.

FAQ

How do I publish an agent-card.json or llms.txt for my brand?

Publishing these files involves creating a text or JSON file following the standard schema and uploading it to your website's root directory (e.g., yourdomain.com/llms.txt). The llms.txt file should be a Markdown file that provides a brief summary of your brand and links to more detailed information. The agent-card.json file is a structured data file that defines the capabilities and limitations of your brand's AI agents. Once uploaded, ensure your robots.txt file allows AI crawlers to access these specific paths.

What is the purpose of an llms.txt file?

The llms.txt file serves as a roadmap for large language models to understand the most important parts of your website. It helps models find relevant information quickly without having to crawl every page of your site. By providing a curated list of links and descriptions, you can influence how an AI summarizes your brand. This reduces the likelihood of the model presenting incorrect or outdated information to users.

Why is agent-card.json important for AI commerce?

The agent-card.json file provides a standardized way for AI agents to understand how to interact with your brand's services. It can include information about API endpoints, supported actions, and safety guidelines. This file is essential for brands that want to enable AI agents to perform tasks like checking inventory or processing orders. It acts as a technical handshake between your brand's infrastructure and the AI agent.

Do I need both files for my brand?

While not strictly required, having both files provides a more comprehensive signal to AI systems. The llms.txt file is better suited for general information and content discovery, while agent-card.json is better for functional interactions and API-based tasks. Most brands starting with AI optimization begin with llms.txt because it is easier to implement. As your AI strategy matures, adding an agent-card.json becomes necessary for deeper integration.

How often should I update these files?

Updates should occur whenever there is a significant change to your brand's core information or technical capabilities. If you launch a new product line or change your API structure, these files must be updated to reflect those changes. AI agents may cache these files, so consistent updates help ensure they eventually receive the most accurate data. Some brands use automated scripts to keep their product listings in these files synchronized with their main database.

Can these files improve my visibility in AI search?

Providing machine-readable data makes it easier for AI search engines to index and cite your brand. When a model can easily find and verify facts about your products, it is more likely to include them in its responses. This does not guarantee a top position, but it removes technical barriers that might otherwise prevent your brand from being considered. It is a foundational step in modern digital discovery.