What is AI Discovery Protocol?
The AI Discovery Protocol (ADP) is a technical specification for making web content optimally discoverable by AI systems. Unlike traditional SEO which optimizes for search engine crawlers, ADP adds machine-readable files that AI crawlers and agents, such as those behind ChatGPT, Claude and Perplexity, can use alongside your normal pages.
ADP has two parts. The core is a markdown copy of every content page, linked from the page itself, so a crawler that lands on the page can take a clean copy of the same text. Around that sit optional root files: llms.txt, a meta-index, a knowledge graph and update feeds.
The order follows our own crawler logs. In the 30 days to 12 November 2026 the markdown copies drew 23.9% of AI crawler requests on pressonify.ai, and all the root files together drew 1.5% (the data and the method).
Page Copies and Optional Endpoints
The core: a markdown copy of each page. Every release, article, topic page, learn guide and industry page on Pressonify has a copy under /okf/, served as markdown, carrying the same content as the page, and linked from the page body. The copies are marked noindex with a canonical link to their page, so search engines index the page and crawlers can still read the copy.
Optional root files, all listed in /ai-discovery.json and, with a link to each live response, on the endpoint index:
- /llms.txt - LLM context document (markdown)
- /ai-discovery.json and /.well-known/ai.json - Machine-readable index of the resources
- /knowledge-graph.json - Schema.org @graph entity catalog
- /feed.json - JSON Feed v1.1 for content updates
Alongside both: a standard /sitemap.xml with accurate lastmod dates and a robots.txt that leaves AI crawlers free to read the pages and their copies.
ADP HTTP Headers
The markdown copies and the root files send these HTTP headers:
- ETag - Cache validation for efficient re-crawling
- Content-Digest - SHA-256 integrity verification (RFC 9530)
- X-Update-Frequency - Update cadence of the endpoint (hourly/daily/weekly)
Implementing ADP
To implement ADP on your site, in this order:
- Generate a markdown copy of each content page from the same source as the page, and serve it as text/markdown
- Link to the copy from the visible body of the page, and add a rel="alternate" link in the head
- Keep robots.txt open to AI crawlers and your sitemap current
- Implement Schema.org JSON-LD on key pages
- Notify search engines through IndexNow when a page is published or corrected
- Add the optional root files (/llms.txt, a meta-index, a knowledge graph, /feed.json) if you want them
- Log which crawlers request which paths, so you can see what is being read
Pressonify implements ADP v3.1 for every press release published on the Pressonify platform. ADP adds a structured discovery layer on top of the crawlable pages and sitemaps that AI agents read most (see our crawl data).