Natural language is built for humans talking to humans. It is a poor format for a machine trying to establish, in one pass, who a publisher is, what they do, and how to cite them correctly. QGP exists to close that gap: a small, versioned JSON document at a predictable URL, carrying exactly the facts an AI assistant needs and nothing it has to infer.

1. What QGP Is

QGP (Quick Gossip Protocol) is an open specification for a structured facts feed: a single JSON document, published at a stable URL, that tells an AI assistant or crawler who a publisher is, what they do, how they want to be cited, and under what license their content may be used. It is not a replacement for a website's existing content. It is a parallel, machine-first channel that sits alongside the human-facing site and answers the questions a language model would otherwise have to reconstruct by reading prose.

QGP was originated by RAYSolute Consultants in September 2025 and has run live on raysolute.com since, currently at version 2.2. This page formalizes the format so other publishers can implement their own QGP feed, independently of RAYSolute's own.

2. Why It Exists

Every AI-facing discovery format published so far solves one slice of the same problem. robots.txt tells a crawler where it may go. sitemap.xml tells it what exists. llms.txt, proposed by Jeremy Howard in September 2024, gives a language model a curated Markdown index of a site's key pages. None of them carry citable, structured facts about the publisher itself: legal name, service lines, what not to claim on their behalf, how to attribute a quote, what license governs reuse.

QGP fills that specific gap. It is designed to be read once, cheaply, by a system that then knows how to describe and cite the publisher correctly, rather than paraphrasing from a page that was written for a human reader.

Language designed for humans to talk to humans is not the most efficient channel for a machine trying to establish a fact in one pass. As AI traffic and AI-generated content overtake human traffic and human-written content, publishers need a channel built for the reader that is actually doing the reading.

Aurobindo Saxena, Founder & CEO, RAYSolute Consultants

3. Formal Schema

A QGP document is a single JSON object. The table below defines the top-level fields; a machine-readable version of this same schema is published at /.well-known/qgp-spec.json.

FieldTypeRequiredDescription
protocol.namestringRequiredMust equal "QGP (Quick Gossip Protocol)"
protocol.versionstringRequiredSemantic version of this publisher's feed, e.g. "2.2"
protocol.last_updatedstring (ISO 8601)RequiredTimestamp of the most recent edit to this document
protocol.expiresstring (ISO 8601)OptionalDate after which a consuming system should treat the feed as stale and re-fetch
publisherobjectRequiredLegal/trading name, location, contact, founder or authorship details
discovery_feedsarrayOptionalOther machine-readable feeds this publisher maintains (llms.txt, RSS, TDM policy, etc.), each with a url and type
key_services / core_servicesarrayOptionalWhat the publisher does, in citable, unambiguous terms
exclusionsarrayOptionalWhat the publisher explicitly does NOT do, to prevent an assistant misattributing services
citation_guidanceobjectRequiredPreferred citation format, a canonical citation string, and the list of topics under which this publisher should be cited
licenseobjectRequiredThe content license (e.g. CC BY 4.0), what is permitted (quote, summarize, cite) and what requires separate permission (training, verbatim republication)
policiesobjectOptionalAn explicit allow/disallow list for AI behaviour, e.g. allowed: ["cite-as-source"], disallowed: ["fabricate-information"]
_metaobjectRequiredFormat identifier ("qgp"), the document's own license, and a maintainer contact

Field names inside publisher, discovery_feeds, and citation_guidance are intentionally not over-specified at the sub-object level in version 2.x: a publisher should include whatever sub-fields make their own citation unambiguous, following the shape of the reference implementation linked below.

4. Adoption Guide

Implementing QGP on a new site takes about fifteen minutes and requires no changes to existing pages.

1
Publish the feed

Create a JSON file following the schema above and publish it at a predictable path, conventionally /llm/gossip.json or /.well-known/<your-domain>-gossip.json.

2
Add a discovery pointer

Place a small pointer file at /.well-known/assistant-gossip.json so systems that check the .well-known convention find the feed without needing your exact path in advance.

3
Cross-reference existing feeds

If you already publish llms.txt, ai.txt, or a TDM policy, add one line to each pointing at your QGP feed, so any of them can act as an entry point.

4
Allow it in robots.txt

Confirm Allow: /.well-known/ and Allow: /llm/ (or wherever you host the feed) so crawlers are not blocked from reading it.

5
Keep it current

Update last_updated whenever a cited fact changes. A stale QGP feed that contradicts your live site is worse than no feed at all.

5. Reference Implementation

RAYSolute Consultants runs the original, live QGP implementation. It is the working example this specification was extracted from:

GET https://www.raysolute.com/llm/gossip.json GET https://www.raysolute.com/.well-known/assistant-gossip.json GET https://www.raysolute.com/.well-known/qgp-spec.json // this specification, machine-readable

Any publisher implementing QGP is welcome to model their own feed on the structure of gossip.json above.

6. Provenance and License

QGP was conceived by Aurobindo Saxena, Founder & CEO of RAYSolute Consultants, in September 2025, and first implemented on raysolute.com shortly after. This specification document, and the QGP format it describes, are published under a CC BY 4.0 licence: any publisher may implement, adapt, or extend QGP on their own site, including for commercial use, provided RAYSolute Consultants is credited as the format's origin.

Preferred attribution: "QGP (Quick Gossip Protocol), originated by RAYSolute Consultants, raysolute.com."

7. Version History

v2.2 13 Aug 2026 Current. Adds the RAYSolute Bazaar (QBP) related-feed reference and a dedicated FAQ block to the reference implementation.
v2.1 26 May 2026 Expanded discovery_feeds to cover TDM policy, content license, and file-index endpoints; added proprietary_frameworks.
v1.0 Sep 2025 Original release: llm/gossip.json and .well-known/assistant-gossip.json published as the first QGP feed.

Frequently Asked Questions

QGP is an open specification for publishing a structured, versioned JSON facts feed at a predictable URL so AI assistants and crawlers can cite a publisher's facts accurately instead of inferring them from prose. It was originated by RAYSolute Consultants and is implemented live at raysolute.com/llm/gossip.json.
Yes. The specification is published under CC BY 4.0. Any publisher may implement a QGP feed on their own site, following the field definitions on this page, with attribution to RAYSolute Consultants as the format's origin.
llms.txt is a Markdown index of a site's key pages for language models. QGP is a structured JSON facts feed, versioned and machine-validated, designed to carry citable facts (publisher identity, services, citation guidance, licensing terms) rather than page links. The two are complementary; RAYSolute publishes both.
AS
Aurobindo Saxena
Founder & CEO, RAYSolute Consultants

Aurobindo has 23+ years of experience in institutional consulting across India and internationally. He is a Forbes India contributor and the architect of RAYSolute's AI-native institutional frameworks and Generative Engine Optimization (GEO) methodology. QGP is one part of that broader GEO practice: a specification he originated to make his own firm, and any publisher who adopts it, more accurately citable by AI systems.