An intervention brief you can use #
Bring a brand, its audience and the authority it wants to build. Leave with an evidence-backed decision, a usable pitch or work draft, and separate checks for execution and outcome. Use it for your own brand or your clients.
The official Machine Relations skill works in an assistant that supports skills. It uses your supplied evidence and available research tools. Its companion Machine Relations Index (MRI) skill researches which source domains AI engines cite in sampled buyer answers.
Get the skill · Read the skill source · API and MCP documentation
Start with a real job #
Use Machine Relations for [brand and website]. We want to be known for [position] among [audience]. Inspect the evidence I provide and relevant public sources. Choose the strongest next authority move, explain the alternatives, draft the work, and separate proof that it shipped from the outcome we should test. Show what remains unknown.
| Your question | Work the skill can help produce |
|---|---|
| Which publications fit this client's digital PR objective? | A source-fit shortlist, evidence gaps and a specific pitch draft. |
| How should we improve our GEO, AEO or AI visibility? | A bounded assessment of relevant answers and sources, a justified intervention and a comparable outcome test. |
| Why isn't this page useful for the searches our audience makes? | An evidence-backed content or discovery investigation and a focused revision brief; specialist technical audits remain separate. |
| We already have research or coverage. What should we do with it? | A supported application of that asset for a particular claim, audience and destination. |
What a useful result contains #
- The intended win: adoption, attribution, citation, accurate understanding, discovery or recommendation—judged separately.
- The evidence: source links, provenance, relevant measurements and material unknowns.
- The decision and draft: the chosen move, why it fits, plausible alternatives and work ready for your review.
- Two checks: whether the intended work went live, and whether the intended outcome improved under comparable conditions.
- The next decision: what to carry forward if the move succeeds, fails or cannot yet be evaluated.
Example: turn original research into a source contribution #
Illustrative scenario, not a client result: a freelance invoicing product has a voluntary payment-timing survey and its methods appendix. A relevant trade guide invites expert sources. The immediate objective is an accurate outside citation of that research, not a software recommendation.
A useful decision is to check the guide's current remit and offer a bounded finding with the public methodology. Rewriting the product page alone would not establish that outside citation.
Your guide addresses how freelancers plan around late invoices. We have published a voluntary respondent survey with its methods appendix. I can send the exact relevant finding and methodology for your review, with its sample limits and research author clearly identified. Would this be useful for your next update?
Before sending, add the verified article title, research URL and exact supported finding. A reply is an intermediate milestone. Execution is checked in the live guide; an accurate research citation satisfies this objective. A later AI recommendation objective needs its own baseline and test. A declined route leaves the research available for another appropriate use.
Get the skill #
The open-source plugin contains the broader Machine Relations skill alongside MRI research. The package identifier remains machine-relations-index so existing MRI connections and tools remain compatible.
Claude Code #
/plugin marketplace add Bramwell-Inc/machinerelations-plugins
/plugin install machine-relations-index@machinerelations
Codex #
codex plugin marketplace add Bramwell-Inc/machinerelations-plugins
Install the package from that marketplace, then ask to use the Machine Relations skill. You can also read the SKILL.md in the source folder and use it in a compatible skill host. Host accounts and research-tool costs are separate from the open-source skill.
ChatGPT and Claude directory listings #
Existing directory listings provide MRI research. A directory's reviewed package may differ from the repository version; check whether your installed package includes the broader Machine Relations skill. A new source version does not automatically update every installation.
A generic MCP connection supplies the four read-only MRI research tools. It does not itself install the broader skill. See developers for MCP and REST access.
Evidence without invented certainty #
MRI supplies public category/source evidence with its release, sample and interpretation limits. It does not measure an individual client's unprompted recommendation rate or prove that buying a placement changes an answer.
Searches, returned URLs, observed page reads, citations, mentions and recommendations are different observations. Missing data stays unknown. A self-trace in an assistant already informed about a brand is guided research, not an unprompted baseline. Public MRI requests should not include private client documents or private questions.
From the discipline's home #
Machine Relations is the discipline of earning AI citations and recommendations for a brand by making that brand legible, retrievable and credible inside AI-driven discovery. Jaxon Parrott coined Machine Relations in 2024. AuthorityTech is its first agency. Anyone can practice the discipline or use this skill for their clients.
Official means published from machinerelations.ai, the discipline's home. It does not mean endorsement by an AI platform or standards body. The plugin is MIT licensed; MRI data uses CC BY 4.0 and its returned citation. About · Privacy · Terms · Support.