Machine Relations Index

What the Machine Relations Index tracks

The MRI monitors how answer engines — ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overviews — select sources when they respond to commercial research questions. It captures every root domain cited across a monitored set of B2B buyer-intent queries and classifies each domain by its source function using deterministic rules.

The index is not a ranking of brands or a quality score. It is a behavioral map of the source layer that machines use when constructing answers to buyer-intent queries. A domain appears in the MRI because at least one engine cited it in at least one observed query.

The MRI is a public research standard maintained at machinerelations.ai. AuthorityTech is a Machine Relations practitioner that reads the MRI as source-layer evidence. This page summarizes the standard and links to the canonical definitions rather than restating the methodology.

Source-role taxonomy

Every domain in the MRI is classified by deterministic rules into one of nine source roles:

• Editorial publication — news and trade media
• Analyst and consulting research — advisory and consulting firms
• Market and company database — data platforms and market research
• Academic and government source — institutional knowledge
• Community and social platform — user-generated surfaces
• Wire and press-release distribution — syndication networks
• Search or media platform — discovery infrastructure
• Vendor-owned source — explicitly identified company domains
• Other observed source — long-tail domains not yet classified by deterministic rules

What the index reports

The MRI reports observed citation behavior at the level of a source segment: one market category paired with one buyer question type. For each domain it tracks:

• Citation rate — the share of observed answer runs in a segment that cite the domain, used as the primary ranking signal
• Answer-engine breadth — how many of the monitored engines cite the source, revealing cross-engine trust
• Segment spread — how many category and question-type segments the source appears in
• Evidence state — whether a segment has cleared the evidence floor and publishes a rate, or is still collecting observations

Evidence floor and confidence

The MRI publishes a citation rate for a segment only after that segment clears an evidence floor of at least 10 observations across at least 7 distinct run dates. Below that line the segment is marked collecting rather than scored, so thin or early signal is never presented as settled authority. Each published domain carries a confidence grade — tiers A, B, or C, or collecting — reflecting how much evidence stands behind its rate.

Methodology version

The current standard is MRI Score v2.0, effective 2026-07-05. It superseded the earlier six-engine composite, MRI Score v1.1, which combined several weighted signals into a single number. v1 scores are not comparable to v2 citation rates; the two measure different things. The one-time v1-to-v2 crosswalk is the transition receipt. Earlier writeups that describe the MRI as a single weighted composite reflect the retired v1.1 methodology.

Public boundary

The public dataset reports citation rates, rankings, and evidence counts. It excludes internal query identifiers, raw cited URLs, and answer-engine provider payloads.

Why the MRI matters

For Machine Relations practitioners, the MRI answers a direct question: when a buyer asks an AI engine about your category, which sources does the engine reach for — and is yours among them?

The MRI was coined by Jaxon Parrott and is maintained as a public research artifact at machinerelations.ai/index. Machine-readable versions are available as JSON and Markdown.

Frequently Asked Questions

What does the Machine Relations Index track?

The MRI monitors how answer engines — ChatGPT, Perplexity, Gemini, Claude, Google AI Mode, and Google AI Overviews — select sources when they respond to commercial research questions. It captures every root domain cited across a monitored B2B buyer-query set and classifies each domain by its source function using deterministic rules.

How does the MRI decide which sources to publish?

The MRI reports a citation rate for a source segment only after that segment clears an evidence floor of at least 10 observations across at least 7 distinct run dates. Segments below the floor are shown as collecting, and each published domain carries a confidence grade reflecting the evidence behind its rate.

Why does the MRI matter for Machine Relations?

For Machine Relations practitioners, the MRI answers a direct question: when a buyer asks an AI engine about your category, which sources does the engine reach for — and is yours among them?