Datasets:
Tasks:
Text Retrieval
Formats:
csv
Languages:
English
Size:
10K - 100K
Tags:
ai-search
generative-engine-optimization
answer-engine-citations
machine-relations
citation
answer-engines
License:
| license: cc-by-4.0 | |
| language: | |
| - en | |
| pretty_name: Machine Relations Index | |
| size_categories: | |
| - 10K<n<100K | |
| task_categories: | |
| - text-retrieval | |
| tags: | |
| - ai-search | |
| - generative-engine-optimization | |
| - answer-engine-citations | |
| - machine-relations | |
| - citation | |
| - answer-engines | |
| configs: | |
| - config_name: domains | |
| default: true | |
| data_files: | |
| - split: train | |
| path: domains.csv | |
| - config_name: segments | |
| data_files: | |
| - split: train | |
| path: segments.csv | |
| - config_name: engines | |
| data_files: | |
| - split: train | |
| path: engines.csv | |
| # Machine Relations Index | |
| Machine Relations Index (MRI) by Jaxon Parrott, Machine Relations Research — machinerelations.ai. | |
| Publisher: Machine Relations Research. Creator: Jaxon Parrott. License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). | |
| Machine Relations (MR) is the discipline of earning AI citations and recommendations for a brand by making that brand legible, retrievable, and credible inside AI-driven discovery. It was coined by Jaxon Parrott, founder of AuthorityTech, in 2024. | |
| Keywords: AI search, generative engine optimization, answer engine citations, Machine Relations. | |
| Weekly public snapshot of the Machine Relations Index. Version `mri_score_v2.1+2026-11-10+32349d1896b9`, generated 2026-11-10, window 2026-05-10 to 2026-11-10. Methodology `mri_score_v2.1`. 36876 cited source domains, 157 segments, 6 answer engines. Evidence floor: 10 observed runs and 7 distinct run dates. | |
| `machine-relations-index.json` is the public artifact copied byte for byte. Its sha256 is `7cb48ba880040a84917c82a756aa7ea3b9a8e20898c29416cca845866e026921` and matches `mri-release-manifest.json`. The CSV files are flat views of that artifact for the dataset viewer. They are not a second measurement. | |
| ## Citation | |
| If you use this dataset, cite it as Machine Relations Index (MRI) by Jaxon Parrott, Machine Relations Research — machinerelations.ai. | |
| Concept DOI: https://doi.org/10.5281/zenodo.23652423. Zenodo record: https://zenodo.org/records/23652423. | |
| Hugging Face dataset: https://huggingface.co/datasets/machine-relations/machine-relations-index. | |
| ```bibtex | |
| @misc{parrottMriDataset, | |
| author = {Parrott, Jaxon}, | |
| title = {Machine Relations Index}, | |
| year = {2026}, | |
| version = {mri_score_v2.1+2026-11-10+32349d1896b9}, | |
| publisher = {Machine Relations Research}, | |
| url = {https://machinerelations.ai/index}, | |
| doi = {10.5281/zenodo.23652423}, | |
| license = {CC-BY-4.0} | |
| } | |
| ``` | |
| ## Links | |
| - Index: https://machinerelations.ai/index | |
| - Methodology: https://machinerelations.ai/research/machine-relations-index-methodology | |
| - API: https://machinerelations.ai/api/mri/v2 | |
| - MCP: https://machinerelations.ai/mcp | |
| - Weekly Index Report: https://machinerelations.ai/research/machine-relations-index-report | |
| ## Files | |
| - `machine-relations-index.json`: public Index artifact. | |
| - `mri-release-manifest.json`: release id, methodology version, artifact sha256. | |
| - `mri-crosswalk-v1-v2.json`: frozen v1 to v2 comparison. | |
| - `domains.csv`: one row per cited source domain. | |
| - `segments.csv`: one row per category and question shape. | |
| - `engines.csv`: the answer-engine roster for this release. | |
| - `release.json`: release id, publication date, artifact sha256. | |
| - `CITATION.cff`: Citation File Format for this snapshot. | |
| - `LICENSE`: CC BY 4.0. | |
| - `zenodo-metadata.json`: the Zenodo metadata sent with this snapshot. | |
| ## Data dictionary | |
| ### domains.csv | |
| - `rank` | |
| - `domain` | |
| - `display_name` | |
| - `source_role` | |
| - `source_role_label` | |
| - `citation_rate` | |
| - `runs_cited` | |
| - `runs_observed` | |
| - `engine_count` | |
| - `engines` | |
| - `confidence` | |
| - `days_cited` | |
| - `days_observed` | |
| - `temporal_consistency` | |
| - `position_avg` | |
| - `position_best` | |
| - `first_cited` | |
| - `last_cited` | |
| - `full_universe_rank` | |
| - `full_universe_percentile` | |
| `engines` is a `|`-separated list of answer engines that cited the domain. `citation_rate` is cited runs divided by observed runs. | |
| ### segments.csv | |
| - `category` | |
| - `question_shape` | |
| - `status` | |
| - `not_collectable` | |
| - `collecting_since` | |
| - `runs_observed` | |
| - `distinct_domains_cited` | |
| - `run_dates` | |
| - `distinct_domains_cited`: On a segment (strata and by_category strata): how many different source domains answer engines cited at least once in that segment. A count of domains, not of answer runs. | |
| - `runs_cited_domains`: The same number as distinct_domains_cited, kept under its earlier name for existing readers. Despite the name it counts domains, not answer runs. | |
| - `cited_domain_observations`: In stratum_summary: distinct_domains_cited summed over every segment, so a domain cited in several segments counts once in each. Not a count of answer runs or of citations. | |
| - `resolved_source`: How a citation is counted: under the domain of the cited page, except that a page on a hosting platform which the Index has mapped, with at least 0.8 confidence, to the channel or publication that made it counts for that channel's or publication's own domain instead. Mapped in this release: youtube.com. That domain counts only its pages that are not mapped, never the whole platform. | |
| - `hosted_platforms`: The hosting platforms whose pages the Index maps to the channel or publication that made them (resolved_source). Wherever one of these domains is named, it counts only its pages that are not mapped, never the whole platform. | |
| - `runs_citing_any_source`: In collection_regimes: how many observed answer runs cited at least one source. A run that cited no source still counts in runs_observed. | |
| - `median_domains_per_citing_run`: In collection_regimes: among the runs that cited at least one source, the median number of different source domains one run cited. Null when no run cited a source. | |
| ### engines.csv | |
| - `engine` | |
| - `requested_model` | |
| - `resolved_model` | |
| - `status` | |
| - `days_observed` | |
| - `runs_total` | |
| - `runs_observed` | |
| - `last_success_date` | |
| - `last_observed_date` | |
| Model fields are empty when the provider reports none. | |