In the September 17 Machine Relations Index v2 release, no editorial publication appeared in more than 9.73% of observed AI Visibility & GEO how_choose answer runs. The highest-cited source in the segment was YouTube at 25.66%. The next three were small AI visibility vendors or operator sites: Brandviz.ai at 13.27%, Mybrandi.ai at 11.50%, and Gen-optima.com at 11.50%.
That is the finding: when buyers ask answer engines how to choose an AI visibility or GEO tool, the engines are not primarily borrowing authority from the trade press. They are assembling answers from a platform layer, vendor-owned or vendor-adjacent explanations, community/social surfaces, and a long tail of specialized pages.
The release is mri_score_v2.0+2026-09-17+d5e23024fc20. Its window runs from 2026-05-10 through 2026-09-17. The public release covers 22,026 cited source domains, 123,538 citation events, 15,678 answer runs, 905 monitored prompts, and six engines: ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews, and Perplexity. The specific segment studied here is ai-visibility-geo/how_choose, with 113 observed answer runs across 7 run dates.
A companion study, AI visibility is one of the three least editorially sourced categories AI answer engines cite, measures the same gap at category level across all 16 measured subject categories in the September 18 release. This page is the segment-level view: which sources engines actually cite in the buyer decision shape.
The ranked source layer is not the editorial layer #
The top of the ai-visibility-geo/how_choose segment is a mixed source layer, not a media list.
| Rank in segment | Domain | Source role | Runs cited | Citation rate |
|---|---|---|---|---|
| 1 | YouTube | Search or media platform | 29 / 113 | 25.66% |
| 2 | Brandviz.ai | Other observed source | 15 / 113 | 13.27% |
| 3 | Mybrandi.ai | Other observed source | 13 / 113 | 11.50% |
| 3 | Gen-optima.com | Other observed source | 13 / 113 | 11.50% |
| 5 | Community and social platform | 12 / 113 | 10.62% | |
| 5 | Stackmatix.com | Other observed source | 12 / 113 | 10.62% |
| 7 | Airfleet.co | Other observed source | 11 / 113 | 9.73% |
| 7 | Averi.ai | Other observed source | 11 / 113 | 9.73% |
| 7 | Community and social platform | 11 / 113 | 9.73% | |
| 7 | Shadow.inc | Editorial publication | 11 / 113 | 9.73% |
Rank 7 is a four-way tie at 9.73%: Airfleet.co, Averi.ai, Reddit and Shadow.inc. The table as first published listed two of the four. It was completed on 2026-09-18 against the September 18 release, in which this segment is unchanged at 113 observed runs, 7 run dates and 448 domain rows.
The first editorial publication in the segment is Shadow.inc, tied at 9.73%. The next editorial publication is PR Newswire at 7.96%, followed by Gracker.ai at 6.19% and Demand Gen Report at 5.31%. Medium and HackerNoon each appeared in 2.65% of observed runs. TechRadar, CNBC, CMSWire, Search Engine Journal, Indie Hackers, and several smaller publication-classified domains each appeared in 0.88%.
The segment clears the MRI v2 evidence floor: at least 10 observations across at least 7 distinct run dates. The cited rate for each domain is a marginal run rate, not market share. A run can cite more than one domain, so these rates should not be added together or narrated as displacement.
What the gap means #
The editorial gap matters because how_choose is a buyer decision shape. It is the question a buyer asks before a shortlist: what criteria should I use, what category language matters, which tools or services deserve evaluation, and what evidence should I trust?
In this segment, answer engines often found that decision support outside conventional editorial media. YouTube was present in more than a quarter of observed runs. LinkedIn and Reddit were both in the top tier. Small vendor or operator domains appeared beside them. The largest editorial-publication rate was below 10%.
That does not mean editorial coverage is useless. It means the current source environment has an opening: the editorial layer has not yet become the canonical authority layer for AI visibility buying questions. For publications, analysts, and practitioners, this is a supply-side gap. For Machine Relations, it is the reason a source-side index matters: the market needs a neutral way to see which sources engines actually cite by category and question shape.
This study measures the market's source selection. It does not claim that any owned page wins the intent, and it does not use the Index as a scoreboard for any Bramwell domain.
Why this is a Machine Relations study #
Most AI visibility advice explains how brands can appear in answer engines. This study measures the supply side: which domains answer engines use as sources when the question is how to choose. That distinction is what Machine Relations exists to measure.
The Index ranks the cited source layer so that engines and market participants can quote a neutral, versioned source-selection record instead of a vendor claim. This page is one finding from one release; the Index publishes daily, and every figure here is stated against the September 17 release identifier.
The practical implication is simple: the editorial layer can still become the reference layer for AI visibility buying questions, but the September 17 release shows it is not there yet. A publication that wants to own this decision shape needs evidence that answer engines can reuse: criteria, comparisons, source lists, and versioned measurement. A vendor that wants to be cited cannot assume brand pages alone are enough; in this segment, engines mixed video, community, social, vendor-adjacent and editorial sources.
Methodology #
Machine Relations Research analyzed the public Machine Relations Index v2 JSON release generated on 2026-09-17. The release identifier is mri_score_v2.0+2026-09-17+d5e23024fc20, located from the market_source_selection observation with public artifact hash d5e23024fc206f9cb339ded268e441f98867bda82712ba7e708fcb1a7564b487.
The measurement was collected by the Machine Relations Index v2 pipeline, which observes six answer engines across a fixed basket of buyer and research prompts, records the source domains cited in answer runs, and publishes segment rates only after the evidence floor is met. The public release states the overall window, engine list, run count, source-event count, and the evidence floor; this study uses those public fields and does not use raw provider payloads or hidden query identifiers.
The segment filter was:
- category:
ai-visibility-geo - question shape:
how_choose - status:
published - observed answer runs: 113
- run dates: 7
- basket version:
mri_taxonomy_v2.0
For each domain in the release, the analysis selected its mri_score_v2.strata entry matching that category and question shape. It sorted the resulting 448 domain-stratum rows by citation_rate descending and read each domain's source_role and source_role_label from the same public record.
A domain's citation rate is calculated as runs citing that domain divided by observed runs for the segment. The rate is a marginal domain-present run rate. It is not citation-slot share, exclusive market share, or proof that one source displaced another. The confidence basis for the segment is the published MRI evidence floor; Shadow.inc, the top editorial-publication domain in this segment, carried an overall MRI confidence grade of B in the same release.
The evidence artifact used for this study was fetched from the public Machine Relations Index data endpoint. The rank-7 tie was completed on 2026-09-18 by re-reading the same segment in the September 18 release, mri_score_v2.0+2026-09-18+8fa38e54dd0a, where every figure in the table is unchanged.
FAQ #
What is the main finding? #
No editorial publication exceeded a 9.73% citation rate in the September 17 MRI v2 ai-visibility-geo/how_choose segment. YouTube led the segment at 25.66%, and several small AI visibility or operator domains ranked above the first editorial-publication source.
Does this say editorial publications do not matter? #
No. It says that in this specific buyer-question segment and release window, editorial publications were not the dominant source layer. The finding describes observed AI source selection, not the intrinsic value of editorial coverage.
Is 9.73% market share? #
No. It is a marginal citation rate: 11 of 113 observed answer runs cited the domain at least once. One run can cite multiple domains, so the rates are not mutually exclusive market-share percentages.
How does this relate to the category-level study? #
This page reads one buyer segment in one release. The companion study, AI visibility is one of the three least editorially sourced categories AI answer engines cite, sums citation weight by source role across every published segment in all 16 measured categories, and finds editorial publications carrying 5.9% in AI visibility and GEO against an 11.3% average. Read that page for the category comparison and this one for the segment ranking.