This is the first edition on the corrected methodology, mri_score_v2.1, and it compares against a release still on v2.0, so the week-over-week rates below divide by two different denominators. v2.1 drops 948 runs from the 81 engine-days on which Gemini or one of the two Google answer surfaces returned no citation at all across five or more runs: runs that could not count toward any source's rate, removed from every rate's denominator alike. Read like for like, Reddit's rate fell from 11.27% to 10.02%, 1.25 points rather than the 0.75 the movers table shows: on the corrected basis the week-earlier release had 20,380 runs, Reddit was cited in 2,296 of them, and it is cited in 2,399 of 23,940 now. It is last week's dilution again, and larger. Reddit was cited in 103 more runs, but observed runs grew 17% as the eligible question set widened from 1,073 to 1,108. The four largest decliners are again platforms (Reddit, YouTube, Medium and LinkedIn), and each was cited in more runs than a week earlier. Forbes holds fourth at 4.09%, still the only editorial publication in the top five.
In the October 12, 2026 release of the Machine Relations Index (mri_score_v2.1+2026-10-12+7925b6c15feb), reddit.com was cited in 10.02% of observed AI answer runs, 2,399 of 23,940: the highest citation rate of the 29,665 source domains the Index has seen cited, at confidence tier A. In the release a week earlier (mri_score_v2.0+2026-10-05+0a0c7e02957b), its rate was 10.77%, 2,296 of 21,328 runs. The two releases use different methodology versions (mri_score_v2.0 a week earlier, mri_score_v2.1 now), so part of every change between them comes from the method, not from how engines cite; the correction record states what changed. Across the 149-day window it was cited by Gemini, Google AI Mode, Google AI Overviews and Perplexity, and not by ChatGPT or Claude.
This is the weekly Machine Relations Index Report, built by code from the Index's daily public release and compared with the release of October 5, 2026. Every figure links to the Index page that publishes it. The previous edition is Machine Relations Index Report, October 7, 2026: reddit.com is cited in 10.44% of 22,344 observed AI answer runs.
The release at a glance #
| Measure | Release 2026-10-05 | Release 2026-10-12 | Change |
|---|---|---|---|
| Observation window | 142 days | 149 days | +7 days |
| Observed answer runs | 21,328 | 23,940 | +2,612 |
| Citation events | 164,938 | 183,746 | +18,808 |
| Cited source domains | 27,431 | 29,665 | +2,234 |
| Eligible questions in the basket | 1,073 | 1,108 | +35 |
| Published segments | 103 of 151 | 115 of 151 | +12 |
| Segments still collecting | 48 | 36 | -12 |
| Graded domains, tiers A / B / C | 17 / 80 / 574 | 19 / 101 / 648 | +97 |
A segment is one subject category paired with one question shape. It publishes once it has 10 observed runs across 7 distinct run dates. A citation rate is the share of observed answer runs in which at least one engine cited the domain.
The ten most-cited sources #
| Rank | Source | Source role | Citation rate | Runs cited | Tier | Engines citing | Rank a week earlier |
|---|---|---|---|---|---|---|---|
| 1 | reddit.com | Community and social platform | 10.02% | 2,399 of 23,940 | A | 4 of 6 | 1 |
| 2 | youtube.com | Search or media platform | 7.32% | 1,753 of 23,940 | A | 6 of 6 | 2 |
| 3 | linkedin.com | Community and social platform | 5.87% | 1,406 of 23,940 | A | 6 of 6 | 3 |
| 4 | forbes.com | Editorial publication | 4.09% | 979 of 23,940 | A | 6 of 6 | 4 |
| 5 | medium.com | Community and social platform | 3.94% | 943 of 23,940 | A | 6 of 6 | 5 |
| 6 | arxiv.org | Academic and government source | 2.85% | 682 of 23,940 | A | 6 of 6 | 6 |
| 7 | techradar.com | Editorial publication | 2.85% | 682 of 23,940 | A | 5 of 6 | 7 |
| 8 | g2.com | Market and company database | 2.64% | 631 of 23,940 | A | 6 of 6 | 9 |
| 9 | nih.gov | Academic and government source | 2.56% | 613 of 23,940 | A | 6 of 6 | 8 |
| 10 | gartner.com | Analyst and consulting research | 2.20% | 527 of 23,940 | A | 6 of 6 | 10 |
Rates are measured over all 23,940 observed runs, so they read lower than any single segment's rate. "Engines citing" counts the engines observed citing the domain at least once in the window. The public release does not publish a per-engine citation rate, so this report does not rank sources engine by engine.
The ten most-cited editorial publications #
| Rank among 1,706 classified editorial publications | Publication | Citation rate | Runs cited | Tier | Engines citing |
|---|---|---|---|---|---|
| 1 | forbes.com | 4.09% | 979 of 23,940 | A | ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity |
| 2 | techradar.com | 2.85% | 682 of 23,940 | A | ChatGPT, Claude, Gemini, Google AI Overviews and Perplexity |
| 3 | yahoo.com | 2.12% | 507 of 23,940 | A | ChatGPT, Claude, Google AI Mode, Google AI Overviews and Perplexity |
| 4 | healthline.com | 1.80% | 430 of 23,940 | B | ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity |
| 5 | nerdwallet.com | 1.63% | 390 of 23,940 | A | ChatGPT, Claude, Google AI Mode, Google AI Overviews and Perplexity |
| 6 | techtarget.com | 1.42% | 341 of 23,940 | A | ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity |
| 7 | nytimes.com | 1.04% | 248 of 23,940 | B | Google AI Mode, Google AI Overviews and Perplexity |
| 8 | cnbc.com | 1.01% | 241 of 23,940 | B | ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity |
| 9 | peoplemanagingpeople.com | 0.89% | 214 of 23,940 | B | ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity |
| 10 | pcmag.com | 0.87% | 208 of 23,940 | B | Gemini, Google AI Mode, Google AI Overviews and Perplexity |
Engine coverage #
| Engine | Model | Status | Days answered | Runs returned | Runs with a citation | Runs returned a week earlier |
|---|---|---|---|---|---|---|
| ChatGPT | gpt-4.1-mini-2025-04-14 | healthy | 145 | 5,386 | 4,731 | 4,554 |
| Claude | claude-haiku-4-5-20251001 | healthy | 148 | 5,484 | 4,477 | 4,648 |
| Gemini | gemini-3.8-flash | degraded | 127 | 5,483 | 2,061 | 4,648 |
| Google AI Mode | not disclosed | healthy | 107 | 1,792 | 1,410 | 1,664 |
| Google AI Overviews | not disclosed | healthy | 90 | 1,252 | 802 | 1,159 |
| Perplexity | sonar | healthy | 149 | 5,491 | 5,490 | 4,655 |
Engine counts come from the release manifest at machinerelations.ai/data/mri-release-manifest.json. An engine that answers on fewer days contributes fewer runs to every rate above.
Largest week-over-week movers among graded domains #
Of the 670 domains graded A, B or C in both releases, these rates moved most. Change is the difference between the two published cumulative rates, in percentage points. It records that a rate moved, not why: in the same week the basket went from 1,073 to 1,108 eligible questions and observed runs from 21,328 to 23,940, so a moved rate can reflect what was asked as much as how engines cite. The two releases use different methodology versions (mri_score_v2.0 a week earlier, mri_score_v2.1 now), so part of every change between them comes from the method, not from how engines cite; the correction record states what changed.
Rates that rose #
| Source | Tier now (a week earlier) | Rate now | Rate a week earlier | Change |
|---|---|---|---|---|
| rework.com | C (C) | 0.29% | 0.15% | +0.14 pts |
| nhimg.org | B (C) | 0.52% | 0.40% | +0.12 pts |
| guideflow.com | A (B) | 1.37% | 1.26% | +0.11 pts |
| worldmetrics.org | B (B) | 0.74% | 0.64% | +0.10 pts |
| alibaba.com | B (B) | 0.89% | 0.80% | +0.09 pts |
Rates that fell #
| Source | Tier now (a week earlier) | Rate now | Rate a week earlier | Change |
|---|---|---|---|---|
| reddit.com | A (A) | 10.02% | 10.77% | -0.75 pts |
| youtube.com | A (A) | 7.32% | 7.62% | -0.30 pts |
| medium.com | A (A) | 3.94% | 4.17% | -0.23 pts |
| linkedin.com | A (A) | 5.87% | 6.08% | -0.21 pts |
| nih.gov | A (A) | 2.56% | 2.76% | -0.20 pts |
Segment leaders that changed #
| Category | Question shape | Leader a week earlier | Leader now | Rate now |
|---|---|---|---|---|
| AI Infrastructure | Problem-first research | arxiv.org and reddit.com (tied), 15.76% | arxiv.org | 14.36% |
| AI Visibility & GEO | Top lists | tryprofound.com and youtube.com (tied), 22.31% | youtube.com | 22.50% |
| Consumer Finance | Problem-first research | youtube.com, 24.61% | experian.com | 23.64% |
| Cybersecurity | Comparisons | exabeam.com, 24.82% | vectra.ai | 26.62% |
| Deep Tech & Hardware | Comparisons | reddit.com, 19.26% | nvidia.com and reddit.com and youtube.com (tied) | 17.31% |
| Education & Training | Is it worth it | reddit.com, 18.46% | theworkademy.com | 18.98% |
| Education & Training | Problem-first research | indeed.com and linkedin.com (tied), 13.08% | indeed.com | 14.38% |
| Education & Training | Comparisons | reddit.com, 21.71% | uopeople.edu | 29.45% |
| Enterprise Software | Problem-first research | zapier.com, 11.03% | reddit.com and zapier.com (tied) | 10.14% |
| Enterprise Software | Top lists | gartner.com, 11.11% | procuredesk.com | 17.69% |
| Fintech | Top lists | stripe.com, 25.19% | crassula.io and stripe.com (tied) | 23.53% |
| Healthcare Services | Problem-first research | medicalbillersandcoders.com and youtube.com (tied), 9.49% | youtube.com | 11.92% |
| HR & Talent | Problem-first research | paychex.com, 17.56% | linkedin.com | 17.99% |
| iGaming & Betting | Best tools | linkedin.com, 18.87% | tigsportsbook.com | 21.09% |
| iGaming & Betting | How buyers choose | softswiss.com and track360.io (tied), 17.14% | jadexconsulting.com | 19.67% |
| iGaming & Betting | Problem-first research | reddit.com, 8.49% | optikpi.com | 8.94% |
| iGaming & Betting | Top lists | smartico.ai, 18.37% | geocomply.com | 28.00% |
| iGaming & Betting | Comparisons | track360.io, 20.59% | rotowire.com | 22.66% |
A leader is the graded domain outside the publisher's related parties with the highest citation rate in the segment, on the segment's own denominator; domains tied on rate are listed together.
Segments that crossed the evidence floor #
| Category | Question shape | Runs observed | Run dates | Leader | Rate |
|---|---|---|---|---|---|
| Industrial | Best tools | 100 | 7 | iqsdirectory.com | 12.00% |
| Industrial | How buyers choose | 97 | 7 | industrialmonitordirect.com | 8.25% |
| Industrial | Is it worth it | 99 | 7 | linkedin.com | 11.11% |
| Industrial | Problem-first research | 104 | 7 | youtube.com | 6.73% |
| Industrial | Top lists | 103 | 7 | iqsdirectory.com | 4.85% |
| Industrial | Comparisons | 106 | 7 | stamtec.com | 29.25% |
| Local Services | Best tools | 106 | 7 | angi.com | 27.36% |
| Local Services | How buyers choose | 104 | 7 | serviceagent.ai | 16.35% |
| Local Services | Is it worth it | 104 | 7 | reddit.com | 22.12% |
| Local Services | Problem-first research | 106 | 7 | angi.com | 12.26% |
| Local Services | Comparisons | 101 | 7 | consumeraffairs.com | 18.81% |
| Logistics & Freight | Problem-first research | 70 | 7 | linkedin.com and wearewarp.com (tied) | 8.57% |
Segment leaders by category #
The leader of every published segment, with its citation rate on that segment's runs. "Collecting" segments have not yet cleared the evidence floor of 10 runs across 7 run dates and carry no leader.
Do the cited sources support the answer? #
Answer-Source Fidelity is a separate measurement of claim support, not citation frequency. This edition preserves the published Fidelity snapshot generated 2026-10-11T11:44:56.712Z; it is not a week-only rate and may cover a different observation period from the Index.
Legacy baseline — frozen corpus, grade-time fetch basis #
Semantic cohort: c74383df31771cfe. Cohorts with different measurement semantics are not pooled.
- ChatGPT (reported): Of the 1255 resolvable citations for this engine, 54.8% were Supported; resolvable coverage was 77% (1255/1630).
- Claude (collecting): Engine still COLLECTING: only 88 of 1361 citations were resolvable (coverage 6.5%) — below the evidence floor, so no fidelity rate is reported.
- Gemini (collecting): Engine still COLLECTING: only 392 of 4349 citations were resolvable (coverage 9%) — below the evidence floor, so no fidelity rate is reported.
- Perplexity (reported): Of the 1298 resolvable citations for this engine, 50.1% were Supported; resolvable coverage was 47.6% (1298/2726).
Collection-time series — snapshot basis, bounded snapshot retry + deferral (F4-3) #
Semantic cohort: 06ddb22c15fa05d0. Cohorts with different measurement semantics are not pooled.
- ChatGPT (reported): Of the 9492 resolvable citations for this engine, 53.2% were Supported; resolvable coverage was 76.3% (9492/12442).
- Claude (collecting): Engine still COLLECTING: only 0 of 0 citations were resolvable (coverage 0%) — below the evidence floor, so no fidelity rate is reported.
- Gemini (collecting): Engine still COLLECTING: only 0 of 0 citations were resolvable (coverage 0%) — below the evidence floor, so no fidelity rate is reported.
- Perplexity (collecting): Engine still COLLECTING: only 0 of 0 citations were resolvable (coverage 0%) — below the evidence floor, so no fidelity rate is reported.
Published dataset SHA-256: 4791ce612b50d145309442497078516d76375b29891224dd29fd1bcf7c149855. Current measurement and method · Current dataset. The linked dataset advances; the snapshot above stays fixed with this edition.
How this report is built #
- Source: Machine Relations Index release
mri_score_v2.1+2026-10-12+7925b6c15feb(methodology mri_score_v2.1, window 2026-05-10 to 2026-10-12), compared withmri_score_v2.0+2026-10-05+0a0c7e02957b. The current release is public at machinerelations.ai/data/machine-relations-index.json; figures for the earlier release are read from the same file as committed on 2026-10-05. - Rates are cumulative over the observation window, not rates for the week alone. Week-over-week change compares two published rates and is never read as a cause.
- Only segments the release publishes, at 10 or more observed runs across 7 or more run dates, are reported. Only domains with a confidence tier (A, B or C) are ranked or listed as movers; tiers are the Index's own.
- Domains operated by the publisher of this site or its related parties (authoritytech.io, machinerelations.ai, jaxonparrott.com, christianlehman.com, presspool.ai, paralabs.ai, paralax.ai) are left out of every ranking in this report. Their Index profiles remain public.
- The method is set out at machinerelations.ai/index, and known data issues and corrections are logged in the Index methodology and update log. Corrections to an edition are made there and in the edition.
FAQ #
What is the Machine Relations Index Report? A weekly reading of the daily public release of the Machine Relations Index: the most-cited sources, the largest movers among graded domains and the leader of every published segment, each figure tied to a release id and linked to the Index page that publishes it.
Does a first place mean a source is better? No. It means answer engines cited it most often for that question shape in that category during the window. The Index measures citation, not accuracy; whether a cited source supports the answer is a separate measurement.
Why can a rate fall while a source gains citations? A rate is citations over observed runs. reddit.com was cited in 2,399 runs this week against 2,296 a week earlier, while observed runs grew to 23,940; the rate moves with both.