If you work in PR, SEO, or digital marketing, you have probably had this conversation in the last month: “Which publications actually show up in AI answers?” For years the answer was educated guesswork — a mix of anecdote, rank trackers, and press-release hype.
There is now at least one measured answer. The Machine Relations Index, released October 17, 2026, ran 26,794 AI answer runs across six engines and recorded every cited domain. Forbes came first — but the more useful findings are about how fragmented those citations are, and how much the “right” list changes with the question being asked.
Key Takeaways
- Forbes was cited in 1,041 of 26,794 AI answer runs (3.9%) across six engines in the October 17, 2026 Machine Relations Index release — first of 1,027 classified editorial publications, followed by TechRadar (758) and Yahoo News (536).
- The top share is only 3.9%, which shows how fragmented AI citations are: no single publication dominates the answer layer.
- The list changes with the question shape: TechCrunch is strong on news questions but was cited on no best-tools or top-list questions, while the New York Times is the reverse.
- The Index tracks standardized enterprise-tech buyer questions — treat the ranking as directional, not as proof of what happens for every query.
What Does the Machine Relations Index Actually Measure?
The Machine Relations Index, published by Machine Relations Research, ran 1,227 monitored prompts — from news questions to buyer shortlists — over 154 days (May 10 to October 17, 2026) and recorded every cited domain across six engines: ChatGPT, Claude, Gemini, Perplexity, Google AI Mode, and Google AI Overviews. Ranks describe confirmed editorial publications, and the release figures are read from the Index’s public data release (mri_score_v2.1+2026-10-17).
That methodology matters because it limits what the ranking can tell you. The prompts are standardized enterprise-tech buyer questions — best tools, top lists, comparisons, how to choose. A brand in healthcare, local services, or consumer retail would face a different prompt set and very likely a different leaderboard.
Which Publications Do AI Engines Cite Most?
The top ten confirmed editorial publications in the October 17, 2026 release:
| Rank | Publication | Runs cited | Engines (of 6) |
|---|---|---|---|
| 1 | Forbes | 1,041 | 6 |
| 2 | TechRadar | 758 | 6 |
| 3 | Yahoo News | 536 | 5 |
| 4 | Healthline | 455 | 6 |
| 5 | NerdWallet | 397 | 5 |
| 6 | TechTarget | 358 | 6 |
| 7 | The New York Times | 286 | 3 |
| 8 | CNBC | 243 | 6 |
| 9 | People Managing People | 227 | 6 |
| 10 | PCMag | 226 | 4 |
Source: Machine Relations Index, via authoritytech.io’s briefing on the release.
A few things stand out beyond the headline. Forbes, TechRadar, Healthline, TechTarget, CNBC, and People Managing People were each cited by all six engines — breadth matters as much as rank. Healthline ranks fourth on 455 runs concentrated in just four health and consumer categories, which shows how a vertical-specialist outlet can out-cite general news inside its territory.
An earlier Index release (October 4, per the briefing’s corroborating data) showed Forbes at 856 of 20,821 runs, first of 1,594 publications — a consistent leader, though the classification set changes between releases. And a Search Engine Land analysis of Semrush data covering more than 800 websites across 11 sectors on Google AI Mode, Perplexity, and ChatGPT search found forbes.com one of four domains in the top 50 cited URLs in every sector — alongside Reddit, Wikipedia, and YouTube, and the only traditional publisher of the four.
How Concentrated Are AI Citations, Really?
Not very — and that is arguably the headline. The confirmed editorial class accounts for 15,563 domain citations in this release, and Forbes, the leader, holds 6.7% of them. Across all source types, the release records 174,435 domain citations, of which the 1,027 confirmed editorial publications hold just 8.9%. Most cited domains have not yet been classified at all: in the neighboring October 18 release, 86% of cited domains carried no identity evidence and held 65% of cited-domain observations.
Two takeaways follow. First, no single outlet is the answer layer — a few dozen publications carry most of the editorial weight and a long tail carries the rest, so an earned-media strategy that bets everything on one “must-win” outlet is mispriced. Second, 8.9% is what editorial holds on present evidence, not a ceiling on what earned media wins — reading a site can only move domains out of the unclassified bucket.
How Much Does the Question Shape Change the List?
The most useful finding in the release is how differently one publication performs by question type. Forbes is cited on both news questions (168 runs) and buyer questions (873 runs), with 505 of those on best-tools and top-list questions alone. TechCrunch ranks #13 on 172 runs — but 160 are news questions, and it was cited on no best-tools or top-list question in this release. The New York Times is the mirror image: 284 of its 286 runs are buyer questions, on three engines.
The strategy implication is direct: TechCrunch coverage is a strategy for being in the answer when someone asks what happened in a market, not for being on the shortlist when someone asks which vendor to pick. Sort your target questions by shape — news, shortlist, evaluation — before you choose which publications to pitch, because the engines cite different publications for each.
Why Don’t Wire Services Lead This List?
Because many wire-service rankings come from studies of news citations. A July 2025 preprint on the AI Search Arena dataset analyzed 366,087 citations and identified the news outlets within them — and in that news slice, Reuters and AP lead. But buying and research questions draw on a different pool: in this release Reuters was cited in 22 runs on three engines and AP News in 11 runs on ChatGPT only, too few for an editorial rank. A wire story can still put a brand into a news answer; it will rarely put the brand on a buyer’s shortlist.
What Should You Do With This Data?
- Sort your buyers’ questions by shape. List the five questions a buyer asks before shortlisting you, and mark each as news, shortlist (best, top list, comparison), or evaluation (how to choose, is it worth it).
- Match publications to shapes, not prestige. For shortlist questions, target publications the engines cite on best-tools and top-list runs in your category — the overall top ten averages 25 categories, but your category’s leaders may be different publications.
- Check your category, not the overall list. The Index publishes a live ranking with category pages; the leaderboard inside your vertical matters more than the global one.
- Measure the answer, not the placement. Before and after a placement lands, record whether your brand appears in answers to those five questions on each engine. A placement in a publication the engines cite for a different question shape will not move them.
Keep one more caveat in mind: methodology moves rankings. A 5WPR Q1-2026 audit of AI citations ranked Forbes only #18 (1.38%) — a reminder that different prompt sets, engines, and classification rules produce different leaders. Use the Index as the best measured map available, and treat its ranking as directional.
People Also Ask
Which publication do AI search engines cite most?
Forbes, in the Machine Relations Index release of October 17, 2026: cited in 1,041 of 26,794 answer runs across six engines (3.9%), first of 1,027 classified editorial publications, with TechRadar (758) and Yahoo News (536) next.
Do ChatGPT, Perplexity, and Gemini cite the same publications?
Partly. Forbes, TechRadar, Healthline, TechTarget, CNBC, and People Managing People were each cited by all six engines the Index observes, but others are narrower — the New York Times was cited by three engines and AP News by ChatGPT alone in the same window.
Is TechCrunch good for AI visibility?
For news questions, yes: 160 of TechCrunch’s 172 cited runs were news questions. For shortlist questions, much less: it was cited on no best-tools or top-list question in the October 17, 2026 release.
Why do other lists put Reuters and AP first?
They measure news citations. On buying and research questions, Reuters was cited in 22 runs and AP News in 11 in the same release — too few for an editorial rank.
How often does this ranking change?
The Index publishes a new release daily. Ranks at the top move slowly; category leaders change more often.
What is Machine Relations?
The discipline of earning AI citations and recommendations for a brand by making it legible, retrievable, and credible inside AI-driven discovery — coined by Jaxon Parrott in 2024, and the umbrella that contains GEO, AEO, AI SEO, and AI PR.
