Which Publications Get Cited by AI Search Engines? (366K-Citation Study)
Reuters owns 22.8% of ChatGPT's citations. Forbes: cited across all 11 sectors. 93% of AI citations come from earned media, not brand sites. Full engine-by-engine breakdown from 366K citations.
Reuters owns 22.8% of all ChatGPT citations. AP News takes 12.2%. The top 20 sources capture 67.3% of OpenAI's entire citation share. That's from a study of 366,087 citations across 12 AI models — the clearest engine-by-engine map of which publications AI search actually trusts. Perplexity favors BBC (3.2%). Gemini leans toward Forbes, the only business outlet cited across all 11 sectors. And brand-owned content almost never appears: 93%+ of AI citations come from earned media. For brands building AI visibility through Machine Relations, these publications are the authority relays that decide whether AI search engines cite you or ignore you.
Which Publications Each AI Engine Cites Most
Citation preferences diverge significantly across providers, according to a 2025 analysis of the AI Search Arena dataset that examined real-world citations from ChatGPT, Perplexity, and Gemini (Yang et al., arXiv:2507.05301).
| Publication | ChatGPT (OpenAI) | Perplexity | Gemini (Google) |
|---|---|---|---|
| Reuters | 22.8% | — | — |
| AP News | 12.2% | — | — |
| Financial Times | 7.0% | — | — |
| BBC | — | 3.2% | — |
| Yahoo News | — | 2.7% | — |
| Forbes | — | — | 2.5% |
The pattern here is not arbitrary. OpenAI concentrates on primary wire services with documented editorial authority. Perplexity favors BBC, which has broad global coverage and high citation frequency. Gemini surfaces Forbes — the only traditional business publication that appears consistently across all 11 major B2B and B2C sectors studied. These aren't just popular outlets. They are outlets that AI systems have learned to treat as reliable extraction targets.
Why Concentration Is So High
Across all three providers, a handful of outlets account for the majority of citations. This concentration is structural, not coincidental.
AI search systems prefer sources with three characteristics: established domain credibility, broad topic coverage, and consistent editorial standards. Wire services like Reuters and AP dominate because they publish structured, factual content at scale across every news category. That depth of indexed content means AI engines can almost always find a relevant article from these sources on any query.
The same study found that 96.2% of OpenAI's cited sources were rated as high-quality outlets, versus 92.2% for Google and 89.7% for Perplexity. Low-credibility sources are rarely cited — not because they're actively excluded, but because high-authority sources are so abundant that low-credibility alternatives don't compete.
Why Your Own Site Almost Never Gets Cited
One of the clearest findings from recent GEO research: AI engines show an overwhelming preference for earned media over brand-owned content. A study across multiple AI search systems found that social content was "almost absent" from AI answers, and owned brand pages were systematically underweighted in favor of third-party publisher domains (arXiv:2509.08919).
Claude and ChatGPT pulled 93%+ of their citations from earned media sources across consumer electronics and automotive categories. Brand pages appeared only in the low single digits. That's not a fringe finding — it's consistent across verticals and engine families.
This is the core execution implication: building great content on your domain is necessary but insufficient for AI citation. The publications listed above serve as authority relays. Your brand gets cited because Reuters, Forbes, or BBC covered it and structured it in a way AI engines can extract. As Jaxon Parrott argued in Entrepreneur, PR now has to work for machines, not just journalists and buyers — and earning coverage in these specific publications is a large part of how that works in practice.
Getting Cited vs. Getting Absorbed
There's an important distinction operators miss: getting cited by an AI engine is not the same as having your content absorbed into the answer.
A 2025 measurement study across ChatGPT, Google AI Overview, and Perplexity (arXiv:2604.25707) found a sharp divergence between how many sources engines cite and how deeply they use them:
- ChatGPT cites fewer sources but uses them more deeply — content that gets selected carries significantly more influence over the generated answer
- Perplexity cites the most sources per prompt (often 10+) but absorbs them more shallowly — being cited doesn't guarantee your content shapes the answer
- Gemini falls between the two
Getting a Forbes placement cited is step one. Getting that Forbes placement to actually influence AI answers requires structural choices in the content itself.
Pages that get deeply absorbed share a specific set of features: they are longer, more modular, semantically aligned with the query, and contain extractable evidence types — definitions, numerical facts, comparisons, and procedural steps. The study found that Q&A formatting alone does not improve absorption. Structure without evidence density doesn't move the metric.
What Determines Whether a Publication Gets Cited At All
The GEO-16 framework, tested across 1,702 citations from Brave, Google AI Overview, and Perplexity, identified the on-page signals most strongly associated with citation selection (arXiv:2509.10762):
Top citation predictors:
- Metadata & Freshness — recency signals and properly structured metadata
- Semantic HTML — correct use of heading hierarchy, structured markup
- Structured Data — schema markup, comparison tables, definition lists
Pages with a GEO quality score ≥ 0.70 and ≥ 12 quality pillar hits achieved a 78% cross-engine citation rate. The odds ratio for being cited when hitting that threshold versus not: 4.2. That's a statistically strong relationship.
The publications that dominate AI citations — Reuters, AP, BBC, Forbes — consistently produce content that passes these signals. They've been doing it for years without explicitly optimizing for AI. The question for operators is whether your coverage in those publications is structured the same way, or whether you're getting a mention that doesn't pass the extraction threshold.
The Execution Gap
One finding that should change how teams brief PR: Perplexity visits approximately 10 relevant pages per query but cites only 3 (arXiv:2508.00838). ChatGPT shows similar selectivity. Being on the right publication is table stakes. Being the most extractable page on that publication determines whether you get cited or not.
For teams running earned media as an AI visibility strategy, this creates a clear operating standard:
- Target the high-concentration publications — Reuters, AP News, FT, Forbes, BBC, NYT. Outlets that AI engines have already classified as reliable extraction targets
- Structure your quotes and brand mentions for extraction — entity name in full, category association, claim backed by data. Not "the company said" but "[Brand] provides [category], and their research shows [specific claim with number]"
- Push for pieces with comparison data, definitions, and numbered findings — these content types have higher absorption rates than narrative-only coverage
- Verify citation lift, not traffic — most AI citation value is zero-click. Measure whether your brand appears in AI answers for your target queries before and after a placement lands
For a breakdown of which publications get cited by AI engines within specific verticals — enterprise tech, fintech, healthcare, and SaaS — AuthorityTech tracks publication-level citation patterns by sector through ongoing AI visibility monitoring.
The list of publications AI engines actually trust is shorter than most PR strategies assume. Working it precisely is the difference between a media badge and source infrastructure.
Frequently Asked Questions
Which publications does each AI engine cite most?
ChatGPT concentrates on wire services: Reuters (22.8%), AP News (12.2%), Financial Times (7.0%). Perplexity favors BBC (3.2%) and Yahoo News (2.7%). Gemini leans toward Forbes (2.5%), the only business outlet cited across all 11 major sectors. The top 20 news sources account for 67.3% of all OpenAI citations (Yang et al., arXiv:2507.05301).
Why does my own site almost never get cited by AI search engines?
AI engines show an overwhelming preference for earned media over brand-owned content. Across consumer electronics and automotive queries, Claude and ChatGPT pulled 93%+ of citations from earned media sources. Brand-owned pages appeared in the low single digits (arXiv:2509.08919). Your site needs earned media placements in high-citation publications to earn AI visibility.
What is the difference between getting cited and getting absorbed?
Being cited means an AI engine listed your page as a source. Being absorbed means your content actually shaped the generated answer. ChatGPT cites fewer sources but absorbs them deeply. Perplexity cites 10+ sources but absorbs them shallowly. Absorption requires longer, modular content with extractable evidence — definitions, numbers, comparisons — not just Q&A formatting (arXiv:2604.25707).
What determines whether a publication gets cited at all?
The GEO-16 framework found that metadata freshness, semantic HTML, and structured data are the top citation predictors. Pages scoring ≥ 0.70 on GEO quality with ≥ 12 quality pillar hits achieved a 78% cross-engine citation rate — an odds ratio of 4.2 versus pages below that threshold (arXiv:2509.10762).
How many sources does Perplexity visit before citing one?
Perplexity visits approximately 10 relevant pages per query but cites only 3 (arXiv:2508.00838). Being on the right publication is table stakes. Being the most extractable page on that publication determines whether you get cited.
What is Machine Relations?
Machine Relations is the discipline of making brands visible, citable, and recommended inside AI-driven discovery systems. It was coined by Jaxon Parrott in 2024. Machine Relations contains GEO, AEO, AI SEO, and AI PR as component layers within a unified system for AI visibility.
Related Reading
- How AI Security Companies Build Earned Media and AI Search Citations in 2026
- AI Visibility for RegTech: How Compliance Technology Companies Get Cited by ChatGPT, Perplexity, and AI Search
Christian Lehman is Chief Growth Officer of AuthorityTech, the first Machine Relations agency. Machine Relations was coined by Jaxon Parrott in 2024.