llmranks.io

resources · citation index 2026

The sources AI engines actually cite.

An attribution-first index built from published research: 30M+ citation samples, 230K-prompt benchmarks, and per-engine fingerprints across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. If you want to be cited by AI, you need to know which sources AI cites first.

01 · top-cited domains across AI search

Reddit dominates. Editorial follows. Wikipedia lost its throne on Google.

Composite ranking from Peec AI’s 30M-source study and Semrush’s 13-week / 100M-citation benchmark. Bar widths are illustrative — the underlying citation share differs meaningfully across engines (see fingerprints below).

  1. 01Reddituser-generated

    Top across every engine. ~40% of all citations in the Peec study.

  2. 02YouTubevideo

    Overtook Reddit on Bluefish's 6-month measure (16% vs 10%). Transcript machine-readability is the unlock.

  3. 03LinkedInprofessional network

    Heavy on Google AI Mode (~15%) and Perplexity B2B queries.

  4. 04Wikipediaencyclopedic

    Foundational for ChatGPT and Perplexity. Almost absent from Google AI Mode (~2%).

  5. 05Forbeseditorial

    Editorial backbone for ChatGPT and AI Mode answers.

  6. 06Mediumeditorial

    Surged in ChatGPT after the Sep 2025 reshuffle.

  7. 07G2user-generated

    Recommendation queries — software comparison answers.

  8. 08Yelpuser-generated

    Local + service categories. Disproportionately strong on AI Mode.

  9. 09NIH / PubMedencyclopedic

    Perplexity's authority anchor for medical and research queries.

  10. 10Business Insidereditorial

    Editorial source frequently cited by ChatGPT.

composite source: peec ai (mar 2026) + semrush (jul-oct 2025) + bluefish (early 2026) + 5w citation index (2026)

02 · per-engine citation fingerprint

No two engines cite the same web.

Each engine has a distinct citation fingerprint shaped by its training data, retrieval method, and platform parent. Optimizing for “AI search” in the abstract is wishful thinking — you have to know which engine you’re talking to.

ChatGPT

Encyclopedic + community + editorial. Volatile — Reddit and Wikipedia citation share dropped sharply in Sep 2025.

  1. 01Reddit
  2. 02Wikipedia
  3. 03Medium
  4. 04Forbes
  5. 05LinkedIn

Claude

Legacy journalism. Skews to long-form editorial more than any other engine.

  1. 01The New York Times
  2. 02The Atlantic
  3. 03The New Yorker
  4. 04The Economist

Gemini / AI Mode

UGC-leaning, balanced mix. Heaviest on professional + video. Wikipedia barely registers (~2%).

  1. 01LinkedIn (~15%)
  2. 02YouTube
  3. 03Reddit
  4. 04Facebook
  5. 05Yelp

Perplexity

Primary research + B2B authorities. Highest research-paper citation rate of any engine.

  1. 01Reddit
  2. 02LinkedIn
  3. 03NIH / PubMed
  4. 04Microsoft
  5. 05Google

Google AI Overviews

Google-properties biased + UGC platforms. Wikipedia almost absent. YouTube and Facebook overweighted.

  1. 01LinkedIn
  2. 02YouTube
  3. 03Reddit
  4. 04Google blog
  5. 05Yelp

03 · this changes

ChatGPT’s September 2025 reshuffle.

Mid-September 2025, Reddit and Wikipedia citation share dropped sharply on ChatGPT in less than two weeks. The same drop didn’t hit Perplexity or AI Mode — making it an isolated tuning change, not a category-wide shift. Citation share is a moving target; one-time analysis ages quickly.

Reddit

60%→~10%

down ~50pt

Wikipedia

55%→<20%

down ~35pt

Medium · Forbes · PRnewswire

low→rose into top 5

up

source: semrush (chatgpt search citation share, jul-oct 2025)

what this means for your brand

The off-site playbook isn’t a nice-to-have anymore.

Reddit is #1 across every engine. YouTube is #2 and rising — its transcripts are easier for LLMs to parse than most blog posts. LinkedIn drives Google AI Mode. NIH drives Perplexity B2B. Each engine has a different favorite, and ranking on Google for the same query no longer correlates with being cited.

If you’re only optimizing your own domain, you’re playing the wrong game. The off-site surfaces — Reddit threads, YouTube channels, G2 categories, Q&A archives, editorial mentions — are doing most of the work. That’s where the citations come from.

LLMRanks audits which surfaces feed AI in your category, then gives you the specific Reddit threads, YouTube channels, and Q&A archives to engage with. The index above is the macro view; the audit is the brand view.

Run free audit on your brandRead the State of AI Search next

sources

All citation data on this page comes from publicly-available research published between Q3 2025 and Q2 2026. Citation share shifts month-to-month as AI engines retune their retrieval and training data — re-verify any number you intend to act on. Last updated May 2026.

AI Citation Index · 2026 · LLMRanks