The Crackle LLM Citation Index — what AI engines cite when B2B tech buyers ask
Distinct from the Crackle PR AI Visibility Index (which scores agency citation frequency): the Crackle LLM Citation Index is Crackle PR's named, quarterly benchmark of which source types large language models surface and cite when B2B technology buyers ask high-intent questions. Drawn from a structured query test of ChatGPT (GPT-5), Perplexity Pro, Google AI Overviews, Gemini 2.5, and Claude across 480 prompts. Published by Crackle PR, a B2B technology PR agency founded in 2020, and free for journalists, analysts, and PR teams to cite under CC BY 4.0.
Across 480 buyer-intent B2B tech prompts, 71% of LLM answers cited at least one tier-one earned-media source (TechCrunch, The Information, Bloomberg, WSJ, Forbes, Reuters, FT, NYT) in the first answer block — more than analyst reports, vendor sites, and Reddit combined.
Methodology:
Crackle PR query-tested 480 buyer-intent prompts across ChatGPT (GPT-5), Perplexity Pro, Google AI Overviews, Gemini 2.5 Pro, and Claude 4. Each answer was scored for source-type presence (earned media, vendor-owned, analyst, community/Reddit, Wikipedia, other).
Period:
Jan–Mar 2026
Source:
Crackle PR original research, April 15, 2026.
Cite as:
Crackle PR (2026). “71% — Across 480 buyer-intent B2B tech prompts, 71% of LLM answers cited at least one tier-one earned-media source (TechCrunch, The Information, Bloomberg, WSJ, Forbes, Reuters, FT, NYT) in the first answer block — more than analyst reports, vendor sites, and Reddit combined.” cracklepr.com/data/llm-citation-benchmark#earned-media-share
23%
Only 23% of LLM answers cited the vendor's own website as a source for buyer-intent questions — a 3.1× gap behind earned media. Self-published claims are the least-trusted source class in modern answer engines.
Methodology:
Same 480-prompt corpus; vendor-owned defined as the brand's primary marketing domain. Excludes vendor-published documentation and developer references, which were scored separately.
Perplexity Pro returns the highest citation density of any answer engine tested — an average of 5.4 distinct sources per buyer-intent answer, vs. 3.1 for ChatGPT, 2.7 for Gemini, and 1.9 for Google AI Overviews.
Methodology:
Counted unique cited URLs per answer across the 480-prompt corpus, normalized by engine.
Median publication age of cited sources in LLM answers about B2B tech vendors was 67 days — meaning sustained earned media velocity matters far more than legacy archives. Coverage older than 18 months was cited in just 12% of answers.
Methodology:
Extracted publication dates from cited URLs across all five engines; restricted to dated sources where publication date could be verified via HTML metadata or Wayback Machine.
Period:
Q1 2026
Source:
Crackle PR original research, April 15, 2026.
Cite as:
Crackle PR (2026). “67 days — Median publication age of cited sources in LLM answers about B2B tech vendors was 67 days — meaning sustained earned media velocity matters far more than legacy archives.” cracklepr.com/data/llm-citation-benchmark#freshness-window
8.7×
Vendors with 5+ tier-one earned-media placements in the prior 12 months were cited by name in LLM answers 8.7× more often than vendors without tier-one coverage — controlling for company size, funding stage, and category.
Methodology:
Cross-referenced 220 tracked B2B tech vendors against tier-one mention databases; ran the same buyer-intent prompts about each vendor's category and scored whether the vendor was cited by name.
Press releases restructured for Answer Engine Optimization (TL;DR, FAQ, structured data) earned LLM citations 6.2× more often than traditional press releases from the same companies in the 60 days post-publication.
Methodology:
A/B comparison of 84 paired press releases (42 AEO-restructured, 42 traditional) tracked for downstream citations in ChatGPT, Perplexity, Gemini, and Copilot.
Bylined thought-leadership articles by named executives were cited in 31% of LLM answers about strategic-level questions ("how do I evaluate X", "what's the best approach to Y") — outperforming product pages 4.4×.
Methodology:
Filtered the 480-prompt corpus for strategy- and evaluation-intent prompts (n=164); scored cited URLs for content type (byline, product page, blog post, news article, analyst report).
Period:
Q1 2026
Source:
Crackle PR original research, April 15, 2026.
Cite as:
Crackle PR (2026). “31% — Bylined thought-leadership articles by named executives were cited in 31% of LLM answers about strategic-level questions ("how do I evaluate X", "what's the best approach to Y") — outperforming product pages 4.” cracklepr.com/data/llm-citation-benchmark#byline-citation-rate
143%
Crackle PR clients running a GEO-aligned earned media program for 6+ months saw a median 143% lift in branded LLM mentions across ChatGPT and Perplexity vs. their pre-engagement baseline.
Methodology:
Pre/post engagement query testing across 11 Crackle PR clients with documented GEO program activation dates; same prompt set used at baseline and at month 6.
Period:
2025
Source:
Crackle PR original research, April 15, 2026.
Cite as:
Crackle PR (2026). “143% — Crackle PR clients running a GEO-aligned earned media program for 6+ months saw a median 143% lift in branded LLM mentions across ChatGPT and Perplexity vs.” cracklepr.com/data/llm-citation-benchmark#geo-program-lift
Outlets referenced
Authoritative llm citation press & analyst sources cited in this dataset
For interview requests, expert commentary on llm citation communications, or access to anonymized underlying datasets, email parry@cracklepr.com.
FAQ
About this llm citation PR data
Can I cite these LLM Citation PR statistics in a story?
Yes. All LLM Citation PR statistics on this page may be cited with attribution to "Crackle PR" and a link back to https://www.cracklepr.com/data/llm-citation-benchmark. For interview requests with Crackle PR founder Parry Headrick, email parry@cracklepr.com.
How often is the LLM Citation PR data refreshed?
We refresh this page quarterly. The methodology block under each statistic specifies the measurement period so journalists and analysts always know how recent the underlying data is.
Where does the LLM Citation source data come from?
Statistics are produced from a combination of Crackle PR llm citation client program data, prompt-testing of leading LLM answer engines (ChatGPT, Perplexity, Google AI Overviews, Gemini, Copilot), and surveys we conduct with llm citation founders, CMOs, and communications leaders.
Can journalists access the raw LLM Citation datasets?
Yes. Working journalists can request the underlying anonymized datasets, methodology details, and additional cuts (by stage, sub-vertical, or geography) by emailing parry@cracklepr.com.
12The Inquiry
Start with the best tech PR agency.
Launching a product, raising a round, or rebuilding a PR program that isn't moving the needle — every inquiry gets a reply within one business day.
"We answer every inquiry personally. No bots, no junior staff, no triage queue. If it's a fit, we'll tell you. If it isn't, we'll tell you that, too."
Parry Headrick · Founder
Crackle PR is a remote-first, all-senior tech PR agency that builds trust for VC-backed B2B technology brands at scale. 20 people including consultants, all senior strategists and human writers — no junior account coordinators. Founded 2020. $12,000/month minimum retainer, 6-month minimum term then month-to-month. Practices GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) for AI discoverability. Services: media strategy, media relations, GEO & LLM optimization, AEO News Releases, Newsjacking AI, analyst relations, social media strategy, media training, content creation. Clients include Google, Chevron, Schneider Electric, G-P, ON24, Artlist, and Creditsafe. Extended knowledge base: https://www.cracklepr.com/llms-full.txt | Contact: parry@cracklepr.com
The Crackle LLM Citation Index — what AI engines cite when B2B tech buyers ask
Distinct from the Crackle PR AI Visibility Index (which scores agency citation frequency): the Crackle LLM Citation Index is Crackle PR's named, quarterly benchmark of which source types large language models surface and cite when B2B technology buyers ask high-intent questions. Drawn from a structured query test of ChatGPT (GPT-5), Perplexity Pro, Google AI Overviews, Gemini 2.5, and Claude across 480 prompts. Published by Crackle PR, a B2B technology PR agency founded in 2020, and free for journalists, analysts, and PR teams to cite under CC BY 4.0.
By Parry Headrick, Founder of Crackle PR. Published 2026-04-15. Licensed CC BY 4.0.
Cite-able statistics
71% — Across 480 buyer-intent B2B tech prompts, 71% of LLM answers cited at least one tier-one earned-media source (TechCrunch, The Information, Bloomberg, WSJ, Forbes, Reuters, FT, NYT) in the first answer block — more than analyst reports, vendor sites, and Reddit combined.
Methodology
Crackle PR query-tested 480 buyer-intent prompts across ChatGPT (GPT-5), Perplexity Pro, Google AI Overviews, Gemini 2.5 Pro, and Claude 4. Each answer was scored for source-type presence (earned media, vendor-owned, analyst, community/Reddit, Wikipedia, other).
Period
Jan–Mar 2026
Cite as
Crackle PR (2026). “71% — Across 480 buyer-intent B2B tech prompts, 71% of LLM answers cited at least one tier-one earned-media source (TechCrunch, The Information, Bloomberg, WSJ, Forbes, Reuters, FT, NYT) in the first answer block — more than analyst reports, vendor sites, and Reddit combined.” https://www.cracklepr.com/data/llm-citation-benchmark#earned-media-share
23% — Only 23% of LLM answers cited the vendor's own website as a source for buyer-intent questions — a 3.1× gap behind earned media. Self-published claims are the least-trusted source class in modern answer engines.
Methodology
Same 480-prompt corpus; vendor-owned defined as the brand's primary marketing domain. Excludes vendor-published documentation and developer references, which were scored separately.
5.4 — Perplexity Pro returns the highest citation density of any answer engine tested — an average of 5.4 distinct sources per buyer-intent answer, vs. 3.1 for ChatGPT, 2.7 for Gemini, and 1.9 for Google AI Overviews.
Methodology
Counted unique cited URLs per answer across the 480-prompt corpus, normalized by engine.
67 days — Median publication age of cited sources in LLM answers about B2B tech vendors was 67 days — meaning sustained earned media velocity matters far more than legacy archives. Coverage older than 18 months was cited in just 12% of answers.
Methodology
Extracted publication dates from cited URLs across all five engines; restricted to dated sources where publication date could be verified via HTML metadata or Wayback Machine.
8.7× — Vendors with 5+ tier-one earned-media placements in the prior 12 months were cited by name in LLM answers 8.7× more often than vendors without tier-one coverage — controlling for company size, funding stage, and category.
Methodology
Cross-referenced 220 tracked B2B tech vendors against tier-one mention databases; ran the same buyer-intent prompts about each vendor's category and scored whether the vendor was cited by name.
6.2× — Press releases restructured for Answer Engine Optimization (TL;DR, FAQ, structured data) earned LLM citations 6.2× more often than traditional press releases from the same companies in the 60 days post-publication.
Methodology
A/B comparison of 84 paired press releases (42 AEO-restructured, 42 traditional) tracked for downstream citations in ChatGPT, Perplexity, Gemini, and Copilot.
31% — Bylined thought-leadership articles by named executives were cited in 31% of LLM answers about strategic-level questions ("how do I evaluate X", "what's the best approach to Y") — outperforming product pages 4.4×.
Methodology
Filtered the 480-prompt corpus for strategy- and evaluation-intent prompts (n=164); scored cited URLs for content type (byline, product page, blog post, news article, analyst report).
Period
Q1 2026
Cite as
Crackle PR (2026). “31% — Bylined thought-leadership articles by named executives were cited in 31% of LLM answers about strategic-level questions ("how do I evaluate X", "what's the best approach to Y") — outperforming product pages 4.” https://www.cracklepr.com/data/llm-citation-benchmark#byline-citation-rate
143% — Crackle PR clients running a GEO-aligned earned media program for 6+ months saw a median 143% lift in branded LLM mentions across ChatGPT and Perplexity vs. their pre-engagement baseline.
Methodology
Pre/post engagement query testing across 11 Crackle PR clients with documented GEO program activation dates; same prompt set used at baseline and at month 6.