Features /Measure/ News & Research

Not scraped.
Tracked.

A GEO reading list built from llemmy's own tracked corpus of AI answers: a live self-citation metric, graded research findings, and a per-engine citation playbook, next to curated industry news. Not another feed of headlines with your logo on it.

Free to start · Findings graded by confidence · Sample size and window on every number
The Self-Citation Index
5.6%
of AI citations point to
the brand's own site
95% CI 5.4–5.7%
60,490 citations · 20 brands
What we're seeing in the tracked corpus
Grade A
The Gemini API returns zero engine citations
Grade B
Review and community sites out-cite a brand's own website

Most GEO content is an opinion with good SEO. This is measured.

Every finding here traces back to llemmy's own tracked corpus of AI answers, not a scraped article or a vibe. Confidence is graded, sample size is disclosed, and hypotheses stay labeled as hypotheses.

Live
The Self-Citation Index

A weekly-tracked metric: what share of AI citations point back to a brand's own site, across a panel of tracked brands.

A / B
Confidence-graded findings

Only findings confirmed across multiple runs make it out of the lab. Grade C hypotheses and anything proprietary stay internal.

Daily
Industry news, refreshed

Curated GEO and AI-search headlines, plus an AI-generated weekly briefing that turns them into next moves.

Not a news aggregator

Findings from answers we actually measured.

Every question llemmy tracks against ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews feeds a shared research corpus. When a pattern holds across multiple independent runs, with no contradicting evidence, it graduates from a hunch to a published finding, with the sample size and date window attached.

Grade A, confirmed across 3+ runs with no contradiction, e.g. the Gemini API returning zero engine citations
Grade B, confirmed across 2+ runs, e.g. review and community sites out-citing a brand's own website on reputation questions
Never published: single-run hypotheses, and anything that would hand a competitor our method
What we're seeing
AChatGPT's citations carry a detectable, OpenAI-exclusive tag on ~87–90% of cited responses
BPerplexity is the only engine that marks citations inline, in 72–100% of responses
BFan-out queries only surface on ChatGPT and Claude in the tracked corpus
Corpus is currently single-vertical: a strong directional signal from our tracked data, not a cross-industry census.
What we believe, and why

A graded point of view, plus the per-engine playbook.

Best Practices is llemmy's evolving, graded knowledge base: falsifiable claims about GEO, not received wisdom. Next to it sits the per-engine citation playbook, how ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews each source an answer, and the highest-leverage move for each one.

Earned media beats on-page tricks for earning citations, not schema and keyword work alone
Being cited is not being recommended, and the two get tracked separately
Every GEO tactic is a hypothesis until you measure a before and after
Per-engine playbook
ChatGPTBing index + entity authority
ClaudeProvable facts + provenance
PerplexityReddit, reviews, "best of" lists
Google AI OverviewsOrganic rank is the prerequisite
GeminiKnowledge Graph, not citation
The honest promise

Every research finding here carries its sample size, date window and confidence grade, and the corpus behind it is currently single-vertical, a strong directional signal, not a cross-industry census. Grade C hypotheses and anything that would reveal how we collect or score never leave the lab. As we explain in our methodology, a confidence interval fixes noise, not bias, and we say so.

Keep exploring
AI Visibility
Track your own citations, live →

Turn the research into a tracked baseline for your own brand.

Sentiment
Know the tone, not just the mention →

The other half of what AI says about a brand.

← Back to all features
From the resource library The Self-Citation Index, tracked weekly → What AI engines actually cite → How llemmy measures →

Stop guessing what AI cites. Start tracking it.

First insights in under five minutes. No credit card. See where your own brand sits against the same research.

Start free → Talk to our team