How we score 500+ crypto outlets and your AI citation footprint. Hover any signal to see exactly what it measures, how we collect it, and why it's weighted the way it is.
Each of the 500+ outlets is scored 0–100 from five live signals. Hover a signal to inspect it.
Measured across 20 buyer prompts × 5 engines, 100 observation points per audit, re-scored weekly. Every weight below is the weight in production.
How often you're named when AI answers category prompts. The dominant signal, deliberately: a category leader AI recommends in nearly every answer has to score like one.
How often your own domain is cited as a source URL, not just mentioned in passing. Also surfaced on its own as the Citation Gap.
A blend, renormalised over whatever is measurable: source corroboration 0.35, entity presence 0.35, domain rating 0.18, technical readiness 0.12.
Entity coherence is no longer a standalone signal. As of 11 July 2026 it sits inside authority, alongside three other inputs. Weights renormalise over the components actually present, so an unwired signal neither helps nor penalises.
When AI names you, do credible outlets back it up? Measured per engine from the answer's own source list.
Cross-platform footprint: Wikipedia, CoinGecko, Reddit, YouTube, X. The largest domain-level authority signal in crypto.
Classic backlink authority. Roughly 3.6% of the composite, and that is intentional. Backlinks are the weakest wired input here.
AI-crawlable, schema present, llms.txt, speed. Can the engine reach and parse you at all.
Engines are weighted to 2026 AI search share. Query types are weighted by how much they reveal about real buyer visibility.
ChatGPT 0.24 · Gemini 0.20 · Google AI Overviews 0.20 · Perplexity 0.14 · Claude 0.11. Grok carries a published weight of 0.06 but is dormant until its API key ships, so today's audits run on five engines and 100 observation points. Weights normalise, so an engine that is off does not distort the score.
Category "best/top X" 1.20× · comparison 1.00× · brand due-diligence 0.30×. You always win questions about yourself, so those are discounted, not excluded.
An engine returning fewer than 5 usable answers is excluded from the composite. Below that the rate is noise, not signal.
The same prompt rarely returns the same brand list twice. We probe N times per query and merge on the majority, then surface per-engine stability.
Weights are only credible if they survive contact with data. Across 39,948 citation events from 72 audited crypto brands, spanning 4,395 distinct source domains, with the pages behind 1,357 of those citations fetched and tagged by hand (192 pages — 3.4% of corpus volume, and a sample of the most-cited pages rather than a random draw):
The remaining 92.22% of citations pointed at domains the brand does not control. Own-site publishing is the smallest lever, which is why citation rate carries 0.20 and not more.
Per brand, being the subject of a page averages 6.22 citations [5.51, 7.05] against 3.78 [2.59, 5.45] for being ranked in someone else’s list. Non-overlapping intervals. This reverses what we published before 25 August 2026.
Listicles are 33% of citation volume across the whole corpus, the largest single archetype. They took 48% among the most-cited pages we hand-read, which is the sample, not the corpus. Both are real; they answer different questions.
The engines pulled from 4,395 distinct source domains, so there is no compact shortlist to buy your way onto. Concentration figures (top-10 share, share cited exactly once) are being recomputed against the wider corpus and are withheld until they are.
Corpus figures are share of total citation volume, not rate of appearance per prompt. The two are different measurements and we keep them named separately.
See the AI score deep dive →Before any citation work lands, your site has to be readable by AI engines. We score the technical foundation, then place your visibility against the players AI actually cites.
Seven checks: AI-crawler access in robots.txt (GPTBot, ClaudeBot, PerplexityBot, Google-Extended), llms.txt, sitemap, Schema.org, page metadata, HTTPS, and Core Web Vitals: scored 0–100 with a prioritized fix script.
Your AI-visibility score against your sector's most-cited brands on the same category prompts, plus a live "score my rival" check that runs any domain through the same method.
Where you sit versus the median citeOS score for your vertical: computed only from anonymized peer audits, shown once 3+ peers exist.
Every data point on this site comes from a measured audit. Scoring weights are published policy choices, not measured values. Audit engine live at citeos.io.
The scoring weights, the aggregate findings and a log of every figure we have published and later retracted are open data under CC BY 4.0 at github.com/emunchained/citeos.
One submission, two reports, your PR citation score and AI citation score, within 48 hours.
Get my free audit →