AI Sources for Crypto Citations: No Room for CoinGecko and CoinMarketCap

Ask ChatGPT what the best DeFi protocol is at the moment or whether a stablecoin is really safe, and chances are the model will never mention CoinGecko, CoinMarketCap, or DeFiLlama. This is the most surprising result of a study conducted by ICODA in collaboration with Semrush, which examined AI sources for crypto citations by analyzing over 100 controlled queries on ChatGPT and Perplexity. The data that emerges overturns many of the certainties on which crypto marketing is based today: data aggregators, considered for years the benchmark of the industry, received not a single citation.

Summary

• Key points
• ICODA and Semrush's research on AI citations in crypto queries
• Zero citations for CoinGecko, CoinMarketCap, and DeFiLlama
• Who really cites AI: the four winning sources
  • SEO sites and roundups outperform data tables
  • Financial media, industry press, and regulatory analysis
  • Companies cite themselves: the cases of Circle and Tether
  • Community platforms: Reddit remains marginal
• How AI decides whom to cite: intent, structure, and overlap between engines
• FAQ
  • Why weren't CoinGecko and CoinMarketCap cited by AI in crypto queries?
  • What types of sources do models like ChatGPT and Perplexity cite for crypto queries?
  • How consistent are the citations between ChatGPT and Perplexity?
  • What drives AI models to choose one source over another?

Key Points

• ICODA and Semrush tested over 100 controlled crypto queries on ChatGPT and Perplexity in August 2026.
• CoinGecko, CoinMarketCap, and DeFiLlama were not cited in any category of queries on either engine.
• Four types of sources were consistently cited: specialized SEO sites, financial media and industry press, the companies themselves, community platforms.
• Sites like Coin Bureau outperformed data tables on "advice" or recommendation-type queries.
• Circle and Tether were directly cited in queries about the safety of stablecoins due to their verified reserve reports.

ICODA and Semrush's research on AI citations in crypto queries

The study aimed to understand which sources AI models really choose when responding to questions about the crypto world. ICODA, working together with Semrush and the Semrush One platform, executed over 100 cycles of queries on ChatGPT and Perplexity in clean sessions, without history or customizations that could influence the results. The questions were grouped into four categories: recommendation-type queries (e.g., "what is the best DeFi protocol"), breaking news, security and legitimacy (such as the solidity of a stablecoin), and queries explicitly related to community platforms like Reddit. Each category was replicated on both engines, with the complete collection of responses and generated citations. Nine representative runs were published as examples in the full report, but it is the entire dataset that outlines the pattern that the research defines as structural. The queries were executed in August 2026.

Zero citations for CoinGecko, CoinMarketCap, and DeFiLlama

The central data is clear: in none of the four categories and on none of the two tested engines, CoinGecko, CoinMarketCap, and DeFiLlama appeared. Not a single mention, not a secondary reference, nothing. For an industry that has built its narrative around the authority of these aggregators as primary sources of price and on-chain data, the result is far from obvious.

Why it matters: it means that the authority perceived by industry operators for years does not automatically translate into visibility in AI search engines. Companies that have invested in traditional SEO based on their reputation as data sources risk remaining invisible precisely in the channel that more and more users are using to inform themselves about Bitcoin and cryptocurrencies in general.

Who Really Cites AI: The Four Winning Sources

Instead of aggregators, four types of sources have emerged consistently across all analyzed queries.

SEO Sites and Roundups Outperform Data Tables

Platforms specifically built to answer advisory-type questions, such as Coin Bureau, Strategy Arena, and Datawallet, have won the recommendation queries without rivals. Coin Bureau was cited independently by both engines on the same question, the only case of spontaneous overlap between ChatGPT and Perplexity in the entire dataset. The explanation according to researchers does not concern the authority of the site, but the structural correspondence: a page built to answer "what are the best DeFi protocols in 2026" beats a price feed every time that is the question asked.

Financial Media, Industry Press, and Regulatory Analysis

On queries related to security, names like Investopedia, Reuters, Yahoo Finance, and Business Insider emerged. When the question required regulatory depth, the engines turned to a completely different set of sources: The Block, paulhastings, and Investors.com, with no overlap with the first group. paulhastings is a law firm that publishes legal analyses: when the query required regulatory depth, the engine chose a source specifically built to provide it.

Companies Cite Themselves: The Case of Circle and Tether

On a query related to the security of stablecoins, Circle and Tether were cited directly as primary sources, not summarized by third parties. The specific condition that allowed this result is not the size of the company nor its notoriety, but the availability of a consolidated archive of disclosures on reserves subject to periodic audits. It is the verifiable data, not the brand, that makes the difference in citations.

Community Platform: Reddit Remains Marginal

Reddit emerged only in one query that explicitly mentioned it, and even then it appeared alongside a blog that had already summarized the same sentiment secondarily. The reach of communities is therefore second-hand by definition, useful as confirmation but rarely as a primary source for AI.

How AI Decides Who to Cite: Intent, Structure, and Overlap Between Engines

Every citation observed in the dataset followed the same logical sequence: query intent, source structure, relevance, and only at the end the citation. The engines do not seek the most authoritative site in the category, but the source whose structure best matches what the user is actually asking for. This is an optimization problem different from what much of crypto marketing is facing today.

The overlap between ChatGPT and Perplexity is not a fixed value, but changes radically based on the type of question. For roundup-type queries, the overlap was high, with both engines independently citing Coin Bureau. For breaking news, it dropped to zero, with the same event bringing up completely different outlets on each engine. For the query regarding the security of stablecoins, the overlap was again zero, but for a different reason: ChatGPT turned to primary sources, while Perplexity referred to comparison blogs.

Reinforcing this pattern is the AI Visibility Index from Semrush, which, measuring 126 million prompts, identified the same structural dynamics on an industrial scale: the overlap between mentioned brands and the sources actually cited can drop to as low as 30% on a single platform like Gemini. Aggregating citation data across different query types into a single average score, according to researchers, obscures exactly the structural differences that determine what gets cited and what does not.

Why it matters, looking ahead: for those operating in the crypto sector, the message is that optimizing for brand authority is no longer enough. Companies and platforms that want to appear in AI-generated responses will need to build content whose structure directly addresses the intent of the question, whether it’s investment advice, a security check, or a regulatory update.

FAQ

Why were CoinGecko and CoinMarketCap not cited by AI in crypto queries?

AI citation depends on the intent of the query and the structural match of the source: these data aggregators, despite their authority, did not structurally align with the needs of the posed questions.

What types of sources do models like ChatGPT and Perplexity cite for crypto queries?

SEO roundup sites built to answer recommendation questions, financial media for security queries and breaking news, official company communications for questions about stablecoins, and community platforms only selectively.

How consistent are the citations between ChatGPT and Perplexity?

The overlap varies based on the type of query: it is high for roundup content but drops to zero for breaking news and queries related to the security of stablecoins.

What drives AI models to choose one source over another?

The sequence followed is query intent, then source content structure, then relevance: the model does not simply cite the most authoritative site in the category, but the one whose structure best answers the specific question.

Content created with the assistance of artificial intelligence and human editorial review.