Asked about the money problems stablecoins genuinely solve, AI engines named a crypto product a reader could actually use 6 times in 360 answers. They raised crypto as a concept in 127. The category is not hidden — the products are.
That is the finding edition one of the Kunzum index was built to test, and it is the entire commercial case for stablecoin AEO. The demand exists in the answer. The names in it belong to Wise, PayPal and Payoneer. The full report publishes all 480 responses.
In short
- Across 360 unprompted answers about cross-border money, engines raised crypto 127 times and named an actionable crypto product 6 times.
- Wise was named in 246 of 480 responses. The highest-placed crypto-native name was Circle at 14.
- Say the word stablecoin in the question and every engine surfaces crypto. The gap is retrieval and naming, not sentiment.
- Crypto recommendations carry a caution 2.8 times as often as traditional ones — and on a hand classification, 82% of those cautions were accurate.
- This is the one crypto category with a published Kunzum corpus measuring it directly.
What the engines actually name
Ten ordinary money problems — a freelancer paid from abroad, savings eroding to inflation, a business paying international contractors — described in plain language to four engines, three times each. Kunzum scored the products named out of the answers rather than against a shortlist, so this is an open-vocabulary count.
| Product | Responses naming it | Type |
|---|---|---|
| Wise | 246 | Traditional |
| PayPal | 156 | Traditional |
| Payoneer | 139 | Traditional |
| Stripe | 127 | Traditional |
| Circle | 14 | Crypto-native |
| BitPay | 13 | Crypto-native |
| Coinbase | 13 | Crypto-native |
| KAST | 13 | Crypto-native |
The shape is the finding. The traditional incumbents are named in the hundreds; the crypto-native names sit in the low teens, and most of those appear only once the question already contains the word crypto. Circle is an issuer rather than something a reader signs up for, which pushes the first genuinely actionable crypto product lower still.
Why this is a retrieval problem
The obvious explanation is that engines dislike crypto. The corpus does not support it. Every engine surfaced crypto 127 times unprompted and 100% of the time once asked directly, and the hand classification of warnings found crypto cautions were more accurate than warnings about traditional products. These engines are not refusing. They are reaching for names they have sources for.
Which makes it workable. In edition one, ChatGPT ran a web search on 54% of its answers and Claude on 65%; reddit.com was the most-cited domain at 293 citations and editorial press accounted for 26 of 3,987. The places these answers came from are documented, and they are not the places a stablecoin issuer's PR budget usually goes. The full source basis is published with its data.
The work itself is the method: find the questions buyers really ask, write the passages that answer them in the buyer's vocabulary, place them where the citation data says these engines read, then re-ask monthly.
What this page does not claim
Every prompt was asked from a US location, so this measures a US-located asker rather than a global one. 54% of ChatGPT's answers ran a web search, meaning the rest came from training data nothing published could reach. And this is one category in one month. The prompts, the raw responses and the scoring scripts are at the dataset page so anyone can rescore it and disagree.
Related categories with their own measurement: crypto cards and neobanks, measured across nine markets, and perpetual DEXes, where the category had no written definition at all.
Sources
- Google Search Central, “AI features and your website”. Google. Checked 2026-09-12.
- OpenAI’s crawler documentation. OpenAI. Checked 2026-09-12.
- Aggarwal and colleagues, “GEO: Generative Engine Optimization” (2023). arXiv:2311.09735. Checked 2026-09-12.
- Lewis and colleagues, “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks” (2020). arXiv:2005.11401. Checked 2026-09-12.
Published 2026-09-12. Written by Narender Charan, who runs Kunzum.