What AI tells collectors and players before they buy
A collector asks which marketplace is safe, whether a collection is legit and whether NFTs are still worth buying. We measured what the answers are built from, because that is where a buyer decides about you.
On 23 October 2026 we put 10 questions to Perplexity and to ChatGPT with web search, the kind a collector or a player asks: the safest NFT marketplace and the one with the lowest fees, how to check that a collection is legit, the best blockchain games, the best wallet for NFTs, where to buy on Solana, whether NFTs are still worth buying, which brands run collectible programmes worth joining, how to avoid scams, and where to sell digital art. We read which pages each answer was built from.
| What we measured | Result | What it means for you |
|---|
| Cited pages that are a ranked list or a comparison | 47% on Perplexity (47 of 99 citations), 61% on ChatGPT (17 of 28) | Collectors are pointed to lists and comparisons. Being on the lists that rank for your kind of game or collection, described correctly, is a channel of its own. |
| Citations on the own site of a project the answer recommends | 5% on Perplexity (5 of 99). 2 of 28 on ChatGPT | Projects’ own sites are almost never the source. A collection page that answers what it is, who made it and what holders get is the page that could change that. |
| OpenSea and Magic Eden, the marketplaces named most often | Each named in 7 answers across both surfaces. OpenSea’s own site cited in 1 of its 7, Magic Eden’s in none | Even the biggest marketplaces are recommended on other people’s pages. For a collection, the marketplace page and the coverage about you matter more than your own site. |
| Citations on another company’s blog or learn pages | 21% on Perplexity (21 of 99), in 9 of 10 answers. The most-cited domain is an exchange’s crypto wiki: 7 citations in 5 answers, none about its own product | An exchange’s generic wiki was the most-cited site in the set. The explainers for your category are an open seat a serious project can take. |
| Answers citing media: crypto, tech or business press | 8 of 10 on Perplexity; 10 of the 23 media citations are the outlet’s own ranked list. 5 of 10 on ChatGPT | Media matter through their lists and their reporting. Coverage that explains what the project does is quoted; a mint announcement is not. |
| ChatGPT answers given without searching the web | 2 of 10, both about scams and checking a collection. 6 of the 8 that searched added “2024” to the query | On scams and legit checks ChatGPT answered from memory, and most answers that searched looked for 2024 lists. A stale record is what a cautious buyer is told. |
| Sources behind the three trust questions on Perplexity | Is a collection legit, are NFTs worth buying, how to avoid scams: no product named, and no marketplace or NFT project among the 30 sources. Security vendors, exchanges’ learn pages and an audit firm’s blog instead | On every trust question, no marketplace and no project was a source. Security vendors and exchanges answer “is it legit” for you unless your own page does. |
Asked for the safest NFT marketplace, ChatGPT searched for 2024 lists and recommended the Kraken NFT Marketplace, which Kraken closed in February 2025. Perplexity, asked whether NFTs are still worth buying, opened with “Usually no—not as an investment”. That is the market a collection now launches into: a buyer who has been told to doubt, reading answers built on old lists.
Method: Perplexity (sonar) and ChatGPT (gpt-4.1-mini with web search) through the API, 23 October 2026, one answer per question, US. 22 questions in two sets: 12 a developer or a business asks before choosing blockchain infrastructure and 10 a collector or player asks. No question names a brand. A source is a distinct page the answer cites; ChatGPT’s logo links above each product it lists are not counted. Domains were classified by hand and five answers checked against the raw responses. A small sample, read as a direction: 10 to 12 answers a set on each surface, and Perplexity returns at most ten sources an answer, so its ten is a ceiling, not a finding.