AI Search Visibility for Web3 Companies: What Our Audit of 75 Medium-Cap Crypto Projects Found
Joshua Etheridge
Only 2 of 75 medium-cap web3 and crypto companies reached the Strong AIO presence band when audited against roughly 50 real AI prompts on 30 June 2026.
That single figure tells the core story. Every company in our sample was named at least once, and the average company was named in about 22 of 50 prompts. But appearing occasionally in an AI answer is not the same as being the name that ChatGPT or Google Gemini reaches for by default. For most of the 75 projects we audited, AI search is already live and already producing winners, and the gap between those winners and everyone else is stark.
What we measured and how
On 30 June 2026, AIO ran its live audit pipeline against 75 real web3 and crypto companies. Selection was systematic: we used the CoinGecko market-cap ranking, took the medium band (export const BLOG_CONTENT: Record
Each company faced the same battery of roughly 50 real prompts put to consumer ChatGPT and consumer Google Gemini, with query location set to Global. The questions were worldwide category questions, for example "best decentralized exchanges" or "top crypto lending protocols", not brand or ticker lookups. We recorded, for each company, in how many of those prompts the company was actually named by the AI. That raw count, named in N of about 50 prompts, is the headline metric throughout this article.
One methodological note on selection bias: these 75 companies are medium-cap projects with real market presence and established communities. They are almost certainly more visible to AI systems than the thousands of smaller web3 startups that never built durable, well-cited public content. The true picture across the long tail of crypto is almost certainly weaker than what we measured here.
The band distribution: most companies cluster at the bottom
We assign every audited company to one of four AIO presence bands. The table below shows how the 75 companies distributed across those bands.
| AIO Band | Companies (of 75) | What it means |
|---|---|---|
| No presence | 35 | The audit’s lowest presence band. AI engines rarely surface this company as a category answer, though the band is a qualitative label and some companies in it were still named in as many as 17 of 50 prompts |
| Weak | 28 | Named inconsistently, with shallow presence and no category ownership. The bands are qualitative labels rather than fixed count thresholds, and their prompt counts overlap with neighbouring bands |
| Moderate | 10 | Named across many prompts with some consistency; recognised in its category but not dominant |
| Strong | 2 | Named across nearly all relevant prompts; AI engines treat this company as a default category answer |
The two companies that reached Strong were PancakeSwap (CAKE, decentralized exchange), named in 47 of 50 prompts, and Uniswap (UNI, decentralized exchange), named in 45 of 50 prompts. Both are long-established, widely used decentralized exchanges. We measured only how often each was named as an answer, not why.
The top of the table: Moderate companies and what separates them
Ten companies reached Moderate. Here is the full list with their prompt counts.
- Jupiter (JUP, decentralized exchange): named in 50 of 50 prompts
- NEXO (NEXO, crypto lending protocol): named in 46 of 50 prompts
- Aave (AAVE, crypto lending protocol): named in 43 of 50 prompts
- Arbitrum (ARB, layer 2 scaling network): named in 42 of 50 prompts
- VeChain (VET, layer 1 blockchain): named in 42 of 50 prompts
- Ethena (ENA, crypto lending protocol): named in 41 of 50 prompts
- Jito (JTO, liquid staking protocol): named in 41 of 50 prompts
- Filecoin (FIL, decentralized storage network): named in 40 of 50 prompts
- Curve DAO (CRV, decentralized exchange): named in 38 of 50 prompts
- Celestia (TIA, layer 1 blockchain): named in 35 of 50 prompts
Notice that Jupiter was named in 50 of 50 prompts yet sits in Moderate rather than Strong. The band reflects not just raw count but the consistency and category ownership signalled across prompt types. This is an important distinction: a high raw count does not automatically equal category dominance if the appearances are spread thinly across many unrelated query types.
The Weak band: named, but not the answer
Twenty-eight companies landed in the Weak band. A selection that illustrates the range:
- Blockchain Capital (BCAP, crypto lending protocol): named in 40 of 50 prompts
- Pyth Network (PYTH, oracle network): named in 40 of 50 prompts
- TAC (TAC, layer 2 scaling network): named in 38 of 50 prompts
- Pendle (PENDLE, crypto lending protocol): named in 35 of 50 prompts
- Ondo (ONDO, real world asset protocol): named in 33 of 50 prompts
- Bittensor (TAO, AI crypto network): named in 32 of 50 prompts
- KuCoin (KCS, centralized crypto exchange): named in 32 of 50 prompts
- Tezos (XTZ, layer 1 blockchain): named in 24 of 50 prompts
- Cosmos Hub (ATOM, interoperability protocol): named in 21 of 50 prompts
- Algorand (ALGO, layer 1 blockchain): named in 21 of 50 prompts
Cosmos Hub and Algorand are well-known projects with multi-year track records and substantial developer ecosystems. Both were named in 21 of 50 prompts. That is not invisibility, but it is not category ownership either. A user asking an AI which interoperability protocol to explore may or may not receive Cosmos in the answer, and that inconsistency is the practical problem for any team relying on AI-driven discovery.
The No presence band: 35 companies, named rarely
The largest single group, 35 of 75 companies, sits in the No presence band. The range runs from named in 17 of 50 prompts down to named in 5 of 50 prompts. A selection:
- Render (RENDER, decentralized GPU compute network): named in 17 of 50 prompts
- Artificial Superintelligence Alliance (FET, AI crypto network): named in 17 of 50 prompts
- POL (ex-MATIC) (POL, layer 2 scaling network): named in 15 of 50 prompts
- Injective (INJ, layer 1 blockchain): named in 9 of 50 prompts
- Gnosis (GNO, layer 1 blockchain): named in 9 of 50 prompts
- Monad (MON, layer 1 blockchain): named in 8 of 50 prompts
- Conflux (CFX, layer 1 blockchain): named in 7 of 50 prompts
- Lighter (LIT, decentralized exchange): named in 5 of 50 prompts
- Plasma (XPL, layer 2 scaling network): named in 5 of 50 prompts
Several of these, Injective, Gnosis, Render, are projects that any experienced crypto researcher would recognise immediately. Their low prompt counts reflect a gap between real-world community knowledge and the structured, citable content that AI retrieval systems draw on. Being well-known inside crypto Twitter does not translate automatically into AI search presence.
What the audit does not tell us
This audit measured how often a company was named by ChatGPT and Google Gemini on one date, against one prompt battery, with one query location setting. It does not measure the quality or sentiment of those mentions, revenue impact, or what happens when prompts are localised to specific regions. The findings are correlational: we observed which companies are named and how often. We have not studied the internal workings of either AI product to explain why.
We also want to be direct about what this sample represents. These 75 companies are medium-cap, established projects, not early-stage tokens. The selection almost certainly over-represents visibility compared with the broader web3 ecosystem. Smaller projects, newer protocols, and niche DePIN networks that did not make the CoinGecko medium-cap band on 30 June 2026 are likely to show weaker numbers than what we found here.
What teams can do with this
The companies that reached Strong or Moderate presence tend to be the most established, widely-referenced names in their category. Our audit measured only how consistently each company was named by the AI, not its content, documentation, or citation trail, so we cannot attribute the gap to any single cause from this data alone. In our wider experience the levers that tend to matter are clear, category-explicit content on a company’s own properties and a strong trail of independent third-party references, and that is where AIO focuses.
AIO's approach to closing that gap centres on building the content and citation structure that AI retrieval systems can actually use. If you want to see where your own project sits against this benchmark, you can run a free audit at aio.io.
Frequently asked questions
How did AIO select the 75 companies in this audit?
We used the CoinGecko market-cap ranking and took companies in the medium band (export const BLOG_CONTENT: Record
What does "named in N of 50 prompts" actually mean?
Each company faced roughly 50 real category-level prompts put to consumer ChatGPT and consumer Google Gemini on 30 June 2026, with query location set to Global. Examples include questions like "best decentralized exchanges" or "top crypto lending protocols". If the AI named the company in its answer to a given prompt, that counted as one. The total across all prompts is the figure we report. It is a raw count, not a score or a percentage.
Why do some companies with high prompt counts sit in a lower band than companies with fewer mentions?
The AIO presence band reflects more than the raw count. It captures the consistency and category ownership signalled across different prompt types. A company named many times but spread thinly across unrelated query categories signals differently to our audit than a company named consistently within its core category. Jupiter, for example, was named in 50 of 50 prompts but sits in Moderate rather than Strong because the band methodology accounts for how that presence is distributed.
Does this audit cover all crypto companies?
No. This is a focused audit of 75 medium-cap companies selected by a specific methodology on one date. It is not a census of the crypto industry. Because these are established, medium-cap projects, they are likely more visible than the thousands of smaller web3 startups outside this band. The findings should be read as a benchmark for medium-cap projects, not a claim about the broader ecosystem.
Which AI products were tested, and on what date?
We tested consumer ChatGPT and consumer Google Gemini on 30 June 2026, with query location set to Global. We describe them as the consumer products available on that date rather than by internal model version, because what matters for AI search visibility is how the products behave for real users.
How can my project find out its own AIO presence band?
You can run a free audit at aio.io. The audit uses the same pipeline we applied to these 75 companies and will tell you how many of our prompt battery your project is named in, which band you sit in, and where the content and citation gaps are.