3 audits a week. Run by hand. Currently 3 slots left this week.
Why a crypto-specific checker exists
Four things break when a general AI visibility tool is pointed at a crypto brand. Each one produces a confident number that is wrong in a specific direction.
Ticker and name collisions
Generic AI visibility tools assume your brand name resolves to one entity. In crypto it often does not. A token that shares a name with a dead fork, with a project on another chain, or with an ordinary English word gets answers that are about somebody else entirely. The audit scores whether the engine resolves you before it scores whether it recommends you.
Too young to be in the model
A protocol that launched eight months ago may not exist in the model's weights at all. That is a different problem from existing and being ignored, and it needs a different fix. Generic tools report both as a zero. This one separates them, because the fix for absence is placement and the fix for being ignored is the writing.
The rug question
The highest-stakes answer a crypto brand faces is not “best DEX for low fees”. It is “is this a rug”. Every generic checker treats any mention of your brand as a point scored. If ChatGPT answers a trust query by raising an exploit, a lawsuit, or a project you have nothing to do with, that is a negative, and a scale that only goes up cannot see it. Trust queries are scored separately and the report quotes what the engines actually say.
Your citation surface is not G2 and Capterra
AI answers about crypto get sourced from CoinGecko, DefiLlama, Messari, protocol docs, governance forums and X. Every cited URL in your 45 answers is logged, so the report shows which surfaces are feeding the answers in your category. That is where the fix lives.
What is in the report
A PDF and a written read. Eight parts:
- Your AI Visibility Score out of 135, split into a money score out of 90 and a brand and trust score out of 45.
- Your band benchmarked against the other protocols audited in your vertical.
- The full 45 cell grid every query, every engine, every score, with the reason each cell got the number it got.
- The competitor set the two or three brands that actually get named and cited on your money queries, which is usually not who you think.
- Citation source breakdown which domains the engines pulled from, ranked by how often.
- Entity resolution check whether the engine knows who you are, on the right chain, in the right category.
- Trust query transcript what the three engines say when somebody asks whether you are safe.
- A prioritised fix list ranked by how many points each fix is likely to move, with the cheapest ones first.
How it works
- You submit your domain and vertical. Takes a minute.
- I run the sweep by hand. 15 buyer queries across 3 engines, logged out, incognito, first answer kept. No dashboard averages and no sampling.
- You get a PDF and a written read within 48 hours. No call required, no card, no signup.
Some tools promise this in 60 seconds. Those are sampled from a cache. This one is a person opening 45 tabs.
How the score is built
Each of the 45 cells is scored 0 to 3 against one rule, so two people scoring the same answer land on the same number.
0 absent. The answer does not mention you.
1 named. You appear, without a link.
2 cited. You appear and the engine links a page of yours as a source.
3 recommended. The engine puts you forward as an answer to the question, not as a passing mention.
| ChatGPT | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Perplexity | |||||||||||||||
| Google AIO |
10 money queries, 90 points 5 brand and trust queries, 45 points
45 cells, 0 to 3 each, 135 points in total.
Trust queries use the same scale with the sign reversed where the answer is hostile, which is why they are counted separately rather than folded into one total. The measurement approach behind all of this is written up in the published methodology, prompt by prompt, with the scripts.
Request the audit
The last field decides which 15 queries get asked, so put a real sentence in it.
Questions
Is this really free?
Yes, and there is nothing to buy at the end of it. It costs me about two hours per audit, which is why it is capped at 3 a week. The reason it exists is that most crypto founders have never seen what an engine actually says about their category, and the fastest way to show that is to go and read it.
How is this different from Ahrefs Brand Radar or a generic AI visibility checker?
Three differences, and they are all crypto-specific. A generic tool assumes your brand name resolves to one entity, which in crypto is often false. It cannot separate a protocol that is missing from the model's weights from one that is present and ignored. And it treats any mention as a point scored, so it reads an answer that raises an exploit or a lawsuit as visibility rather than as the problem it is.
Why only 3 audits a week?
Because 45 answers are read by a person. Each query is sent logged out and incognito to three engines, the first answer is kept, every cited URL is logged, and the cell is scored by hand. Running more than 3 a week would mean sampling, and a sampled audit is the thing this is meant to replace.
Do you keep or sell my domain and email?
No, and nothing is sold or shared. What I store is the six fields you submit, in Cloudflare KV, so I can run the audit and send it to you. There is no tracking pixel, no email sequence and no CRM. Ask and I will delete the record.
What if my protocol is too new to show up anywhere?
Then the report says so, and that is a genuinely useful result. A protocol that launched eight months ago may not be in the model weights at all, which is a different problem from being present and unrecommended and it takes a different fix. The audit separates the two rather than reporting both as a zero.
Can I re-run it later?
Yes. Ask again after 90 days and I will run the same 15 queries and send you both grids side by side. Sooner than that and the movement is mostly noise, because engines vary between runs even with nothing changed.
Find out what the engines say about your category before your next investor call. Request the audit.
Published 2026-09-14. Written by Narender Charan, who runs Kunzum.