AI search visibility for crypto
That is the annoying part. Every box got ticked, and the buyer still never heard your name.
Find out what they say instead. 45 checks, run by hand, back in 48 hours.
Five wins no engine can see
Five things crypto teams were proud of this year. Every one of them true, every number real. And not one of them changed what a model says about you.
“We booked eight KOLs.”
Every post of it lives on X.
Only Grok trains on X.
“Airdrop hit 40,000 wallets.”
Does ChatGPT know that happened?
Will it quote you over a competitor?
“We rank first on Google.”
The AI Overview sits above you.
First place under it is second place.
“Impressions up 300%.”
Impressions of what, exactly?
Your buyer never loaded a results page.
“Thread did 400k views.”
Nothing indexed it. Nothing cited it.
Ask any engine today. It never happened.
Nobody here made a bad call. The KOLs were the going rate. The airdrop was the playbook. The ranking is real and it took months.
But every one of those five wins was measured somewhere your buyer was not standing.
What actually happened
Somewhere in the middle of all that, the person who would actually integrate you stopped opening Google. They opened ChatGPT and typed the thing they wanted, in plain words, the way people type when nobody is watching.
Three names. Then they opened three tabs, and every one of those tabs was a competitor.
You did not lose that deal. You were never in it. And here is the part that should genuinely bother you: nothing in your analytics recorded any of it. There was no click to log. No impression, no bounce, no session. From where you are sitting, that buyer never existed.
This is happening right now, to your protocol, and your dashboard says everything is fine.
Here is the truth.
It has been six years since DeFi summer. Six. The airdrop playbook is finished and everyone at this point knows it is finished, because everyone farmed it, including you, including me. Points programmes bought you a spike and a mercenary user base that left on schedule.
Crypto is growing into the boring part of its life. Revenue that shows up in a spreadsheet. Users who stay because the product is good. Compliance teams. Buyers who do diligence before they commit capital, and who are not doing that diligence on Crypto Twitter.
They are doing it in a chat window. The paragraph a model writes about your category is the new front page, and right now somebody else is in it.
That paragraph is not fixed. It is not an algorithm you wait out. It gets assembled from what is written about you and where models go looking, and it can be changed on purpose.
I know it can, because I have done it.
What that looks like
Velar made a bet on Stacks. Perpetual futures settled on Bitcoin through Stacks, rather than a wrapped asset parked on some other chain. It is a good bet and a genuinely hard thing to explain to somebody in one sentence.
So we did not try to explain it. We wrote six words instead.
“World’s first bitcoin perpdex.”
Was Velar actually first? Depends who you ask and when you ask them. Nobody had written it down. That is the whole point, and it is the part most teams never notice about their own category: a category nobody has defined is a category nobody is being cited for, and the definition goes to whoever writes it properly first.
So we wrote it properly. Then we put it everywhere a model goes looking, which is the slow unglamorous half and the reason the position holds once you have it.
Velar was not the first time. I did the same thing for RemoteStack, a job board I own and run, which now gets pulled into AI answers about remote hiring. That one had no client, no budget and nobody else to blame, which is the only reason I know the method works and not just the brand.
Then I built kymo, because after all of it I still could not tell you which crawler had read what, or which citation actually sent a human. Nothing on the market would say. So every month of client work now reports on numbers from a tool I wrote myself.
Research
1.7%
I put the same ten money problems to four engines — a freelancer in Lagos, a designer in Buenos Aires, an agency paying eight contractors. 480 responses. They raise crypto a third of the time and describe it accurately. They almost never say anybody’s name.
Which means the thing standing between a protocol and the answer is not a model that dislikes crypto. It is retrieval, and retrieval can be changed on purpose.
Read the full report →Free to read, raw data published. All research →
Or read both editions as one guide: The Crypto AI Visibility Playbook →
First the SEO bros ruined Google. Now it is our turn to ruin ChatGPT.
Sorry in advance. Somebody was going to.How
Measure the baseline before touching a page.
Match the site to the words buyers actually use.
Retrieval takes one block and judges it cold.
Vendor docs and forums, not press. Then re-run it.
This is the only one we have not run yet.
Twenty minutes. I run your questions before we talk, so you are not spending the call explaining your protocol to a stranger.
You are talking to Narender. Six years marketing crypto products, from inside the market, not from an agency floor.
Answer engines are the sharp end of it, not the whole job. Positioning, content, GTM, launches, the community work that actually moves numbers — I run all of it, hands-on.
No call, no card. 45 answers read by hand, scored out of 135, in your inbox within 48 hours.