Crypto GEO is the work of measuring and improving how AI assistants treat a Web3 brand in their answers: whether they retrieve its pages, cite them, mention the brand by name, and recommend it for a buyer's need. We pair answer-engine optimization (restructuring pages so answers can be extracted) with generative-engine optimization (content and third-party sources worth citing) in one retainer with one prompt panel and one monthly report. Google SEO stays a separate, related discipline.
- Crypto brands with healthy Google rankings that rarely appear in ChatGPT, Perplexity or Google AI answers
- Buyers comparing an AEO agency with a GEO agency: the shared work is billed once, not twice
- MiCA-bound exchanges, FCA-registered firms and licensing partners whose regulatory copy must hold up when quoted
- B2B fintech and tokenization platforms whose buyers research vendors inside AI tools first
- Teams that track organic clicks but cannot yet show pipeline from AI-assisted research
Quick Facts
| Parameter | Value |
|---|---|
| Monthly fee | From $2,800 USD |
| Minimum term | 3 months |
| Systems tracked | ChatGPT search, Perplexity, Gemini, Claude, Google AI Overviews and AI Mode. The list is confirmed per engagement; model versions are logged per run, not promised |
| Prompt panel | 22 buyer prompts chosen from your ICP in discovery, spanning commercial, comparison, technical and regulatory questions plus a branded control group, run weekly. Full answer text, cited URLs and brand mentions are logged |
| Metrics, reported separately | Retrieved, cited, mentioned, recommended. Never merged into one score |
| AEO page work / month | 4–6 priority pages restructured: H1 disambiguator, Quick Facts, direct-answer first sentences ≤30 words |
| GEO content / month | 2–4 long-form pieces, each built to answer one buyer question |
| Structured data | Organization, Person (real profiles only), Service, Article, BreadcrumbList, and FAQPage where the FAQ is visible. JSON-LD, validated per page. It helps machines read the site; it does not cause citations |
| Crawler access | Search bots stay allowed (OAI-SearchBot, PerplexityBot, Googlebot). Training-bot policy (GPTBot, ClaudeBot, Google-Extended) is the client's decision |
| llms.txt | Optional manifest with canonical links, served at 200 and re-checked quarterly. Not a ranking or citation factor |
| Reporting | Monthly report: per-system rates for each of the four metrics, source URLs, prompt-by-prompt comparison against three named competitors |
What do retrieved, cited, mentioned and recommended mean?
Retrieved: the system fetched your page. Cited: your URL appears as a source. Mentioned: your brand name appears in the answer. Recommended: the answer suggests you for the buyer's need.
These are four different outcomes and we log them separately. A page can be retrieved and never cited. A brand can be mentioned without a link. A recommendation is the strongest outcome and the least stable one, because it depends on the exact prompt, the system and the day.
Every logged result carries the prompt, the system, the date and the answer text, so a number in a report can be traced back to a real answer. We do not roll the four metrics into a single visibility score: a blended score hides which of the four moved.
What is the difference between AEO and GEO?
AEO restructures existing pages so a self-contained answer can be extracted. GEO produces new content and third-party sources worth citing. We measure both on one prompt panel.
AEO (answer engine optimization) works on pages you already have: an H1 that disambiguates the service, a Quick Facts table near the top, H2 questions with a direct first sentence under 30 words. GEO (generative engine optimization) works on what assistants have available to quote: new pieces that answer one buyer question, dated claims, named experts, links to primary sources, and the third-party pages that describe your brand.
Google SEO targets ranked links on a results page. AI search optimization targets the passages that assistants quote or link inside an answer. The two surfaces overlap, and how much depends on the site, so we scope it in discovery instead of quoting a percentage.
Why is it one retainer?
AEO and GEO share the same infrastructure: schema, internal linking, named-expert bylines and the prompt panel. Billing them as two retainers would charge twice for the same base.
The shared base is built once: entity data and schema, internal links between service and case pages, byline and reviewer data for named experts, and the prompt panel that shows whether any of it moved a result. Page restructuring and new content then draw on that base in the same month, which is why the fixed scope lists AEO pages and GEO pieces side by side.
How much does a crypto GEO retainer cost?
From $2,800 a month on a three-month minimum. Combined with Crypto SEO, the two retainers are quoted at $5,200 a month instead of the additive $6,000.
$2,800 is the base rate for crypto exchanges and DeFi or SaaS products; the pricing page shows how a different niche moves it. The fee covers 4–6 restructured pages, 2–4 long-form pieces, the weekly prompt panel and the monthly report, as listed in Quick Facts. Some clients run AI search alone because their Google rankings are already healthy. Others combine it with Crypto SEO to avoid a second discovery phase.
How is the prompt panel built?
We build the panel in discovery from your ideal customer profile, current sales conversations and a crypto-specific prompt corpus, then run it weekly on every tracked system.
The panel mixes commercial prompts (who should I hire), comparison prompts, technical prompts, regulatory prompts and a branded control group. Each prompt runs weekly. For every run we capture the full answer, the sources cited and their order, and brand mentions in unbranded category queries.
How are the results reported?
Monthly, as rates per system for each of the four metrics, with the weekly trend and a prompt-by-prompt comparison against three named competitors.
Answers differ between runs, even for the same prompt on the same day, so a single result is never reported as a trend. Each citation carries a simple quality note: favorable, neutral, or comparative against a competitor. We lead with the method rather than a headline count, because a count without a method is easy to game.
Which signals can you influence, and which can't you?
We influence your pages, your structured data, crawler access and the third-party sources about you. We do not control model updates, personalization or what any system decides to answer.
What we can work on: how clearly each page answers one buyer question, whether search crawlers can reach the page, whether the entity data on the site is consistent, and whether credible third-party sources describe your brand accurately.
What we cannot control: model and index updates, personalization, location and conversation history, and the order of sources inside an answer. Any of these can move a result without a change on your side. That is why we report ranges, why we do not guarantee citations or recommendations, and why a drop in one week is read against the trend before anyone changes a page.
Do schema, llms.txt and crawler access change what AI systems cite?
Crawler access is a precondition, so we keep search bots allowed. Schema and llms.txt are not documented as levers that cause citations; schema helps machines read your entities.
Google's guidance for its AI features is that the same SEO fundamentals apply and that no special markup or extra files are needed (Google Search Central). OpenAI's publisher guidance says a site that wants to appear in ChatGPT search must allow OAI-SearchBot, and that GPTBot is a separate choice about training use (OpenAI Publishers FAQ). We keep search bots allowed and leave the training-bot decision to you.
We still implement JSON-LD, because clean entity data (a single Organization, real Person profiles, Service and Article pages that point at each other) removes ambiguity about who you are. We ship a short llms.txt with canonical links as a courtesy manifest, and we re-check that it is served. We know of no provider documentation that ties it to citations, so it does not appear in any forecast.
How does regulatory copy survive being quoted by an AI?
The quoted sentence still traces back to your page. We review YMYL copy before publication; your compliance lead or counsel approves it.
When an assistant quotes a line from your site such as "guaranteed APY" or "regulator-endorsed", the quote points back to your page. The page is where the wording can still be controlled, which is why regulatory review sits before publication, not after the first citation.
Our review checks every YMYL piece against four banned-claim categories: guaranteed returns, risk-free framing, implied regulator endorsement without a specific licence reference, and universal-suitability suggestions. For MiCA-bound clients we keep a per-jurisdiction copy-rules document. This is an editorial check, not legal advice. Legal approval stays with your in-house compliance lead, your retained counsel or a reviewer you contract.
Where can I read case studies for this work?
Our case studies describe the scope, period and what was counted. Two are public (CryptoLawIndex, Fast Offshore Licenses) and two are NDA cases.
Read the CryptoLawIndex case and the Fast Offshore Licenses case for the public examples, and the tokenization platform case for an NDA engagement where the client's technical buyers research in AI tools. Each describes the scope, the period and the counting rules for its figures.
Frequently asked questions
How is this different from your Crypto SEO retainer?
Crypto SEO builds access, authority and Google rankings. Crypto GEO measures and improves how AI assistants retrieve, cite, mention and recommend you. They share a technical base.
Many clients run both as one program, which saves duplicate discovery. We quote the combined scope at $5,200 a month instead of the additive $6,000. Some clients run AI search alone because their Google rankings are already healthy.
Which AI systems do you track?
ChatGPT search, Perplexity, Google AI Overviews and AI Mode, Gemini and Claude. We confirm the list per engagement because products and their modes change.
Model versions are logged for every run. We do not promise results on any named model, since a model can be replaced without notice.
Can you work with our existing Google SEO agency?
Yes. We coordinate on shared infrastructure (schema, internal links, sitemap discipline) and stay clear on different surfaces: their blue-link work, our AI-answer work.
Standard pattern: a quarterly call between the two teams, a shared content calendar so work is not duplicated, and clear ownership lines so neither agency has to ask permission to ship.
Does llms.txt make AI assistants cite us?
No provider documents that, and we know of no evidence for it. It is an optional manifest; we ship one but do not count it as a citation lever.
We deploy it in month one and check quarterly that it is still served at 200, so it never points at dead URLs.
How do we measure AI-driven pipeline when referrers are missing?
Three layers: branded-search lift in Search Console, self-reported source on lead forms, and referrer logs where a platform passes one, such as utm_source=chatgpt.com.
Last-click attribution undercounts AI-assisted leads because buyers often finish on direct or branded search after an assistant named you. The three layers together catch what referrers alone miss.
Do you guarantee citations or recommendations?
No. AI answers differ between runs and systems change without notice. We report rates and ranges, and we show the logs behind them.
If the work is done and the numbers have not moved, we say so in the monthly report and extend the scope with you rather than extending the invoice.