NEWWorld's first AI visibility audit tool for Web3 is live.Run free audit →
Deep guide · ~4,800 words · Last reviewed April 2026

The AI Visibility Guide for crypto sites.

Crypto buyers research through ChatGPT, Perplexity and Claude more than they research through Google. AEO matters more for crypto than for any other vertical. This guide covers the four readiness dimensions, citation testing methodology, prompt design, monitoring and the specific anti-patterns crypto sites fall into.

View pricingTest how ChatGPT and Perplexity cite your site
Chapter 01
// Foundations

AEO vs SEO: what is actually different

SEO optimizes for ranking position in search engine results pages. AEO optimizes for citation in AI-generated responses from ChatGPT, Perplexity, Claude and similar engines. The technical foundation overlaps significantly; the citation model diverges meaningfully.

Traditional SEO signals

Optimized for SERP rank

  • Backlink count and Domain Authority graph
  • Click-through rate and behavioral signals
  • Keyword density and topical depth
  • Meta title and meta description CTR optimization
  • Link velocity and anchor text profile
vs

AEO signals

Optimized for AI citation

  • Schema correctness and structured fact extraction
  • Factual density (named entities, numbers, dates)
  • robots.txt access for AI bots specifically
  • Authority source citations (themselves cited by AIs)
  • Server-rendered content (not behind JS render)

Three patterns matter for crypto specifically. First, AI search engines weight schema more heavily than Google does because they extract facts directly from JSON-LD. Second, AI engines penalize JavaScript-only renders aggressively because the majority of AI crawlers do not execute JS. Third, AI engines weight authority citations differently: a site cited by DefiLlama scores higher with crypto-focused prompts than a site with twice the generic backlinks.

Why AEO matters more for crypto

Crypto is technical, fast-moving and trust-sensitive. Buyers use AI search disproportionately for crypto research because traditional search results are noisy with promotional content. AI engines pull from a curated set of sources, which favors sites with strong AEO signals.

Chapter 02
// The model

The 4-factor LLM citation model

Different LLMs use different signals but a 4-factor model captures most of the citation behavior across ChatGPT, Perplexity and Claude. A site weak on any one factor gets cited less than its content quality alone would predict.

01

Schema correctness

Does the page expose structured facts the AI can extract reliably. JSON-LD with the right type tells the AI exactly what your page describes. FinancialProduct on a token page tells ChatGPT this is a financial instrument with these specific properties.

02

Factual density

How many direct factual statements per paragraph. Named entities (founders, audit firms, exchanges), numbers (TVL, holders, transactions per second), dates (audit dates, launch dates) all count. Marketing copy without facts gets cited rarely.

03

robots.txt access

Can the AI bot actually crawl the page. Many crypto sites block AI bots accidentally through aggressive Cloudflare filtering or outdated robots.txt. The page is invisible regardless of how good the content is.

04

Authority citations

Is your source itself cited by other sources the AI trusts. AIs prefer to cite chains of authority. If DefiLlama, CoinGecko or audit firm reports cite your project, your domain inherits authority for crypto-specific prompts.

The compounding effect

A site scoring 7/10 on all four factors gets cited more than a site scoring 10/10 on three factors and 2/10 on one. The factors are multiplicative, not additive. Fix the weakest factor first.

Chapter 03
// Factor 01

Schema correctness for AEO

Schema correctness is the highest-leverage AEO fix because it is binary and fast to implement. Either the page exposes the right schema type or it does not. The Crypto SEO Guide covers schema decision tree by page type; this chapter focuses on the AEO-specific patterns.

The single most common mistake on crypto sites: token information pages with generic Product schema instead of FinancialProduct. To an AI extracting facts, a Product is a thing being sold. A FinancialProduct is an instrument with specific monetary properties.

// Wrong · Generic Product on token page
{
  "@context": "https://schema.org",
  "@type": "Product",                       // AI sees: this is a thing
  "name": "PRTC Token",
  "description": "Native governance token"
}
// Right · FinancialProduct with monetary properties
{
  "@context": "https://schema.org",
  "@type": "FinancialProduct",              // AI sees: financial instrument
  "name": "PROTOCOL Token (PRTC)",
  "description": "Native governance token of the PROTOCOL DeFi protocol.",
  "category": "Cryptocurrency",             // AI sees: type of asset
  "feesAndCommissionsSpecification": "https://protocol.com/fees"
}

The AI extraction model treats these very differently. The first gets categorized loosely; the second gets categorized as a financial instrument with specific properties the AI can cite directly when asked about cryptocurrencies.

For deeper schema treatment by page type (token, exchange, wallet, organization), see the dedicated Token Schema Guide with copy-paste examples for every page type.

Schema validation chain

Run JSON-LD through the Google Rich Results Test, then through Schema.org validator, then test the actual citation behavior with the Token Schema Tester. All three checks pass means the schema works for AEO.

Chapter 04
// Factor 02

Factual density patterns

Factual density measures the ratio of direct factual statements to marketing language. AI engines reward content that names entities, cites numbers and dates clearly. Marketing copy without facts gets cited rarely regardless of how well-written it is.

Three transformation patterns convert low-density crypto marketing copy into high-density AEO-friendly content. Each takes the same key fact and exposes it as something an AI can extract and cite.

Pattern 01 · Performance claims

Marketing copy

  • "the fastest blockchain in DeFi"
  • "lightning-fast transactions"
  • "unmatched scalability"
  • "sub-second finality"

High factual density

Citable facts

  • "processes 65,000 transactions per second on testnet"
  • "average finality 0.4 seconds, 99th percentile 1.2 seconds"
  • "benchmarked against Ethereum L2s in Q1 2026"
  • "verified by Independent Benchmark Foundation report"

Pattern 02 · Trust claims

Marketing copy

  • "audited and secure"
  • "battle-tested codebase"
  • "industry-leading security"
  • "trusted by thousands"

High factual density

Citable facts

  • "audited by Trail of Bits in March 2024"
  • "second audit by OpenZeppelin in October 2024"
  • "3 years live, 2.4 million transactions processed"
  • "$847 million TVL across 18,000 active wallets"

Pattern 03 · Tokenomics

Marketing copy

  • "fair launch"
  • "community-owned"
  • "deflationary mechanics"
  • "strong tokenomics"

High factual density

Citable facts

  • "100M total supply, 65M circulating, no team allocation"
  • "54% held by community wallets, 8,400 unique holders"
  • "2% of fees burned per transaction, 4.2M tokens burned to date"
  • "governance threshold 1M tokens, 12 active proposals"

Density compounds

The more direct facts per paragraph, the higher the citation rate. Aim for at least 3 named entities, numbers or dates per 100 words. Below that threshold, AI engines treat the content as marketing rather than reference.

Chapter 05
// Factor 03

robots.txt for AI bots

The third citation factor is binary: either the AI bot can crawl the page or it cannot. Many crypto sites block AI bots accidentally through aggressive Cloudflare filtering or outdated robots.txt rules.

The seven AI bots that matter for AEO in 2026 each serve different purposes. Allow all of them unless you have a specific reason not to.

// robots.txt · the canonical AI bot allow list for 2026
# OpenAI
User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

# Anthropic
User-agent: ClaudeBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

# Perplexity
User-agent: PerplexityBot
Allow: /

# Google AI
User-agent: Google-Extended
Allow: /

# Apple Intelligence
User-agent: Applebot-Extended
Allow: /

The CDN gotcha

robots.txt at the origin is correct but Cloudflare or similar CDN rules block AI bots at the edge. Test with curl using each user-agent string. Edge filtering blocks the bots before they ever reach your robots.txt.

The free Web3 Robots.txt Checker tests for the seven bots above plus catches deprecated user-agent strings (the OpenAI bots renamed several times in 2024-2025) and CDN-level overrides.

One specific case worth flagging: GPTBot trains the model; OAI-SearchBot fetches for ChatGPT search. Allowing only GPTBot does not mean ChatGPT can cite your live page; OAI-SearchBot is the bot that retrieves current pages for citation. Same pattern for ClaudeBot vs Claude-SearchBot.

Chapter 06
// Factor 04

Authority citations and AI weighting

AI engines prefer to cite chains of authority. A site cited by sources the AI already trusts inherits citation weight for related prompts. For crypto specifically, the authority sources differ from generic SEO authority sources.

Three tiers of authority sources matter for crypto AEO. Each tier carries different citation weight depending on the prompt category.

T1

Tier 1: Crypto data and audit

DefiLlama, CoinGecko, CoinMarketCap, Etherscan, Polygonscan, Solscan, Trail of Bits, OpenZeppelin, CertiK, Quantstamp. AI engines cite these heavily for any prompt involving TVL, market data or security context.

T2

Tier 2: Crypto media and research

CoinDesk, The Block, Decrypt, Cointelegraph, Messari, Delphi Digital, a16z crypto blog. Cited frequently for current events, comparison prompts and category overview prompts.

T3

Tier 3: General financial media

Bloomberg, Reuters, WSJ, FT. Lower weight for crypto-specific prompts but high weight for YMYL trust signals. Mention here helps for regulatory and institutional credibility prompts.

The strategy that works

Get listed on DefiLlama and CoinGecko first. These two cover ~60% of all crypto-related AI citation prompts. Audit firm coverage on Trail of Bits, OpenZeppelin or CertiK adds another ~20%. Tier 2 and Tier 3 mentions cover the remaining ~20%.

Avoid the temptation to chase generic high-DA backlinks instead. A site with 5 Tier 1 crypto authority citations outperforms a site with 50 generic high-DA backlinks for crypto-specific AEO. The compute is different.

Chapter 07
// Methodology

Citation rate testing methodology

The only way to measure AEO progress is direct citation rate testing. Build a prompt set, run it across multiple LLMs, score citations. Re-test on a fixed cadence to measure improvement.

The five-step methodology below is what the Crawlux audit runs internally. The methodology is open; other auditors can use it without Crawlux.

01

Build the prompt set

Generate 30 to 50 category-relevant prompts that real users would ask. Cover four intent types: investigative (what is X), comparative (X vs Y), transactional (best X for Y), educational (how does X work).

02

Run prompts across 3 LLMs

Test each prompt against ChatGPT, Perplexity and Claude. Record whether your domain was cited, position of citation in response and which competitive sources got cited alongside.

03

Score citation rate per dimension

Calculate citation rate per LLM, per intent bucket and per competitor. Identify systematic gaps: which LLM cites you least, which intent type you fail to capture, which competitor consistently displaces you.

04

Diagnose failure modes

For prompts where competitors cite but you do not, diagnose against the 4-factor model: schema gap, factual density gap, AI bot blocking, authority citation gap. The 4 factors map to 4 fix categories.

05

Re-test quarterly

Run the same prompt set quarterly to measure improvement. LLM citation patterns shift faster than Google rankings. Quarterly cadence catches drift; monthly cadence measures incremental gains during active optimization.

The free AI Citation Checker runs 5 prompts per audit; the full Crawlux Pro audit runs 30+ category-specific prompts with prompt-by-prompt diagnostics.

Chapter 08
// Prompt sets

Prompt set design per crypto category

Generic prompt sets miss crypto-specific citation patterns. The Crawlux audit detects category automatically and runs the relevant prompt set. Four crypto categories cover most projects.

DeFi protocol prompts

~30 prompts in the set

  • What is the best DeFi protocol for [use case]?
  • Compare [protocol A] vs [protocol B] for yield
  • How does [specific feature] work on [protocol]?
  • Which DeFi protocol has the lowest fees for [action]?
  • Is [protocol] audited and what are the audit findings?
·

Exchange prompts

~30 prompts in the set

  • What is the best crypto exchange for [region]?
  • Compare [exchange A] vs [exchange B] fees
  • Which exchange supports [specific token]?
  • Is [exchange] regulated in [jurisdiction]?
  • What are the trading limits on [exchange]?

NFT and GameFi prompts

~30 prompts in the set

  • Which NFT marketplace has the lowest fees for [chain]?
  • What is the floor price of [collection]?
  • Compare [game A] vs [game B] tokenomics
  • Which NFT marketplace supports [feature]?
  • How does [game] handle in-game token economy?
·

Wallet prompts

~30 prompts in the set

  • What is the best [chain] wallet for [use case]?
  • Compare [wallet A] vs [wallet B] security
  • Does [wallet] support [specific feature]?
  • Is [wallet] open source and audited?
  • How does [wallet] handle [transaction type]?

Why category-specific matters

Generic SEO prompt sets ask "what is the best blockchain". Real crypto buyer prompts ask "what is the best Solana DEX with sub-cent fees for sub-$1000 trades". The specificity is where citation rate diverges.

Chapter 09
// Operating cadence

Monitoring AEO over time

Citation rate drifts faster than Google ranking position. LLM training data refreshes, prompt-routing models update independently and authority weighting shifts as AI engines learn from real user behavior. A site that ranked highly in Q1 may rank lower in Q2 without changing anything.

Three monitoring patterns catch the different types of drift. Each operates on a different timescale.

Quarterly deep audit

90-day cadence

  • Run full prompt set across all 3 LLMs
  • Score against the 4-factor model
  • Diagnose failure modes per prompt
  • Build 30-60-90 day fix roadmap
  • Use Crawlux Pro audit for the deep run
·

Monthly snapshot

30-day cadence

  • Run abbreviated prompt set (10-15 prompts)
  • Track per-LLM citation rate trend
  • Catch new competitive entrants
  • Use free AI Citation Checker for sampling
  • Skip if you are not actively optimizing

Continuous: schema and robots.txt

Weekly automated checks

  • Validate schema on key templates
  • Test robots.txt with all 7 AI bot user-agents
  • Catch CDN rule changes that block bots
  • Detect schema regressions from deploys
  • Alert on negative changes only
·

Annual deep retrospective

Once per year

  • Review 12 months of quarterly audits
  • Identify systematic gaps that did not close
  • Re-evaluate prompt set against new buyer questions
  • Adjust strategy for next 12 months
  • Compare against industry benchmarks

The monitoring trap

Some teams monitor weekly and burn cycles chasing noise. Quarterly is the right deep cadence; monthly is fine for active optimization; weekly is overkill except for automated schema and robots.txt checks.

The companion Crypto SEO Guide covers the full quarterly audit workflow including the other 7 dimensions beyond AEO. The Token Schema Guide goes deeper on schema selection per page type.

// AI Visibility Guide FAQ

Common questions

Eight questions covering the AEO terminology, methodology and decision points.

What is the difference between SEO and AEO?

SEO optimizes for ranking in search engine results pages (SERPs). AEO (Answer Engine Optimization) optimizes for citation in AI-generated responses from ChatGPT, Perplexity, Claude and similar engines. The technical foundation overlaps significantly: schema, factual content, page speed and authority signals matter for both. The divergence is in the citation model. Search engines reward backlink graphs and behavioral signals; AI engines reward direct factual statements, schema correctness and authority citations. For crypto sites specifically, AEO often matters more than traditional SEO because crypto buyers research heavily through AI search.

How do AI search engines decide what to cite?

Different LLMs use different signals but a 4-factor model captures most patterns. First, schema correctness: does the page expose structured facts the AI can extract reliably. Second, factual density: are key facts stated directly with named entities, numbers and dates. Third, robots.txt access: is the page crawlable by the AI bot. Fourth, authority citations: is the source itself cited by other sources the AI trusts. Sites scoring well on all four dimensions get cited more than content quality alone would predict.

Which AI bots should I allow in robots.txt?

For crypto sites optimizing for AEO, allow GPTBot (OpenAI training), OAI-SearchBot (ChatGPT search), ClaudeBot (Anthropic training), Claude-SearchBot (Claude search), PerplexityBot (Perplexity), Google-Extended (Gemini training) and Applebot-Extended (Apple Intelligence). Common mistake: aggressive Cloudflare bot filtering blocks these accidentally even when robots.txt allows them. Test with curl using each user-agent to verify access.

How do I measure AI citation rate?

Build a prompt set of 30 to 50 category-relevant questions real users would ask. Run each prompt against ChatGPT, Perplexity and Claude. Track citation rate per LLM, per intent bucket and per competitor cited alongside you. Re-run quarterly to measure trend. The free Crawlux AI Citation Checker runs 5 prompts per audit; the Pro audit runs 30+ prompts with category-specific prompt sets.

What is factual density and how do I improve it?

Factual density measures the ratio of direct factual statements to marketing language in your content. AI engines reward content that names entities, cites numbers, dates, founders and audit firms. Crypto sites often hide facts behind metaphors: replace "the fastest blockchain" with "processes 65,000 transactions per second on testnet". Replace "audited and secure" with "audited by Trail of Bits in March 2024 and OpenZeppelin in October 2024". Density compounds: the more direct facts per paragraph, the higher the citation rate. Aim for at least 3 named entities, numbers or dates per 100 words.

How often should I re-test AI citation rate?

Quarterly minimum, monthly if you are actively improving AEO. LLM citation patterns shift faster than Google ranking patterns because LLM training data refreshes and prompt-routing models update independently. A site that got cited heavily in Q1 might get cited less in Q2 if a new authority source gets weighted higher. Quarterly testing catches drift; monthly testing measures incremental gains during active optimization.

Do I need separate AEO strategies for ChatGPT vs Perplexity vs Claude?

No, the underlying signals are largely the same. Schema correctness, factual density, robots.txt access and authority citations matter for all three. The differences are at the margins: Perplexity weights recency and source diversity heavily; ChatGPT weights training data and search results separately; Claude weights document structure and citation depth. Optimizing for one tends to lift the others. Track citation rate per LLM to catch divergence but optimize for the shared signals first.

Is AEO the same as Generative Engine Optimization (GEO)?

Mostly yes. AEO and GEO are used interchangeably across the industry, though AEO is more common in technical SEO communities and GEO is more common in marketing communities. Both refer to the practice of optimizing for citation in AI-generated responses. Some practitioners distinguish AEO (broader, includes traditional answer engines like Google featured snippets) from GEO (narrower, specifically generative AI). For practical purposes the work is the same.

RUN YOUR FIRST AUDIT FREE

See Crawlux on your own crypto site.

No signup, no credit card. Full Web3-tuned audit report in 60 seconds.

Free first audit · No signup · 60 seconds · Full PDF report