AI visibility 101: how to get cited by ChatGPT, Claude, and Perplexity

Three failure modes account for ~90% of "we don't show up in ChatGPT" cases we audit: a blocked crawler, weak structured data, or no concise answer block on the page. Fix those three and you'll see your domain start appearing in cited answers within weeks — not months.

The bots that actually matter

AI tools don't crawl the web with one bot per company — they crawl with several, each with a different policy and a different ranking impact. Knowing which is which is half the battle.

• OAI-SearchBot — indexes for ChatGPT Search. Must be allowed if you want to appear as a cited source.
• ChatGPT-User — on-demand fetch when a user asks ChatGPT to browse. Without it you'll never be quoted live.
• GPTBot — OpenAI's training crawler. Blocking it stops your content being used for future model training but does not affect ChatGPT Search citations. Most teams should allow it; the trade-off is yours.
• ClaudeBot, Claude-User, Claude-SearchBot— Anthropic's training, on-demand, and search crawlers respectively. Same logic as the OpenAI trio.
• PerplexityBot — Perplexity's core indexer. Block this and Perplexity cannot cite you.
• Google-Extended — the toggle Google uses for Gemini and AI Overviews. Allow if you want eligibility for AI Overview citations.
• Bingbot — quietly powers ChatGPT Search via Bing's index. Keep it allowed even if you don't care about Bing itself.

Failure mode 1: blocked bots

We've audited sites where Cloudflare's "Super Bot Fight Mode" was silently blocking PerplexityBot, OAI-SearchBot, and Claude-SearchBot — all without the operator realising. Two fixes:

1. Audit your robots.txt per bot. Our AI crawler matrix shows each bot's allow/block state in one view.
2. Check your CDN / WAF rules. Cloudflare, Vercel Firewall, AWS WAF all have bot-protection presets that block non-Google bots by default.

Failure mode 2: weak structured data

AI tools quote pages with clear, machine-readable structure faster than pages with beautiful design and zero schema. The four schemas that matter most:

• Organization — tells AI tools who you are. Required for entity attribution.
• FAQPage — Perplexity and Claude pull FAQ answers verbatim.
• BreadcrumbList — gives AI tools context about page hierarchy.
• LocalBusiness (if applicable) — required for "best X in Y city" AI answers.

Failure mode 3: no concise answer blocks

AI tools prefer to quote 30–80 word answers wrapped in clear headings. If your "What is X?" answer is buried in a 4,000-word essay, the AI will skip you and quote your competitor's FAQ snippet instead.

The fix: add an FAQ block to every page that targets a question-style query. Mark it up with FAQPage JSON-LD. Keep each answer 40–80 words and start with the direct answer ("Yes, we offer same-day install in Sacramento.") not the preamble ("Great question — let's dive into the details…").

The 7-step checklist

1. Audit AI crawler accessibility (per-bot allow/block matrix)
2. Allow OAI-SearchBot, ClaudeBot/SearchBot/User, PerplexityBot, Google-Extended, Bingbot
3. Add Organization + FAQPage + BreadcrumbList schema to your homepage
4. Publish a spec-compliant /llms.txt at root
5. Add an FAQ block (5+ Q&A) to every page that targets question queries
6. Set 10–25 target AI prompts and start monitoring weekly
7. Build a citation-worthy About page + Contact page + Trust signals

Each of these is a low-effort, high-impact action. Most teams ship the whole list in a single sprint and start seeing citations within 30 days.