Claude Optimization

This skill tunes an AEO program for Claude. The cross-engine fundamentals apply; this skill says where to add weight. Behavior evolves — treat these as current emphases and confirm by testing. It works for any business, B2B or B2C.

How Claude builds an answer

Claude answers from trained knowledge (a snapshot with a cutoff) and, when web search is enabled, by retrieving and citing live sources. Relative to other engines it cites conservatively and selectively — it is cautious about asserting things it can't support. That shapes the strategy in two ways:

• From memory: Claude can only describe a brand it has formed a clear, consistent picture of — which comes from being a recognized entity with accurate, corroborated information across the web.
• When searching: it favors accurate, well-sourced, technically precise pages, original data, and clearly credentialed authors, and it shies away from overstated or thin content.

Net: Claude rewards substance and accuracy over recency and community chatter. The way to win is to be genuinely correct, well-sourced, and recognizable — not loud or merely fresh.

When to use this skill

• A brand cares specifically about how Claude describes and cites it.
• Diagnosing why competitors are named by Claude and the brand is not.
• Prioritizing AEO for a Claude-heavy audience.

Before you start

Read aeo-foundation.md for the prompt set, competitors, and entity facts, plus a Claude baseline if available. If it is missing, point to the aeo-foundation skill.

Where Claude puts weight

• Accuracy and trustworthiness. Precise, verifiable, well-sourced content is favored; overclaiming or thin content is skipped. Lean on evidence-and-citations and claim-accuracy.
• Author authority and credentials. Named, credentialed experts and demonstrated expertise carry real weight — prioritize author-authority.
• Original data and depth. Distinctive first-party data and genuine information gain are strongly rewarded; Claude handles substantive, technical material well.
• Entity recognition. A clear, consistent entity helps Claude describe the brand confidently from trained knowledge — prioritize entity-optimization.
• Retrievability for web search. When it searches, it needs reachable, readable, well-structured pages — confirm ai-crawler-access (ClaudeBot, Claude-SearchBot, Claude-User), ai-retrievability, and answer-engineering.
• Less weight on community/recency. Reddit-style signals and aggressive freshness matter less here than on Perplexity — though content still must not be outdated to stay accurate.

Quick wins

• Remove overclaims and unsupported superlatives; back every strong claim with a source or a number (run a pass with claim-accuracy + evidence-and-citations).
• Surface original data the brand owns as quotable, well-sourced passages.
• Add credentialed author bylines and author pages.
• Establish/complete the entity so Claude knows the brand from memory.
• Confirm Claude's crawlers aren't blocked (run the ai-crawler-access checker).

Workflow

1. Baseline in Claude (with and without web search where possible): how it describes the brand, what it cites, who wins, and any inaccuracies.
2. Tighten accuracy and sourcing on target content — specific, verifiable, well-cited claims; remove overstatement.
3. Strengthen author authority — credentialed bylines and demonstrated expertise.
4. Surface original data and depth where the brand has it.
5. Reinforce the entity so Claude knows the brand from training.
6. Confirm retrievability for Claude's web search.
7. Re-test and iterate, correcting any misrepresentations (cadence: see ai-visibility-tracking).

Output format

• A Claude baseline (description + citations) and gap list.
• Prioritized moves weighted to accuracy/sourcing, author authority, original data, and entity strength.
• Target content, authors, and entity facts to act on.
• (assumption) tags where behavior was inferred rather than tested.

Common pitfalls

• Overclaiming or thin content — exactly what a conservative citer skips.
• Anonymous content with no credentialed authority.
• A weak or inconsistent entity, so Claude is unsure who the brand is.
• Forgetting retrievability for Claude's web search.
• Assuming Claude weights Reddit/freshness like Perplexity — it leans on accuracy and authority.

Related Skills

• evidence-and-citations / claim-accuracy — the accuracy and sourcing Claude rewards.
• author-authority — the credentialed expertise it favors.
• entity-optimization — recognition from trained knowledge.
• answer-engineering — extractable structure for when it searches.
• ai-retrievability / ai-crawler-access — reachability for Claude's web search.