How to Optimise for Claude AI: The Complete 2026 Guide

Most AEO guides treat all AI engines as roughly equivalent. They are not. Claude, built by Anthropic, has measurably different citation preferences from ChatGPT and Perplexity. The most important difference: Claude uses Brave Search for web retrieval, not Google or Bing. This single fact changes the indexation pathway and shifts the optimisation priorities significantly.

This guide covers the specific signals that drive Claude citation, how they differ from ChatGPT and Perplexity optimisation, and the exact steps to take.

How Claude retrieves web content

Claude uses retrieval-augmented generation (RAG) for queries requiring current information. When Claude searches the web, it goes through Brave Search, not Google Search or Bing. This has a direct implication: your Google ranking has limited influence on whether Claude can find and cite your content. What matters is whether your content is indexed by Brave and whether it scores well on the signals Claude uses to select passages.

Brave Search operates its own independent web index. Sites that are well-indexed on Google but not on Brave may be invisible to Claude. If you have never checked your Brave indexation, open Brave Search and run a site: query for your domain.

Allow ClaudeBot in your robots.txt

Anthropic's crawler is called ClaudeBot. There is also a secondary identifier: anthropic-ai. Both need to be allowed in your robots.txt. Many sites that carefully whitelisted GPTBot during the 2023 AI crawler wave never added ClaudeBot: this is one of the most common reasons a site is cited by ChatGPT but not by Claude.

User-agent: ClaudeBot
Allow: /

User-agent: anthropic-ai
Allow: /

Claude's E-E-A-T weighting: the key differentiator

Of all the major AI engines, Claude places the highest weight on E-E-A-T signals. In CiteRank's per-engine scoring model, E-E-A-T accounts for 28% of Claude's citation likelihood score. This is the highest E-E-A-T weighting of any engine analysed. By comparison, ChatGPT weights E-E-A-T at 18% and Gemini at 15%. If your content has weak author signals and no Organisation schema, you are specifically penalised by Claude relative to the other engines.

• Named author with a visible byline and schema:author markup
• Author bio with specific, verifiable credentials (not "our team of experts")
• sameAs links from author schema to LinkedIn or professional profiles
• About page with founding date, team details, and domain expertise stated explicitly
• Organisation schema with name, URL, logo, and contact point

Content signals Claude favours

Academic and formal citation patterns

Claude was trained on a larger share of academic and long-form research content than most other LLMs. As a result, content that mirrors academic writing conventions performs better: claims supported by cited sources, clear definitions on first use of a term, numbered references or linked footnotes, and formal-but-readable language. This does not mean you need to write a research paper. It means you should link to primary sources (studies, official documentation, government data) when you make factual claims.

Crawlability comes first

Claude's E-E-A-T weighting only applies to content it can actually read. In CiteRank's analysis, crawler access accounts for 22% of Claude's citation score. No amount of author markup or structured data compensates for a blocked ClaudeBot or a slow page that Brave's crawler skips. Fix access before refining content signals.

Schema markup: the second-highest lever

Schema markup accounts for 18% of Claude citation likelihood. FAQPage and Article schema are the most impactful types. For Claude specifically, the Article schema author field carries significant weight. A page with Article schema that includes a named Person author with sameAs URLs gives Claude the information it needs to verify the source before citing it.

What does not work for Claude

Meta title and OG tags carry zero weight in Claude citation scoring. Claude does not use these as citation signals the way some engines do. Spending optimisation time on meta tag A/B testing is unlikely to move Claude citation rates. Focus instead on E-E-A-T, schema, and crawler access.

Testing whether Claude is citing you

The most direct test is manual: open Claude.ai, ask a question in your category, and observe whether your site appears as a cited source. Run 5-10 queries covering your main topics. Note which competitors appear when you do not. Then run a CiteRank audit to identify the specific signal gaps between your site and the cited ones.

Common mistakes that quietly block Claude citation

Most sites that fail to get cited by Claude are not doing anything dramatically wrong. They are making one or two small, correctable mistakes that compound because Claude weights fewer signals more heavily than other engines. The most common ones, in order of how often CiteRank audits find them:

• ClaudeBot allowed at the root but blocked on a CDN, WAF, or bot-management layer that sits in front of robots.txt: check server logs, not just the robots file, to confirm ClaudeBot requests are actually reaching origin
• Author bylines present in the HTML but missing from Article schema: Claude reads the structured data first, so a visible byline with no matching schema:author field is invisible to the citation model
• Organisation schema with a mismatched or outdated sameAs list: stale LinkedIn or Crunchbase links reduce entity confidence rather than adding it
• Content that answers the question only after several paragraphs of scene-setting: Claude favours passages where the direct answer appears in the first sentence or two
• No Brave Search indexation check, ever: teams optimise for Google and assume it transfers, but a page can rank well on Google while being thin or absent on Brave

A monthly Claude citation checklist

Because Claude weights E-E-A-T and crawler access so heavily, and both can drift over time (a CDN rule changes, an author bio gets simplified, a schema field gets dropped during a redesign), a short recurring check catches regressions before they cost months of citation.

1. Confirm ClaudeBot and anthropic-ai are still allowed in robots.txt and are not being blocked upstream by a CDN or WAF rule
2. Spot-check 3-5 high-priority pages for Article, Organisation, and FAQPage schema validity using a structured data testing tool
3. Search your domain on Brave Search directly to confirm recent pages are indexed
4. Ask Claude 5-10 real questions in your category and log whether your domain appears, and in what position relative to competitors
5. Review any pages published or substantially edited in the last month for a named author and up-to-date Article schema