Claude SEO: how to optimise your site for Claude in 2026
Claude SEO is the practice of making your website reachable, legible and quotable for Anthropic's Claude, across both the training data it answers from by default and the live web search it uses in claude.ai, the API and Claude Code. This guide covers every signal that matters, in priority order.
Key takeaway: Anthropic exposes three different web agents with different jobs: Claude-SearchBot for search, Claude-User for user-directed retrieval and ClaudeBot for model training. If Claude search visibility matters, keep the first two available according to your policy and make the public content clear and supportable. ClaudeBot training access is a separate choice, not a search requirement.
What is Claude SEO?
Claude SEO (a branch of GEO, generative engine optimisation) is the process of making your website citable and recommendable by Anthropic’s Claude. It is not traditional SEO with a new name: the crawlers are different, the selection criteria are different, and the surfaces where answers appear are different.
Claude matters more than its chat-window market share suggests. It is one of the most heavily used models among developers, and a large share of its usage never touches claude.ai at all: it reaches people through the Anthropic API, embedded inside other companies’ apps, assistants and internal tools, and through Claude Code on engineers’ machines. Many of those products run their own retrieval and hand the results to Claude to answer from. Optimising for Claude therefore means optimising the fundamentals those products all depend on, not gaming one interface.
How does Claude decide what to recommend?
Claude surfaces a business in two ways, and your playbook has to cover both:
- Training data (the default). Claude answers from what it learned before its knowledge cutoff. A distinct, consistently described, well-referenced brand footprint earns recall here. You cannot edit it on demand; you build it over time.
- Web search and tool use (the live path). When enabled, Claude can search and retrieve current web content. Search/discovery and user-directed fetches are separate from training, and Anthropic does not publish a universal timeline for when a page change will affect a cited answer.
A brand can win one path and lose the other, which is why a Claude-specific audit reports them separately. For the mechanics in depth, see how Claude decides which websites to cite.
The five signals that decide Claude visibility
1. Crawler access
Anthropic documents three agents that honour robots.txt: Claude-SearchBot for search, Claude-User for user-directed retrieval and ClaudeBot for model training. Blocking Claude-SearchBot or Claude-User can reduce live web visibility; blocking ClaudeBot is a training-policy choice and should not be described as removing you from Claude search. SearchScore’s July 2026 interactive audit sample found 6.9% of sites blocked at least one major AI-related crawler or control checked by the audit.
2. Server-rendered, readable content
If your key content only appears after client-side JavaScript runs, crawlers see an empty page. The answer must exist in the HTML that gets fetched.
3. Quotable, answer-first structure
Claude picks the source it can lift a clean, self-contained answer from. Across the 850,000+ sites in the SAVI Report (April 2026 edition), average on-page structure scores just 23.1/100, which is why most sites are readable but never quoted. Lead each page with the direct answer, use question-format headings, and keep one idea per paragraph.
4. Entity clarity
Make the entity easy to resolve with consistent visible naming, clear About information and accurate Organisation, Person or Product markup where appropriate. FAQPage can describe genuine visible FAQs. Structured data can reduce ambiguity, but it does not pin a citation to your brand or guarantee source selection.
5. Recommendation strength
Claude leans toward sources the wider web already trusts: the third-party mentions that shaped how your brand appeared in training data, plus the reviews and references a live web search can surface now. Claude needs a reason in both to name you over the competitor beside you.
The Claude SEO playbook, step by step
Step 1: set the right Anthropic crawler policy
Open yourdomain.com/robots.txt. If you want Claude web-search visibility, confirm Claude-SearchBot and Claude-User are not accidentally disallowed, directly or through a wildcard rule. Decide ClaudeBot separately according to your training policy. Anthropic says all three agents honour robots.txt; also check CDN or security rules that may block legitimate requests.
Step 2: make important pages easy to find and interpret
Keep internal links, sitemaps and key public pages clean and accessible, with the important answer present in the fetched page. If you maintain /llms.txt, treat it as optional agent-readiness housekeeping; Anthropic does not document it as a Claude search-ranking or citation input, and adoption scarcity is not evidence of an advantage.
Step 3: restructure your key pages answer-first (this week)
Take the five to ten pages you most want Claude to cite. Move the direct answer into the first paragraph as one self-contained sentence. Convert section headings into the questions buyers actually ask. Cut vague brand copy from openings. Claude cites text it can extract without rewriting; make extraction trivial.
Step 4: pin your entity (this week)
Add accurate Organisation and Person markup where appropriate and keep core identity details consistent across the site and credible profiles. Use named authors and verifiable credentials when they reflect genuine authorship. Anthropic does not publish a weighting formula that lets SearchScore claim authorship or schema has a fixed citation weight.
Step 5: build the referenced footprint (ongoing)
Earn mentions from reputable industry publications, directories and review platforms. This is the slowest lever but the only one that reaches Claude’s training path: the footprint you build now is what future model versions recall. Consistency matters as much as volume, because scattered, contradictory references fragment your entity.
How is Claude SEO different from ChatGPT SEO?
The fundamentals rhyme, but three differences change your checklist:
- Different agent roles. Claude-SearchBot handles search, Claude-User handles user-directed retrieval and ClaudeBot handles training. Audit each provider separately and do not collapse search eligibility and training permission into one decision.
- No search-index intermediary. ChatGPT’s live browsing leans on Bing’s index, and Gemini grounds on Google. Claude’s web search fetches pages directly, so neither your Bing nor Google presence substitutes for your own site being reachable and quotable.
- Developer surfaces. Claude answers inside APIs, coding tools and embedded assistants at unusual scale. You optimise for all of them at once by fixing the fundamentals, because those products retrieve content and hand it to Claude rather than exposing their own knobs.
For the ChatGPT equivalent of this playbook, see the ChatGPT SEO guide.
How long until Claude visibility improves?
Changes become eligible to matter after the relevant search/retrieval path revisits the content. Anthropic publishes no universal recrawl or citation timetable, and model-memory changes follow separate update cycles. Measure the target questions instead of promising near-instant, weekly or quarterly effects.
How do you measure Claude visibility?
Two layers. For a point-in-time diagnostic, run the free Claude Visibility Checker: it tests each Anthropic crawler’s access separately, scores citable structure and entity clarity, and returns a ranked fix list in about 60 seconds. For movement over time, SearchScore’s Tracker puts real prompts to six live engines including Claude every week and counts exactly how often each one cites you, with a dedicated Claude column.
Frequently asked questions
Is Claude SEO worth it for a non-technical audience?
Yes. Claude’s chat products serve a broad consumer and professional audience, and its API surface means Claude answers questions inside tools your customers already use. The work also overlaps heavily with ChatGPT, Perplexity and Gemini optimisation, so very little of it is single-engine effort.
Can I appear in Claude without a big brand?
Yes, on the live path. Claude’s web search cites the page that answers the question most quotably, and in SearchScore audits small sites with strong signals regularly outscore far larger brands. A small site with clean access, sharp answer-first structure and a consistent entity can be cited ahead of a household name with none of those.
What should I fix first?
Access, always. Check robots.txt for the three Anthropic agents before touching content. Then structure, then entity, then footprint: that is the order in which each layer caps the next. If Claude still never names you after those fixes, work through why your website isn’t showing up in Claude for the full diagnostic.
Related articles
- How Claude decides which websites to cite →
- Why your website isn’t showing up in Claude →
- How to check your Claude visibility →
- How to rank in AI search results →