How to get cited by Claude: the 2026 playbook

Claude is the engine most people underestimate. Anthropic doesn't run a flashy consumer search product the way Perplexity does, doesn't have ChatGPT's brand pull, and rarely shows up in casual "I asked an AI" anecdotes outside of Anthropic's own X feed. But by mid-2026, Claude Sonnet 4.5 and Opus 4.7 are inside more enterprise stacks than any competitor's foundation model — every major Notion alternative, every developer-tools incumbent, half of Cursor's installed base. Claude Search rolled out broadly in Q1 and now answers questions inside Claude.ai itself, with inline citations. If your buyer is a builder or a knowledge worker, Claude is touching their workflow far more than search-volume estimates suggest.

The catch: Claude doesn't pick sources the way ChatGPT or Perplexity do. Its constitutional training pushes it toward sources it can defend, not sources it can rank. The mechanics this post unpacks are not the same patterns that win you a Perplexity citation. If you've already read our Perplexity citations playbook and our breakdown of why ChatGPT doesn't recommend your brand, file this one as the third side of the triangle.

"Claude SEO" or AEO? Same practice, different name

Some teams call this "Claude SEO" or "Anthropic SEO" or "AI search optimization for Claude". We use AEO (Answer Engine Optimization) as the umbrella because Claude is one of six engines we track, and the foundational signals overlap. But if you arrived searching for "Claude SEO" — yes, this is exactly the playbook you came for. Claude has unique citation behavior that doesn't transfer cleanly from ChatGPT or Google, which is why a Claude-specific take exists.

How Claude decides who to cite

Three things make Claude's citation behavior distinctive:

1. It over-weights authoritative, established sources. Wikipedia, government domains, academic publishers, mid-tier journalism (TechCrunch, The Verge, niche industry trades). Less so the brand-owned blog, even when the brand blog is technically correct.
2. It under-weights promotional language. Anthropic's RLHF pipeline trained Claude to flag hype and rhetorical claims. Pages with "the #1 best", "revolutionary", or stuffing-style keyword density get implicitly demoted in the re-rank stage, even when they're the canonical resource.
3. It rewards specificity over breadth. Claude prefers a page that gives a precise, narrow, well-sourced answer to a page that lists 47 generic best practices. This is the opposite of how Google's classic SEO playbook taught content teams to write.

Put together, you get a model that cites the boring authoritative source before the exciting marketing page. For most marketing teams this feels backward — and they keep losing citations to sources they consider less qualified.

The 5 signals Claude actually weights

We've reverse-engineered hundreds of Claude citations across SaaS, ecommerce, and B2B verticals. Five patterns repeat:

1. Content depth + specificity

Claude prefers 1,200-word focused answers to 4,000-word omnibus posts. The "what is X" article that gives a single tight definition + 3 illustrative examples beats the "ultimate guide to X" with 17 sub-sections. The retrieval pass scores chunks against the user's query, and chunks pulled from focused pages score higher per-token.

2. Structured data that maps to the claim

Claude reads JSON-LD when it's there. A FAQPage schema with the actual user question as Question.name and a 50-word answer as Answer.text gets pulled into context heavily. So does Article schema with mainEntityOfPage + author set. You can generate clean schema in 30 seconds with our schema generator — and FixAEO's own pages all emit the right types as a dogfood test.

3. Author E-E-A-T signals

Claude over-weights pages with named authors, especially when those authors have on-domain bios with real credentials. "By Jane Smith, Director of SEO at Acme, ex-Moz" anchored to a real /team/jane-smith/ page that has a Person schema, a LinkedIn rel-author link, and prior work — that whole graph signals trust to Claude in a way an anonymous "FixAEO Team" byline does not.

4. Wikipedia / Wikidata entity presence

If your company has a Wikipedia page or a Wikidata QID, you're in Claude's training data more deeply than you realize. Foundation models are pre-trained on Wikipedia at extreme weight. Anthropic doesn't disclose ratios, but multiple model-card hints suggest Wikipedia gets ~5-10× the per-token weight of typical web crawl during training. Translation: a Wikipedia entry is one of the highest-leverage AEO investments any brand can make.

5. Third-party validation

For SaaS: G2 and Capterra reviews. For local: Yelp + Google Business Profile. For consumer products: Wirecutter, RTINGS, Consumer Reports. Claude's re-ranker reads these as orthogonal trust signals — independent sources confirming the brand exists and is taken seriously. A G2 profile with 50+ verified reviews moves Claude citations on commercial queries more than a 5,000-word product page does.

The 6 tactics that actually move Claude citations

These are ranked by leverage per hour invested, not by how much the AEO industry talks about them.

Tactic 1 — Ship a clean Organization JSON-LD + claim your Wikidata entry

Twenty minutes of work, six months of payoff. Your Organization schema needs name, url, logo, sameAs (linking to LinkedIn, X, Crunchbase, Wikipedia), and description. Then go to wikidata.org and either find or create your entity (it's free and lightweight — way easier than getting a Wikipedia page approved). Once your QID exists, link to it from sameAs. Claude resolves these graphs at retrieval time and uses them to disambiguate "Acme" the company from "Acme" the road runner brand.

Tactic 2 — Publish first-person research or original data

Claude's reranker over-indexes on sources nobody else has. A blog post titled "We analyzed 10,000 AI citations and found 3 patterns" lands in Claude's context every time the topic comes up, even when the blog isn't a top-100 domain by DR. Original data — even modest amounts — beats restating what everyone else has said. If you have a free tool that generates data (FixAEO's leaderboard is exactly this — every scan generates a brand-specific data point), publish the aggregate findings.

Tactic 3 — Earn mid-tier publication coverage

Claude trusts TechCrunch, The Verge, Ars Technica, The Information, plus deep-niche trades (Marketing Brew for marketing, The Pragmatic Engineer for dev, etc.) more than it trusts content marketing blogs. One TechCrunch mention earns more Claude citations than 20 self-published posts. The PR motion: pitch original data (see tactic 2), respond to journalist queries on Qwoted or Help A B2B Writer, sponsor a niche industry newsletter, or build something genuinely noteworthy and let it speak.

Tactic 4 — Match Claude's query patterns

Claude users ask longer, more nuanced questions than ChatGPT users. Look at any sample of Claude queries: they average 18-25 words versus ChatGPT's 8-12. That changes your content strategy. Title questions like "What's the best CRM for a 10-person SaaS startup with a $30k/yr software budget that integrates with Slack?" — yes, that long — get retrieved on the long-tail variant queries Claude is actually fielding. Long-tail conversational keywords have been an AEO bet since 2024; for Claude specifically they're the bet.

Tactic 5 — Cut the marketing voice

Audit your top-10 indexed pages. Count instances of: "industry-leading", "best-in-class", "revolutionary", "game-changing", "the #1". Claude's reranker implicitly down-weights pages dense with these. Rewrite into specific, falsifiable claims. "Industry-leading email deliverability" → "97.3% inbox placement on the Litmus seed list, audited Q4 2025." The second version cites. The first doesn't.

Tactic 6 — Ship a working llms.txt

Claude's web crawler (ClaudeBot) reads /llms.txt if it's there. Most sites don't have one. Having one with a curated overview of your most-cited pages effectively tells Claude "start here." Use our llms.txt generator — it produces a spec-compliant file in under a minute and gives Claude a clean entry point. Pair it with explicit ClaudeBot / anthropic-ai allows in your robots.txt.

What NOT to do (and what we see most teams doing wrong)

Three anti-patterns crater Claude citations faster than anything else:

• Keyword-stuffed meta descriptions and H1s. Claude's training implicitly modeled keyword spam as a low-quality signal. Pages where the H1 reads like "Best CRM Software Tools Platform Solution for Startups 2026" get retrieved less often than pages with conversational, specific H1s.
• Synthetic reviews on G2 / Trustpilot. Claude reads third-party signals AND cross-references them. A G2 page with 100 5-star reviews posted in two weeks signals fraud, not authority. The re-ranker has been documented down-ranking these. (We've seen brands lose citations to worse-quality competitors with fewer but real reviews.)
• AI-generated content with no human edit. Claude can detect its own kind statistically. AI-bulk content gets retrieved at lower rates and re-ranked further down. Original first-person writing — even simple, conversational — beats sophisticated AI-generated content.

The temptation to scale content production with LLMs is real. The actual ROI of doing it badly is negative.

How to verify your work

The closed-loop check: scan your domain across Claude (plus the other five engines) and watch the citation rate change. You can do this manually by asking Claude itself "what's the best [X]?" and reading whether your brand shows up — but it's tedious and quickly drifts.

The cleaner path is to run your domain through our AI visibility checker — it queries Claude alongside ChatGPT, Gemini, Perplexity, Grok, and DeepSeek for a configurable set of prompts in your niche, then scores citation rates, sentiment, and which competitors are eating your share of voice. Re-scan weekly after each AEO change and the deltas tell you which tactics are working.

The full AEO tools catalog covers the other pieces — schema generation, llms.txt, citation source radar, query generation — that you'll want once you have a measurement baseline.

TL;DR

Claude rewards what the AEO industry has been undervaluing: depth, specificity, authoritative third-party validation, structured data, and a working entity graph. It punishes promotional language, synthetic reviews, and AI-generated bulk content. The fastest wins are (1) clean Organization schema + Wikidata, (2) original research or data, and (3) cutting the marketing voice from your highest-trafficked pages.

It's not a glamorous playbook. It's the boring stuff your CMO has been talking about for ten years. Claude is the engine that finally pays for it.

Related per-engine playbooks

The other engines reward different signals — same AEO foundation, different leverage points:

• How to get cited by Gemini — why Google ranking is still the ceiling
• How to get cited by Grok — winning Grok citations through X, not your blog
• How to get cited by DeepSeek — the open-source and Chinese-market engine