Claude is the stingiest of the major AI assistants when it comes to citing sources — and that changes the game for getting your brand into its answers. Where Perplexity links out in practically every answer, Claude puts a link in roughly one in four. Being cited by ChatGPT tells you almost nothing about whether Claude will cite you either: when both assistants answered the same question with sources, they shared at least one cited site in only about one search in five.
That is the headline from a Sayu study of 360 everyday UK searches run in October 2026, and it makes Claude a different citation problem from the other engines. Here is how its citation behavior works and what makes content citation-worthy for it specifically.
Key Takeaways
- Claude linked to at least one website in only 26.9% of answers (97 of 360), versus Perplexity at 100%, ChatGPT at 69.4%, and Gemini at 53.6% (Sayu, October 2026).
- When Claude does link, it offers around 4 sites per answer — a wider field than ChatGPT’s ~3.1 but far fewer opportunities overall.
- ChatGPT and Claude shared a cited site in only ~1 in 5 same-question answers, so a citation in one buys you nothing in the other.
- With Claude, being named in the answer text can matter as much as the link — and answer-first, self-contained content is what earns both.
Why Does Claude Cite So Rarely?
Anthropic’s design philosophy shows in the numbers. Claude tends to answer from its training knowledge and only reach for the web when a question clearly needs fresh or verifiable information. That is the opposite of Perplexity’s search-first posture, which is why the two sit at opposite ends of the table: Perplexity linked out in all 360 answers with about 7.8 sites each, while Claude linked in 97 answers with about 4 sites each.
The practical consequence is simple and easily missed: there are far fewer citation slots in Claude. Across the Sayu study’s 360 searches, Claude produced 391 cited sites against Perplexity’s 2,800. Fewer slots means each one is worth more — and harder to win.
Which Questions Make Claude Reach for Sources?
Claude is most likely to cite on questions where its training knowledge is insufficient or where accuracy matters enough to check. The Sayu study’s twelve topics — shopping, health, money, travel, motoring, legal, and others — span exactly this mix of everyday decisions where freshness helps.
Think about it from Claude’s perspective: definitional and conceptual questions (“what is entity SEO?”) can often be answered from training data, while anything time-sensitive, product-specific, or comparison-based (“best project management tools for agencies”) triggers web lookup. If your content answers questions that can sit in training data forever, Claude may use the knowledge without ever reaching for your page — and no link will appear.
There is one methodological caveat worth knowing. In the Sayu study, Claude was asked through its developer interface (Haiku 4.5) with web search switched on, which can behave differently from the consumer app most people use. The pattern — stingier than every rival — is what matters, not the exact percentages.
How Do Claude’s Cited Sources Differ from ChatGPT’s?
They barely overlap. When ChatGPT and Claude both answered the same search with sources, they shared at least one cited site in only 19.8% of cases (19 of 96 searches). The closest pairing among the six measured, Claude and Perplexity, still shared a site on only a little over half the searches both answered.
This kills the common assumption that one citation strategy covers every assistant. Each model appears to keep its own source preferences — its own little black book. The Sayu analysts put it plainly: AI visibility is something you check and work on assistant by assistant, not something you simply hope for.
What Does Citation-Worthy Content Look Like for Claude?
Claude’s citation behavior rewards a specific content profile. Write for it like this:
Answer the question outright, early. A page that gives a clear, complete answer to a question in its first paragraphs gives Claude something worth linking to. Claude cites pages that resolve the question, not pages that bury the answer under three screens of buildup.
Be the verifiable source. Claude reaches for the web when it wants to verify or freshen up. Pages with clear authorship, dates, primary sources, and checkable facts give it reason to cite you rather than answer from training data. First-hand experience and demonstrated expertise — the E-E-A-T signals — are increasingly what separates cited pages from uncited ones.
Structure for extraction. Claude pulls citations from content it can quote cleanly. Short paragraphs, descriptive subheadings, lists, tables, and definition-style passages all make your content easier to lift into an answer — and easier to attribute.
Cover the specific, not just the general. Broad explainer content competes with Claude’s training knowledge. Specific, current, opinionated, data-backed content — comparisons, original numbers, tested recommendations — is what a model cannot generate from its own memory and must fetch.
Don’t neglect unlinked mentions. Because Claude links out so rarely, being named in the answer text can carry as much value as the link. A clear, distinctive brand name and quotable phrasing travel further in Claude’s answers than they do in engines where the link is the unit of value.
How Should You Check Your Claude Visibility?
Run the Sayu playbook as a monthly routine, Claude edition:
- List the everyday questions your customers actually type, in their own words.
- Ask each one in Claude, word for word, and note whether it links out — and to whom.
- Ask the same questions in ChatGPT, Gemini, and Perplexity and mark where your site appears.
- Compare: where Claude cites competitors but not you, study what those pages do that yours doesn’t.
Expect volatility. AI answers change between askings, and a single round of questions is a snapshot. What you are tracking is the trend — whether your pages are entering or leaving Claude’s cited set over time.
People Also Ask
Does Claude cite websites?
Yes, but sparingly. In a Sayu study of 360 UK searches (October 2026), Claude linked to at least one site in 26.9% of answers — 97 of 360 — offering about 4 sites when it did, versus Perplexity at 100% with about 7.8.
Why does Claude cite less than Perplexity or ChatGPT?
Claude is designed to answer from its training knowledge and reach for the web mainly when a question needs fresh or verifiable information, while Perplexity is search-first. In the same study, ChatGPT linked out in 69.4% of answers and Gemini in 53.6%.
Is ranking on Google enough to get cited by Claude?
No. About 9 in 10 sites Claude cited in the Sayu study fell outside Google’s top 10 for the same search — and Claude cited from its own source preferences regardless. A Google ranking is a head start, not a guarantee.
How often should I check my Claude citations?
Monthly is a sensible cadence. AI answers vary between askings, so track whether your pages are entering or leaving Claude’s cited set over time rather than reacting to a single check.
Do the same tactics that work for ChatGPT work for Claude?
Only partly. The two assistants shared a cited site in only about 1 in 5 same-question answers, so treat each as its own citation problem and verify Claude separately.
