Claude: the fastest-growing AI referrer, and it cites differently

01Why Claude is on the list now

Growth. In Goodie's brand-averaged B2B panel, Claude rose from 1.4% of AI referrals (May–August 2025) to 18.5% (March–April 2026). SE Ranking, across 101,574 sites, measured Claude's share of web traffic up 386% from January to April 2026, with March alone a 2.6× jump. Previsible saw Claude pass Perplexity in referral sessions in March 2026 and stay ahead.

02Who uses it

Knowledge workers and technical buyers. Goodie describes roughly 30 million consumer monthly users with a heavy skew to professionals, research-stage queries and a high click-out rate relative to platform size. Pierview's 8.7-million-session dataset shows the same B2B tilt.

For an Australian services or B2B brand, that profile is often closer to your actual buyer than the general ChatGPT audience. Small volume, high relevance.

03How Claude searches and cites

Claude runs a web search when a question needs current information and ships citations as part of the answer, each a clickable link. Profound found an 86.7% overlap between Claude's citations and Brave Search's top results, so Brave visibility matters for Claude more than Google rankings do.

Ahrefs' referrer testing found Claude passes referrer data properly, which means your GA4 AI channel will see it. Presenc's review of citation patterns notes Claude favours documentation, technical explainers and primary sources over listicles, and hedges when evidence is thin. A brand that is strong in ChatGPT can still be absent from Claude.

04What to do for Claude

Publish primary material, make it crawlable, and check your Brave presence.

1. 1Allow ClaudeBot. Anthropic documents it; it needs to be allowed in robots.txt alongside the others.
2. 2Write documentation-style pages. Specs, methods, data, definitions, with sources. Claude cites explainers over "top 10" posts.
3. 3Check Brave Search. Search your key questions in Brave; if you are not there, Claude is unlikely to find you.
4. 4Keep facts verifiable. Claude hedges when evidence is thin. Named sources and dates raise your odds.

05How to measure

Claude is the easiest assistant to attribute: add claude.ai to your GA4 AI channel and report it as its own line. Run your prompt list in Claude monthly; the cited domains will differ from ChatGPT's, which tells you where to earn mentions.

Give Claude its own baseline rather than folding it into a blended AI score. Its growth and its audience are different enough to deserve a separate line in the monthly report.

06Why documentation-style content wins in Claude

Claude tends to cite pages that explain how something works and show their evidence: specifications, methods, definitions, data tables, primary sources. "Top 10" posts, which do well in ChatGPT, are cited less. If your best explanatory material is locked in a PDF or a sales deck, publish it as a page.

A practical test: for each of your key topics, do you have a page a careful researcher would bookmark? It should define terms, state figures with sources, describe the process step by step and be dated. That is the page Claude is most likely to cite, and it is also the page a serious buyer wants to read.

Because Claude hedges when evidence is thin, verifiability matters more than on other platforms. Named sources, dates, and a clear author with credentials all reduce the chance Claude qualifies or omits you.

07The Brave Search connection, explained

Claude's web search draws on Brave Search's index; Profound measured an 86.7% overlap between Claude citations and Brave's top results. Brave's index is independent of Google and Bing, so a site can rank well in Google and be weak in Brave.

Check it directly: search your ten key questions in Brave and note whether you appear on page one. If not, the fixes are familiar (crawlable pages, clear titles, specific content, links from relevant sites) but the index is different, so results in Google are not a proxy. Brave also publishes webmaster guidance; make sure your site is not blocking its crawler.

08Claude for Australian B2B and services brands

Claude's audience skews to professionals and technical buyers at the research stage, which is often exactly the person who hires an accountant, a law firm, a software vendor or a consultant. For those brands, Claude can send fewer visits than ChatGPT but better ones.

Because Claude passes referrer data, you can prove it. Add claude.ai to your GA4 AI channel, give it its own row, and watch conversion rate rather than volume. In Goodie's panel Claude was second only to ChatGPT for B2B AI referrals by spring 2026; your own data will tell you whether that holds in your market.

09A Claude-ready page, in practice

Take a service page and add what a researcher needs: a plain definition of the service, the process in numbered steps with timeframes, a price range with what it includes, two or three named people with credentials, a dated "last reviewed" line, and links to the standards or regulations you work under.

That page serves three audiences at once: the buyer doing homework, Google (specific, current, authoritative), and Claude (verifiable, primary, well-structured). It also gives the other assistants accurate facts to repeat, which is the whole point of the official-facts approach I use across SEO, AEO and GEO.

10Three mistakes brands make with Claude

Assuming ChatGPT visibility carries over, hiding the best material in PDFs, and blending Claude into a generic "AI" line so its growth is invisible.

1. 1Assuming carry-over Different index (Brave), different source preferences. Run a Claude-specific baseline.
2. 2Locked-up expertise Whitepapers and decks are not citable. Publish the explanation as a dated web page with sources.
3. 3No separate line Claude passes referrer data; give it its own GA4 row so stakeholders see it.

A fourth, smaller one: forgetting that Claude users are often researching rather than buying today. The pages they land on should make the next step easy (a clear contact path, a related guide) rather than pushing a sale, because the visit is early in the decision.

11Putting Claude into the monthly report

Three lines: Claude referral sessions and conversions from GA4, Claude mentions and citations from the fixed prompt list, and the share of form respondents who chose Claude. Compare quarters; monthly figures are small and noisy.

Goodie's panel shows the share moving quickly, so a brand that set its AI baseline in 2025 may have no Claude line at all. Add one. The growth data says it will matter more next quarter, and because Claude passes referrer data properly, it is one of the few AI platforms where the return can be shown directly in analytics.

12The short version

Claude is small but growing fast, sends trackable visits, and rewards primary sources. Make your best material citable and let ClaudeBot read it.