How to rank in Claude and get your business seen
Direct answer: You do not buy a Claude rank. To get your business seen in Claude, make the basics undeniable: crawlable NZ answer pages, identical name, address and phone everywhere, a tidy Google Business Profile, real reviews, third-party mentions, and a robots.txt that allows Claude-SearchBot and Claude-User. Then re-run the same customer questions in Claude every month. Nobody can honestly guarantee Claude will name you.
This guide follows the way NZ owners actually search ("how to rank in Claude", "get Claude to recommend my business") and sets out what we check from Hamilton.
How to rank in Claude?
Claude does not publish a Google-style results page you can rank on. In practice, "ranking" means your business is named, or your URL is cited, when someone asks Claude a question with web search or retrieval switched on. You earn that by being the clearest, most consistent public entity in your niche, and then measuring.
When we run a free diagnostic, we do not invent a Claude position. We write 8–12 questions customers actually ask, run them in Claude with search or retrieval where available, and note who is named and which URLs appear. Then we compare that with your site, Google Business Profile and directory listings. The gaps are usually boring: mismatched phone numbers, service pages that never answer the question, or blocked crawlers.
How to do SEO for Claude (NZ)
SEO for Claude is classic technical and content SEO plus answer packaging. It is not a separate paid auction. Work through it in order: fix crawl access and identity, package your answers, earn proof, then measure.
| Layer | What to do (NZ) |
|---|---|
| Technical | Indexable HTML, sensible internal links, robots.txt allows for Claude-SearchBot and Claude-User, and Search Console hygiene for the Google overlap. |
| Entity | Matching name, address and phone (NAP); truthful Google Business Profile categories; one consistent trading name. |
| Content | Answer-first pages with honest GST, towns served and shipping detail. No US tax copy. |
| Proof | Reviews gathered within Google's rules; third-party mentions that repeat the same entity details. |
| Measure | A monthly Claude question set, logging who is named. |
On the page itself, that means question-shaped H2s, a 40–80 word answer that can stand alone directly under each one, and FAQ or Article schema only where the matching answer is visible on the page. There is no special Claude-only schema. A public /llms.txt can summarise your citeable facts and key URLs; it complements robots.txt but does not replace SEO or guarantee mentions.
How do I allow Claude-SearchBot and Claude-User in robots.txt?
Add a separate group for each bot with Allow: /. Anthropic's help article documents three crawlers, and blocking or allowing one does not configure the others: Anthropic Help — Does Anthropic crawl data from the web? (updated 2026).
| User-agent | Purpose (per Anthropic) | If you Disallow |
|---|---|---|
| ClaudeBot | Model development / training data collection | Future training collection from your site is limited; search and user fetches are separate |
| Claude-SearchBot | Search indexing / improving search responses | May reduce visibility and accuracy in Claude search results |
| Claude-User | Retrieving your pages when a user asks Claude to | May reduce visibility for user-directed web fetches |
For search visibility, Claude-SearchBot and Claude-User matter most. ClaudeBot (training) is a separate policy decision, and blocking training while allowing the search bots is a valid split. Anthropic says its bots honour robots.txt and will not try to bypass measures such as CAPTCHAs, and notes that IP blocking may not reliably keep an opt-out in place and can stop the bots reading robots.txt, so prefer robots.txt. If you need to rate-limit, Anthropic supports the non-standard Crawl-delay extension (for example, Crawl-delay: 1 under ClaudeBot).
An illustrative allow block has one group per bot, each a User-agent line followed by its rule, then your sitemap line:
| User-agent line | Rule | Note |
|---|---|---|
| User-agent: Claude-SearchBot | Allow: / | Search indexing |
| User-agent: Claude-User | Allow: / | User-initiated fetches |
| User-agent: ClaudeBot | Allow: / | Training: decide deliberately with the owner |
| Sitemap: https://example.co.nz/sitemap.xml | Replace with your own sitemap |
Deploy it, open /robots.txt on the live site to confirm, and repeat for every subdomain you care about. Older strings such as Claude-Web or anthropic-ai may still sit in legacy files; update to the current documented names, because the old ones may not control today's crawlers. Watch for Shopify, Webflow or Cloudflare security apps, password gates and copied "block AI" snippets that insert Disallows owners forget, and for reverse proxies challenging legitimate crawlers into failure.
For reference, our own /robots.txt allows Claude-SearchBot, Claude-User and ClaudeBot (plus legacy Claude-Web / anthropic-ai lines for older tooling), along with OAI-SearchBot, ChatGPT-User, GPTBot, Bingbot and Googlebot.
How to get your business seen in Claude answers?
Be fetchable and quoteable. Put your answers in real HTML text, not locked inside images or PDF brochures. Lead with answer-first paragraphs under clear headings, and state the NZ facts a customer needs: GST labelling, the towns you serve, honest shipping. Assistants prefer pages they can fetch and sentences they can lift without inventing NZ facts.
Hamilton owners often lose to a plainer competitor whose page says "We cover Chartwell, Rototuna and Cambridge" while theirs says "excellence in service".
How to get Claude to recommend my business?
There is no form or official directory for submitting your NZ business to Claude. Recommendations follow businesses that look complete and consistent across the open web. Fix identity first, then publish answers, then earn proof off your site:
- Identity: one trading name, phone and address everywhere: site, Google Business Profile, directories, email signature.
- Answers: pages that state who you serve, where in NZ, GST-honest pricing and the next step, within the first 40–80 words under the heading.
- Proof: reviews within Google's rules, with tidy replies, plus supplier, stockist or chamber mentions that repeat the same entity details.
- Access: Claude-SearchBot and Claude-User allowed in robots.txt.
- Measure: 8–12 fixed questions; log who Claude names each month.
If you sell products, keep Merchant Center availability aligned with Shopify. Shopping-style answers fail when feed data is wrong, even if the brand page looks fine.
How to get mentioned in Claude AI?
In practice, mentioned and recommended both mean your brand is named in the answer; cited usually means a URL or source is shown. The work to earn all three is the same citeability stack. For mentions, make sure the first screen of each key page covers who you serve, where, and what happens next.
How to get cited by Claude AI?
Citations favour sources that resolve cleanly and state facts without fluff. On top of everything above:
- Confirm key pages return HTTP 200 and money URLs have no accidental noindex.
- Check nothing useful sits behind a soft gate or login for anonymous visitors.
- Put the answer in visible text, not only in images or PDFs.
- If you use schema, make sure it matches the visible content.
- Log which URLs Claude shows for your fixed prompts, not just whether you were named.
Google works in a parallel way for its AI features: pages need to be indexed and snippet-eligible, with no special AI schema required (Google Search Central — AI features).
Claude vs ChatGPT vs Google AI Overviews
Foundations overlap (clear pages, entity trust, crawl access); ranking systems do not. Claude is not a copy of Google's top ten, and ChatGPT search uses its own crawler (OAI-SearchBot). Track each tool your customers actually use. See how to rank in ChatGPT, Google AI Overviews (NZ) and our plain-English GEO guide for the side-by-side.
What we check in a Backroom Ops diagnostic
We check NAP across your site, Google Business Profile and directories; whether the first screen of each service or product page answers the query with NZ facts; robots.txt for Claude-SearchBot, Claude-User, ClaudeBot and OAI-SearchBot; reviews and reply hygiene; and, for Shopify, Merchant Center disapprovals, title and availability mismatches, and checkout and app bloat. The output includes a dated log of Claude, ChatGPT and AI Overview answers for your question set, with the URLs shown: a trend over months, not vanity screenshots.
Every change that can hit money, stock or live copy stays behind your approval. That is how we run NZ$790/month store ops (ex GST) and how we implement SEO fixes when you choose done-for-you.
Practical checklist (this month)
- Write 8–12 questions customers ask, including Hamilton–Waikato variants if you are local.
- Download your live robots.txt, remove accidental blanket AI blocks, and allow Claude-SearchBot and Claude-User. Decide ClaudeBot in writing.
- Align trading name, phone and address; fix Google Business Profile categories and hours.
- Publish or rewrite 3–5 money pages with answer-first NZ copy, GST and logistics where relevant.
- Add FAQ schema only for real on-page FAQs; publish or refresh /llms.txt facts.
- Request reviews properly. Never fake them.
- Note the date of each change in your log, and re-run the question set in Claude in 30 days.
What not to do
- Pay for "guaranteed Claude ranking" or "Claude recommendation packages".
- Keyword stuffing, doorway pages, or cloaking (showing bots different content).
- Fake reviews, or review gating that breaks Google's rules.
- Fake "as cited by Claude" badges or "Claude recommends us" claims on your own site.
- US generic FAQ copy that gets NZ tax, shipping or company structure wrong.
- Treating Ahrefs Domain Rating as a Claude rank or citation score.
Illustrative examples
Illustrative examples based on patterns we see with NZ small businesses. Details generalised; not specific client results.
A Waikato service business asked Claude (with web access) for "a good [trade] near Hamilton". Two competitors were named; they were not. The competitors had matching NAP, a service page covering process and GST-inclusive price ranges, and recent Google reviews. The missing business had a polished homepage, almost no service text, and a robots.txt that broadly blocked unknown bots. We proposed aligning listings, deliberately allowing Claude-SearchBot and Claude-User, publishing five answer pages and re-checking the same questions monthly. The owner approved the order; we executed behind that gate. No guarantee of being named, just a clear path to becoming nameable.
Two Cambridge businesses offered the same trade, and only one was mentioned in Claude. That one had a service page answering timeline, GST-inclusive ranges and suburbs served. The other had a homepage slider and a PDF brochure. After the owner approved HTML answer pages and crawler allows, we logged the prompts monthly: clearer odds, still no guarantee.
A nationwide NZ SaaS company asked why Claude never linked its docs. The docs sat behind a soft gate for anonymous users, and robots.txt carried a broad Disallow copied from a "block AI" snippet. Once the owner approved opening the public answer pages and allowing Claude-SearchBot and Claude-User, fetch tests succeeded. Citations were not promised; the blockers were removed.
FAQ
How do I rank in Claude?
There is no official Claude page one. Ranking means your business is named or your URL is cited when someone asks Claude with web search or retrieval. Make yourself crawlable and consistent, then measure with a fixed question set.
What is the difference between ClaudeBot, Claude-SearchBot and Claude-User?
Per Anthropic, ClaudeBot collects training data, Claude-SearchBot indexes for search, and Claude-User fetches pages when a user asks. Allowing or blocking one does not configure the others.
Should I block ClaudeBot but allow the search bots?
That is a valid policy split. For visibility in Claude answers, Claude-SearchBot and Claude-User matter most; training access is a separate choice.
Does allowing the crawlers or adding schema guarantee a citation?
No. Crawler access removes a blocker and accurate schema helps machines parse real on-page content, but neither forces Claude to cite you.
Can Backroom Ops guarantee Claude will name me?
No. We improve citeability and measure with a fixed question set. Anthropic decides what Claude returns.
Next step
If you want this done with you approving every change, book a free diagnostic. We show what search and assistants currently say in your niche, then give you a staged list of fixes. Nothing goes live until you approve it, and store ops stay month to month.
Glossary (Ahrefs, Search Console, Merchant Center): /#glossary. About: /about/.