How to Rank in Gemini & Claude

The one thing that changes everything: different backends

Before any tactic, internalize this: Gemini and Claude do not read the same web the same way. Gemini grounds its answers in **Google Search** — it uses "Grounding with Google Search," and Google's AI Overviews and AI Mode pull live web sources. Claude is different. When Claude searches the web, the backend is **Brave Search**, not Google or Bing.

This isn't a guess. TechCrunch reported (March 21, 2025) that Brave was added to Anthropic's subprocessor list on March 19, 2025, that Claude's search function contains a `BraveSearchParams` parameter, and that Claude and Brave returned identical citations in testing. One independent test put the citation overlap with Brave's top organic results around **86.7%**. Translation: the work you do to rank in Brave Search is the work that gets you cited in Claude.

Allow the right crawlers — or you're invisible by default

This is the most common own-goal in AI SEO. People block AI bots in `robots.txt` to "protect their content," then wonder why they never get cited. Blocking the search/fetch agents removes you from citation eligibility. Here's exactly which AI crawlers matter for each engine.

Gemini: Google-Extended (and the rank myth)

There is no public "Gemini bot." Crawling happens with existing Google user agents — `Googlebot` powers Search, Discover and Images; `GoogleOther` is a generic R&D fetcher. The control that governs Gemini is **`Google-Extended`**, and Google is explicit that it "doesn't have a separate HTTP request user agent string" — it's a robots.txt token used in a control capacity. It decides whether your content is used to train Gemini and for grounding in Gemini Apps / Vertex AI.

Claude: three crawlers, current official names

Anthropic runs three crawlers. Note that the older `anthropic-ai` and `claude-web` agents from 2024 coverage are deprecated — use these three:

• **`ClaudeBot`** — collects public web content that may train Anthropic's models. Blocking it only affects future *training* inclusion.
• **`Claude-User`** — fires when a real person asks Claude to fetch or visit a page (user-initiated browsing).
• **`Claude-SearchBot`** — navigates the web to improve Claude's search-result quality and relevance.

Per Anthropic's guidance (via Search Engine Land, Feb 2026), blocking **Claude-User** or **Claude-SearchBot** reduces your visibility in Claude's search answers. So to be cited, allow both. One more practical note: Anthropic's bots run on cloud-provider IPs, so don't try to block or allow by IP. To *verify* that traffic is genuinely Anthropic, check the published crawler IP list at `claude.com/crawling/bots.json` — a source IP on that list confirms it's real (and exposes spoofers pretending to be Claude).

How to rank in Gemini: own your domain, then go wide

Gemini has one tell that sets it apart from every other engine: it trusts brand-owned domains far more. In Yext's analysis of 6.8M citations, **52.15% of Gemini citations came from the brand's own website** — compared with ChatGPT, where 48.73% of citations came from third-party sites like Yelp and TripAdvisor. Gemini wants structured, factual content on *your* domain.

But don't mistake "rank #1 and you're in" for the strategy. After the Gemini 3 rollout, an SE Ranking study found only about **19% of AI Overview sources overlapped with the organic top 10** — over 60% of queries showed ≤20% overlap. (A separate Moz figure put 88% of AI Mode citations on pages outside the organic SERP.) Ranking well is the floor. Getting cited is its own layer.

The Gemini playbook

• **Invest in your own domain first.** Ship complete **Organization schema**, build out **local landing pages**, keep a complete and consistent **Google Business Profile**, and use consistent subdomains. This is where Gemini looks first.
• **Keep Google-Extended allowed** and maintain solid technical SEO — mobile-friendly, fast (an LCP under ~2.5s helps crawl and comprehension). Snippet-eligibility on Google is the entry ticket.
• **Don't rely on organic rank alone.** With ~19% overlap, you also need presence in the sources Gemini actually pulls — pursue YouTube content and high-authority third-party mentions, not just your own blog.
• **Ship structured, factual, schema-rich pages** (Article, FAQPage, HowTo, Organization) with visible "Last updated" dates. Gemini favors structured, factual content it can lift cleanly.

How to rank in Claude: win Brave, write deep, play the long game

Claude rewards different behavior than Gemini, and treating it like ChatGPT is a classic mistake — wrong backend (Brave, not Bing), different recency window, and a bias toward niche over mainstream. Here's what actually moves Claude.

Optimize for Brave Search

Because Claude searches through Brave, Brave Search visibility is your highest-leverage Claude lever. That means the fundamentals Brave rewards — clean technical SEO, genuine topical authority, real inbound references from other sites — translate directly into Claude citations. If you're invisible in Brave, you're largely invisible in Claude.

Write deep, balanced, durable content

Claude favors nuanced, well-sourced material — it likes content that covers limitations and risks alongside benefits, not one-sided hype. Thin 500-word posts rarely get cited. And Claude plays a longer game than ChatGPT: per Muck Rack's analysis, its citation sweet spot is roughly the **past 10 weeks** (vs ChatGPT's sharp drop-off after about a week), and it's about **3x more likely than ChatGPT to cite content from 2–4 weeks ago**.

Claude also leans on smaller voices. Its top-100 cited outlets have roughly **50% lower average monthly visitors** than ChatGPT's top-100 — it leans on trade and specialist media, citing outlets like NPR, Yahoo Finance, Variety and CNN at least 2x as often as ChatGPT or Gemini do. If you publish in a niche, that's an advantage here, not a handicap.

The Claude playbook

• **Optimize for Brave Search** — the backend that decides Claude's citations.
• **Allow Claude-User and Claude-SearchBot** (and ideally ClaudeBot), and verify legit Anthropic traffic via `claude.com/crawling/bots.json`.
• **Write long, comprehensive, balanced pieces** that cover the downsides too — Claude favors well-rounded, well-cited material over thin posts.
• **Play the long game** — durable, evergreen content fits Claude's ~10-week window better than news-cycle chasing. Niche and trade outlets carry real weight.

What works for both: quotable, structured, corroborated

Most "GEO advice" is folklore. The strongest evidence we have is the peer-reviewed GEO research paper (Princeton, KDD 2024), which tested 9 tactics across ~10,000 queries and measured the visibility lift inside generative answers. Overall, GEO methods lifted visibility up to **~40%**. These are the moves that won — and they help Gemini and Claude alike.

1. 1**Add direct quotations** from relevant, authoritative sources — one of the top performers (~30–40% lift on position-adjusted word count).
2. 2**Add statistics and concrete numbers** — comparable lift, and combined with fluency it was the single strongest pairing tested.
3. 3**Cite your sources inline** — pages that show their work get pulled into answers more. This disproportionately helped lower-ranked sites: citing sources lifted visibility **115% for a site ranked #5**.
4. 4**Write fluently and clearly** (~15–30% lift). Clean prose is easier for a model to extract cleanly.
5. 5**Use an authoritative voice** (~10–20%) — more modest, but it compounds with the above.

Shared operational checklist

• **Structure for extraction** — clear H2/H3, short direct-answer paragraphs, tables, lists and FAQ blocks let a synthesizer lift a clean passage.
• **Ship JSON-LD schema** (Article, FAQPage, HowTo, Organization) and visible "Last updated" dates.
• **Build cross-channel authority** — both engines weight corroboration. Being referenced by other authoritative sites raises your odds of being cited in either.
• **Keep robots.txt permissive** for the AI fetch/search agents you want citations from. Don't blanket-block.
• **Publish llms.txt if you like** — it's low-effort and may help some tools, but treat it as an emerging experiment, not a confirmed ranking factor for either engine.

Common mistakes that cost you Gemini & Claude citations

Each of these maps to something above — and each quietly keeps brands out of the answers.

• **Blocking the wrong bots.** Blocking Claude-User / Claude-SearchBot or Google-Extended to "protect content" silently removes you from citation eligibility.
• **Blocking Google-Extended to "boost Search."** It has zero effect on Google Search rank — it only affects Gemini grounding and training. Pure downside.
• **Treating Claude like ChatGPT.** Wrong backend (Brave, not Bing), different recency window (~10 weeks vs ~1), and a bias toward niche over mainstream outlets.
• **Assuming organic #1 = an AI citation.** Only ~19% of Gemini AI Overview sources overlap with the organic top 10 — a separate strategy is required.
• **Keyword stuffing.** Proven net-negative in generative engines. Classic keyword-density tactics actively work against you.
• **IP-blocking Anthropic bots.** They use cloud IPs — block by user-agent (and per subdomain) instead, and use the published IP list only to *verify* real traffic.
• **Thin content.** Superficial posts are essentially never cited by Claude. Depth, balance and sources win.

How to measure whether it's working

Rankings are the wrong yardstick here — being inside the answer is the win, and Gemini and Claude pull from different places, so you have to watch both directly. Two complementary signals tell the truth.

1. 1

   Track citations and share of voice per engine

   Run a fixed set of category questions against Gemini and Claude on a schedule, and record whether you're named, how prominently, and how you stack up against competitors. Because Gemini 3 showed how fast citations churn, the trend matters more than any single snapshot — and you want it split by engine, since the two behave differently.

2. 2

   Watch your first-party AI traffic

   When these engines read or cite your site, real AI crawlers hit your pages and real AI referrals land on them. That server-side data is ground truth — it tells you which pages are actually being consumed by Gemini's Google crawlers and Claude's bots.

3. 3

   Verify the traffic is real

   AI-crawler user agents get spoofed. Confirming a hit genuinely came from a known engine — Anthropic publishes its IP list at claude.com/crawling/bots.json for exactly this — keeps your measurement honest before you act on it.

4. 4

   Change one thing, re-measure

   Add stats to a page, tighten its FAQ, refresh its data, fix a blocked crawler — then re-run the prompt set. AI answers are non-deterministic, so the only reliable read is a controlled before/after.

This is exactly what SourceWatch is built for: it measures whether ChatGPT, Perplexity, Gemini and Claude cite your brand — your AI visibility and share of voice against competitors — and it captures the real, verified-vs-spoofed AI-crawler and AI-referral traffic landing on your site. There's also an MCP server, so you can pull all of it straight into Claude Code while you work.