May 15, 2026 · 5 min read

How to check your website's rank in Claude and Gemini (not just ChatGPT)

ChatGPT, Claude, and Gemini each use a different search backend — Bing, Anthropic's own crawler, and Google. Same prompt, often very different cited sources. Here's how to check your rank in all three.

Almost every guide on "LLM SEO" talks about ChatGPT. That's a problem, because ChatGPT is only one of three major AI search engines — and the one that's least likely to share retrieval results with the others.

If you're optimising for ChatGPT alone, you're optimising for one search backend (Bing) and ignoring the two that index almost everything else on the web. This guide explains why Claude and Gemini behave differently, and how to measure your visibility inside each.

Each provider has its own search backend

When you ask the model "best CRM for early-stage startups", here's what happens under the hood:

ChatGPT (via `web_search` tool)

OpenAI uses a Bing-powered search index. The model issues 1–3 web queries, gets back the top 3–10 results per query, reads the snippets and sometimes the full pages, then composes an answer with url_citation annotations.

To check your rank: call the latest GPT model with tools: [{ type: "web_search" }], parse response.output[].content[].annotations[].

Claude (via `web_search_20250305` tool)

Anthropic operates its own web crawler and search index. Add the web_search server tool to your messages.create call and Claude will issue 1–5 queries (capped via max_uses), read the results, and return a web_search_tool_result block plus citations attached to the text response.

Anthropic's crawler is newer and less complete than Bing or Google. It tends to favour:

Recently-published content
Well-structured documentation sites
Pages with strong canonical signals (proper <link rel="canonical">)
Sites that haven't blocked the Claude-User user agent in robots.txt

That last point bites a lot of sites. If your robots.txt has a blanket "block AI bots" rule, you've removed yourself from Claude's index. Check yours with our scanner.

Gemini (via `googleSearch` grounding)

Google Gemini uses Google Search itself through the grounding tool. When you enable googleSearch, the model issues queries against Google's main index and returns groundingChunks with the cited URLs.

Quirk: Gemini's web.uri field is a Vertex AI proxy URL (vertexaisearch.cloud.google.com/grounding-api-redirect/...). The actual source domain is in web.title. A naive parser will think every Gemini citation is from vertexaisearch.cloud.google.com — it isn't.

Gemini is the closest thing to "track my Google rank but for the AI answer". If your classic SEO is healthy, Gemini citations tend to follow.

Why your rank differs across providers

Same prompt, three providers, three retrieval pipelines, three answer-composition models. The variance is real:

A SaaS product page might be at rank #1 in ChatGPT (strong Bing presence) and not appear in Claude (small site, not yet in Anthropic's crawl).
A new blog post might appear in Gemini within hours (Google crawls fast) but take days to show up in ChatGPT or Claude.
An old, well-cited research article tends to appear in all three at similar positions, because authority signals carry across backends.

This is why a single-provider rank tracker is misleading. A Pro-tier AI Rank Checker check queries all three in parallel, returns per-provider ranks, and shows you which sources are appearing where.

How to check Claude rank specifically

The Anthropic SDK call looks roughly like:

import Anthropic from "@anthropic-ai/sdk";
const c = new Anthropic();
const response = await c.messages.create({
  model: "<latest-claude-model>", // check Anthropic's docs for the current model ID
  max_tokens: 2048,
  messages: [{ role: "user", content: "your prompt" }],
  tools: [{ type: "web_search_20250305", name: "web_search", max_uses: 3 }],
});

The response contains text blocks with citations[] and (optionally) web_search_tool_result blocks listing every URL Claude saw. Iterate both, dedupe by host, and you have your cited-source list. The position of your domain in that list is your "Claude rank" for the prompt.

How to check Gemini rank specifically

Using @google/genai:

import { GoogleGenAI } from "@google/genai";
const ai = new GoogleGenAI({});
const response = await ai.models.generateContent({
  model: "<latest-gemini-flash-model>", // check Google's docs for the current model ID
  contents: "your prompt",
  config: { tools: [{ googleSearch: {} }] },
});

The grounding data lives at response.candidates[0].groundingMetadata.groundingChunks[]. Each chunk has web.uri (the proxy URL) and web.title (the real source domain). Use web.title as the host identifier and you get a clean per-source ranking.

The faster way: AI Rank Checker

You can wire all three SDKs yourself — they're documented in OpenAI, Anthropic, and Google's docs. Or you can drop your prompt + domain into AI Rank Checker and get a side-by-side comparison in 25 seconds:

Per-LLM rank cards showing your position in each provider's cited list
Aggregate average rank across the three providers
Sources table sorted by how many LLMs cited each URL — so you can immediately see who's beating you across all three

Pro plan ($79/mo) gives you 75 cross-provider checks per month. Starter ($29/mo) covers ChatGPT only.

What to do with the per-provider data

The most valuable insight is the diff between providers. If you rank #1 in Gemini but don't appear in Claude, the problem is almost always one of:

Robots.txt blocking Claude-User — check with our scanner.
Anthropic hasn't crawled the page yet — submit your sitemap and wait a week.
The page lacks structural signals Claude relies on — semantic HTML, canonical link, structured data.

If you rank in ChatGPT but not Gemini, it's usually the opposite — Google has down-ranked you for some technical SEO issue. Run a classic SEO audit.

Knowing which provider you're invisible in tells you what to fix. That's the whole point of multi-LLM rank tracking.