The platforms that crawl your AI site the most cite you the most (except one)
We tested 24 queries across four AI platforms. ChatGPT cites Genymotion 83% of the time. Claude, 21%. Perplexity, 17%. Gemini, 4%. The first three track almost perfectly with how much each platform crawls our AI site. Gemini is a different story entirely.
Fresh data from our citation tracking tests on March 2: we ran 24 queries, the same ones on all four platforms, and measured how many times Genymotion were mentioned (“brand mentions”) and how many times the sources linked to their site (“citations”). We then compared the results to the traffic we measure on our crawler logs on the AI site (rozz.genymotion.com).
We’ve been calling this a “mirror site” in previous articles. We’re dropping that. It’s an AI site. It’s built for a different audience: AI agents. The AI site is more accurately compared to a mobile site: same content, different audience, different artefact. We’re not optimizing for citations, we’re optimizing for AI agents.
Key Findings
- ChatGPT: 83% citation rate, up from 14% before the AI site launched. This one works.
- Claude: 21% citation rate. ClaudeBot crawl activity jumped 24x in one week (21 to 505 requests). Growing.
- Perplexity: 17% citation rate. PerplexityBot tripled week over week (14 to 42). Early but on the same curve.
- Gemini: 4% citation rate despite recommending Genymotion in most queries. Different pipeline entirely.
The correlation (and where it breaks)
Here’s what we measured on March 2, against crawl data from the same period.
| Platform | Crawling the AI site? | Weekly crawl volume | Citation rate | Brand mentioned |
|---|---|---|---|---|
| ChatGPT (GPT-5.2) | Heavy, sustained since January | 1,200+/week | 83% (20/24) | 96% (23/24) |
| Claude | Just activated (Feb 28) | 505 this week, was 21 | 21% (5/24) | 33% (8/24) |
| Perplexity (sonar-pro) | Light, growing | 42 this week, was 14 | 17% (4/24) | 25% (6/24) |
| Gemini | Not crawling | 0 | 4% (1/24) | 38% (9/24) |
For ChatGPT, Claude, and Perplexity, you can draw a straight line: more crawling = more citations. The AI site is doing what it’s supposed to do for these three. Gemini is something else because it ignores the AI site entirely.
ChatGPT: 14% to 83%
Before we built the AI site, genymotion.com showed up in roughly 14% of relevant AI queries. Eight weeks later: 83%. ChatGPT cites Genymotion in 20 of 24 use-case queries we tested. Brand mentioned in 23 of 24. Same product, same company, same market, same website. The only thing that changed is we built an AI site with structured content for AI agents.
Here’s the recap: GPTBot’s initial mass crawl (547 requests on January 7, the event that started this whole article series), followed by sustained ChatGPT-User traffic of 1,200+ requests per week. This week it made 1,228 requests, mostly hitting Q&A pages (714), the homepage (307), and GEO content pages (150). It moved from training, to indexing, to citing.
The pages getting cited are consistently on these major topics: pricing plans (22 visits this week), macOS compatibility (13), free version availability (13), Google Play Store setup (13). These are purchase-decision questions, and Genymotion is in that answer 83% of the time.
When ChatGPT recommends Genymotion, it links directly to genymotion.com pages. In most queries, it ranks at position #1. There are commonly multiple citation links per response. This reinforces the legitimacy and makes the ChatGPT answer a warm referral.
The remaining work is expanding content coverage into use cases where we’re weaker. CI/CD and app performance monitoring both came in at 13% citation rate, compared to 63% for app development and manual testing.
Claude: reading the map, not the pages
Two weeks ago in Entry #4, we wrote: “We’ve been waiting for ClaudeBot to come back for three weeks now. It hasn’t.” Well, it did. 505 requests this week, up from 21: that’s 24x.
But the interesting part isn’t the number. It’s what ClaudeBot chose to read.
GPTBot does mass crawls: hundreds of pages in a day, content first, structure optional. ChatGPT-User hits individual Q&A and GEO pages during live conversations—176 Q&A pages in the last four days alone. Both treat the AI site as a pool of content to pull from.
ClaudeBot is doing something different. It’s reading the site’s organizational structure.
The topic sweep
On March 2, between 10:34 and 10:39 AM, ClaudeBot crawled 13 topic pages in sequence. One every 20–30 seconds. That’s nearly the full topic taxonomy: android-versions, hardware-architectures, virtualization-technologies, arm-platform-and-gpu, billing-and-subscriptions, licensing-and-eulas, macos-security-toolkit, documentation-and-support, ci-cd-tools, network-security-toolkit, system-image-and-bios, software-installation-and-trials, root-access-and-tools.
It had done something similar on February 28: a session that mixed topic pages with GEO content pages (enterprise features, cloud platforms, VirtualBox compatibility, Hyper-V troubleshooting), for a total of 8 content pages plus 1 topic page in two minutes.
ClaudeBot is the only bot that systematically crawls topic pages. In the same four-day window, ChatGPT-User hit two topic pages (both pulled into live conversations by user queries). GPTBot, ByteSpider, PerplexityBot, Meta AI: zero topic page hits.
What does “reading the map” actually mean?
Topic pages are the only pages on the AI site with CollectionPage schema. Each one lists every content page and Q&A in that topic, with titles, descriptions, and links. They’re the site’s table of contents: not answers to questions, but a map of what the site knows and how it’s organized.
By crawling all 13 topic pages, ClaudeBot has a complete picture of the site’s knowledge structure. It knows which topics this domain covers, how many pages address each topic and with which keywords, and how individual pages relate to each other.
What we can’t explain is why ClaudeBot read zero Q&A pages. During the same time, ChatGPT-User read 176. The content that drives 75% of ChatGPT citations, ClaudeBot skipped entirely.
Three possible explanations
1. ClaudeBot is evaluating before committing. The pattern: constant monitoring (robots.txt and sitemap checks every 2–3 hours), then a content sample (Feb 28 session, 8 pages), then a structural assessment (Mar 2 topic sweep). This looks like a staged decision pipeline. Understand the site’s scope and quality through its taxonomy, then decide whether to invest deeper crawl budget on individual pages. If that’s right, a bulk content crawl should follow.
2. Anthropic is building a different kind of index. What if Claude’s retrieval doesn’t work like ChatGPT’s “find page, index page, serve page” model? If Anthropic is building something more like a knowledge graph—where the unit of understanding is “what does this domain know about” rather than “what does this page say”—then topic pages are exactly what you’d want to read first. You’d build a domain expertise profile, not a page-level retrieval index.
3. The Q&A gap explains the citation gap. Claude’s citation rate is 21%. ChatGPT’s is 83%. ChatGPT-User reads Q&A pages constantly. ClaudeBot hasn’t touched one in the last four days. Q&A pages are phrased as questions—the same way users query AI. If ClaudeBot hasn’t indexed them, it’s missing the content format that most directly matches how people ask questions. If ClaudeBot comes back for a Q&A crawl and Claude’s citation rate jumps, that would confirm this.
The crawl itself
Outside those content sessions, ClaudeBot runs a monitoring loop: robots.txt plus sitemap.xml every 2–3 hours, always from the same single IP. One instance, checking in regularly, reading selectively when it decides to.
The citation test reflects this early stage. Claude cited Genymotion with a link in only 5 of 24 queries (21%), including a #1 position for “how can support agents replicate mobile app bugs on a virtual device” and #5 for manual testing. Brand mentioned in 8 of 24.
Three weeks ago it was zero, and three weeks before that ClaudeBot wasn’t crawling at all. The trajectory exists. What we don’t know yet is whether ClaudeBot’s structural approach—reading the map before reading the pages—eventually produces the same citation depth as ChatGPT’s content-first approach, or whether it leads somewhere different entirely.
Perplexity: promising, but we haven’t cracked it yet
PerplexityBot: 42 requests this week, up from 14. Tripled. It’s crawling pages it wasn’t touching before: VirtualBox troubleshooting, release notes, macOS errors, boot issues.
Citation results: 4 of 24 queries (17%). The hits clustered where you’d expect—in the categories where the AI site content is deepest: mobile security (#7 position), manual testing (#7, #8), app development (#6). Brand mentioned in 6 of 24.
The pattern looks like the early days of ChatGPT. Light crawling, citations starting to appear in the strongest content areas, gradually expanding. PerplexityBot has its own crawler, its own index, and its own retrieval pipeline. That’s the same architecture where the AI site has already proven it works.
42 requests a week is still light. What we haven’t figured out is what triggers PerplexityBot to go from light sampling to deep indexing. GPTBot did that jump on January 7 (547 requests in one day). We’re waiting for PerplexityBot’s version of that moment.
Gemini: a completely different game
This is the section we had to rewrite after looking at the actual response data. What we originally assumed about Gemini was wrong, and the real finding is more interesting.
We assumed Gemini’s low citation rate (4%, just 1 of 24 queries) was because it hadn’t crawled the AI site. Then we pulled the full response text from this week’s test.
What Gemini actually does
Manual Testing query: Gemini recommends Genymotion as #2 (“The Best Dedicated QA & Performance Option”), describes its UI, sensor emulation, cloud browser access. Mentions it 6 times. Zero citation links to genymotion.com.
App Development query: Gemini recommends Genymotion as #2 (“Best for Enterprise & CI/CD Testing”), describes cloud integration, lightweight client, sensor emulation. Mentions it 4 times. Zero citation links.
Embed on Website query: Gemini recommends Genymotion for enterprise use, describes its WebRTC streaming, AWS/GCP/Azure hosting. Mentions it 6 times. Zero citation links.
Pricing query: Gemini cites genymotion.com at position #1 with 13 citation links. Gives exact pricing: $206/year for Desktop, $0.05/minute for Cloud, $149/month unlimited, $0.60/hour for PaaS. More specific pricing data than ChatGPT provided.
Gemini doesn’t have a knowledge problem. It has a linking problem. It knows Genymotion well enough to recommend it with specific feature details in every query. It just doesn’t send anyone to genymotion.com.
The pipeline difference
Three out of four queries: Gemini answers from its training data. Good answers, detailed recommendations, no links. The user reads “get Genymotion” and has to go find it themselves.
One out of four (pricing): Gemini triggers a live Google Search, grounds its response in web results, and cites through Google’s Vertex AI redirect URLs. That’s the only query where genymotion.com gets a link.
What this means: ChatGPT sends traffic. Gemini sends word of mouth. When ChatGPT recommends Genymotion, there’s a direct link to genymotion.com right in the response. When Gemini recommends Genymotion, there’s nothing to click. The user has to go Google it—which means they hit Google Search results.
What this means for GEO strategy
The AI site works for platforms that have their own crawler and retrieval index: ChatGPT, Claude, Perplexity. Their bots crawl the AI site, index the structured content, and cite it with direct links.
Gemini doesn’t have its own AI crawler. It either answers from training data (no links) or triggers a Google Search through its grounding API (links to whatever ranks in Google). The AI site at rozz.genymotion.com isn’t in Gemini’s pipeline at all, because Gemini’s pipeline IS Google Search.
For Gemini, the path to citations isn’t better AI site content. It’s Google Search ranking. GEO strategy needs two tracks: one for the crawl-and-cite platforms (ChatGPT, Claude, Perplexity), where the AI site approach works, and another for Gemini, where the rules are different.
Where we are
Four platforms, four different stages:
ChatGPT: solved. 83% citation rate. Direct links to genymotion.com. The AI site approach works. Remaining work is expanding content coverage into weaker use cases.
Claude: structurally aware, content-light. 21% citation rate, growing. ClaudeBot is the only bot reading the site’s topic taxonomy, suggesting it may be building a structural model of domain expertise rather than a page-level retrieval index. Will it come back for a deep content crawl?
Perplexity: promising. 17% with light but growing crawl activity. Same architecture as the platforms where the AI site works. We believe this will follow the ChatGPT curve, but we haven’t triggered the deep indexing event yet.
Gemini: different game. Recommends Genymotion in most queries but doesn’t link. Citations only happen when Gemini triggers a Google Search. The AI site doesn’t reach Gemini’s pipeline. This might require a completely different approach.
One infrastructure handles three of four platforms. The fourth needs something else: Google thrived on SEO and still lives in that world.
Get This for Your Site
ROZZ builds this infrastructure automatically. AI site. Q&A pages from your chatbot. Schema.org markup on every page. llms.txt discovery files.
$997/month | 83% citation rate on ChatGPT
→ Book a call | → See how it works | → rozz@rozz.site
→ Data source: CloudFront access logs for rozz.genymotion.com, February 24 – March 3, 2026 (crawl data). Citation tests conducted March 2–3, 2026: 24 queries tested on ChatGPT, Claude, Perplexity, and Gemini. Bot classification based on User-Agent strings.