# 1 prompt. 4 LLMs. 48 bots.

> Canonical source: [1 prompt. 4 LLMs. 48 bots.](https://ai.answer.cloud/pages/1-prompt-4-llms-48-bots/)

Same prompt, same site, four LLMs. Wildly different bot behavior. ChatGPT self-identifies, Gemini cites without obvious bots, Claude mixes disclosed and anonymous fetchers, and Grok drives the most impact while staying fully anonymous. AI bot “personality” shows up in fetch patterns and disclosure and cost clearly shapes how these systems crawl and cite your site.

I wanted to understand whether AI systems have *distinct bot personalities*, not in what they say, but **how they show up on your site**.

Here’s what I did:

- Asked the **same prompt**
- About the **same site**
- Observed behavior via **server logs + citations**

Here’s what I found 👀

---

### What AI bots actually look like in the wild

| LLM | Citations | AI Bots Observed | Self-Identifying? | Example User Agent |
|---|---:|---:|---|---|
| ChatGPT | 10 | 7 | Yes | `ChatGPT-User/1.0; +https://openai.com/bot` |
| Gemini | 7 | 0 | No | `Googlebot/2.1` (mobile Chrome UA) |
| Claude | 10 | 2 | Mixed | `ClaudeBot/1.0; +claudebot@anthropic.com` |
| Grok | 81 | 39 | No | Standard Chrome on macOS |

---

### A few things stood out

**Self-identification is optional**  
Only ChatGPT consistently raises its hand. Gemini appears to rely heavily on search ingestion. Claude will *both* self-identify *and* send anonymous fetchers.

**“No bots” doesn’t mean no traffic**  
Gemini produced citations without any clearly identifiable Gemini bots. Other LLMs likely used cached or previously ingested data. This aligns with recent cost-optimization behavior in newer models (e.g. ChatGPT 5-2 auto).

**Anonymous ≠ low impact**  
Grok generated the *most* citations while being fully anonymous at the UA level.

**Each LLM has a bot personality**  
Fetch patterns, timing, headers, and disclosure habits differ significantly. We don’t fully understand what drives this behavior yet, but **cost is clearly a major factor**. We observed a sharp drop in bot activity after GPT-5 launched, followed by a partial reversal to improve response quality.

---

AI bot traffic is the best signal we have for understanding **how your site is represented inside LLMs**—but it’s still an imperfect science.