Scanning the web is only half the story
AngryRobot Scanner's core scan tells you how your website performs technically. But there's a deeper question many brands want answered: what do AI models actually think about us?
That's what AI Perception is for.
How it works
When you run AI Perception, we send a structured research prompt about your brand to three AI models simultaneously:
- Gemini 2.5 Flash (Google): with live web grounding via Google Search
- Claude Sonnet (Anthropic): evaluated against its training knowledge
- GPT-4o (OpenAI): for a third independent perspective
Each model returns a structured assessment covering brand sentiment, competitive positioning, perceived strengths and weaknesses, and country/market attribution.
What the report surfaces
The results are synthesised into a unified AI Perception report with several sections:
- Brand Sentiment Score: a -1.0 to +1.0 scale showing how positively AI models perceive your brand, visualised with a full-range colour gauge
- Competitive Tier: where you sit relative to competitors (leader, challenger, adopter, unknown)
- AI Perception Summary: a plain-language synthesis of what AI models say about you
- Citation Sources: which websites the AI models are actually drawing from when they describe your brand
- Key Themes: topics and keywords the models consistently associate with your brand
Why multiple models?
Different AI models have different training data, different recency windows, and different tendencies. A brand that scores well with Google's Gemini might be almost unknown to OpenAI's models. Running all three gives you a more honest picture than any single model could provide.
It also surfaces gaps: if your brand is well-known in one AI ecosystem but invisible in another, that's a strategic insight, one that a single-model scan would miss entirely.
Available on Hobby plans and above
AI Perception is included from the Hobby plan. The multi-model architect comparison (with individual scores from each model) is available on Startup and above. See all plans.