Docs/AI Visibility/Engines

AI Engines & How They Work Cloud

Each AI search engine has different data sources, ranking signals, and response styles. Understanding how they work helps you optimize for each one.

Perplexity

Model	Sonar (via Perplexity API)
Data source	Live web search (RAG pipeline) — every response is grounded in real-time search results
Citation rate	8.79 citations per response on average — the highest of any engine
Key signal	Search ranking + content freshness. If you rank well in web search, you are likely cited by Perplexity.

Perplexity is the most transparent engine — it always shows its sources. This makes it the easiest to optimize for: improve your search rankings and you improve your Perplexity visibility.

ChatGPT

Model	GPT-4o (via OpenAI API)
Data source	Training data + Bing search index. The API does NOT include web browsing.
Brand density	43% of responses include 10+ brands — ChatGPT tends to list many options
Key signal	Training data prominence. Brands with strong web presence at training cutoff are more likely to appear.

Important caveat: The ChatGPT web UI at chat.openai.com includes web browsing and plugins. The API does not. Our checks use the API, which represents the baseline model knowledge. Brands that appear via API are deeply embedded in ChatGPT's training data — a stronger signal than appearing only via web browsing.

Gemini

Model	Gemini 2.5 Flash (via Google AI API)
Data source	Google Search grounding — responses are grounded in Google's search index
Self-citation	52% self-citation rate — Gemini frequently references Google properties (YouTube, Google Maps, etc.)
Key signal	Google Search ranking + Google ecosystem presence (YouTube, Google Business Profile, Maps)

Gemini is location-aware through Google's index. When target markets are configured, location context is included in queries. YouTube content is heavily favored — if you have a YouTube channel, Gemini is where it pays off most.

Claude

Model	Claude Sonnet 4.5 (via Anthropic Messages API)
Data source	Real-time web search via the `web_search` tool — searches the web live during response generation
Authority weight	65% weight on high-authority sources — Wikipedia, established publications, and institutional sites
Key signal	Source authority + web search results. High-authority mentions (Wikipedia, press coverage) carry outsized weight.

Claude supports a native user_location parameter on its web search tool, which allows location-aware results without modifying the query text. When target markets are configured, this parameter is set automatically.

Grok

Model	Grok 3 Fast (via xAI API, OpenAI-compatible)
Data source	X/Twitter firehose as real-time signal + training data
Unique angle	Real-time social signals — trending topics, recent discussions, and viral content on X influence Grok's responses
Key signal	X/Twitter presence and engagement. Active X accounts with relevant discussions boost Grok visibility.

Grok is the only engine that draws heavily from social media in real time. Brands with active X/Twitter accounts and regular industry discussions see stronger Grok visibility.

API vs Web UI differences

This is an important caveat that applies to all AI visibility tools:

API responses differ from web UI responses. The ChatGPT web interface includes web browsing and plugins. The Gemini web app has different grounding behavior than the API. Our checks use the official APIs, which represent the most consistent and measurable baseline.

The advantage of API-based measurement is reproducibility and scale. Web UI responses vary based on user history, A/B tests, and feature rollouts that make consistent measurement impossible.

Which engines matter most

Brand type	Focus engines
B2B SaaS	Perplexity + ChatGPT + Claude — tech-savvy users use these most
Local business	Gemini (Google ecosystem) + Perplexity (web search)
Consumer brand	ChatGPT (largest user base) + Perplexity + Grok (social signal)
Media / content	Perplexity (citation-heavy) + Claude (authority-weighted)
Social-first brand	Grok (X/Twitter data) + Perplexity (Reddit as source)

Engine selection

You can choose which engines to monitor per brand. By default, all 5 engines are enabled. To change your selection:

Go to your brand detail page
Click Settings or edit the brand
Toggle engines on or off

Selecting fewer engines means each check uses fewer API calls (and credits), but each engine gets more runs for higher frequency accuracy. Selecting 1-4 engines runs 5 checks per engine; selecting all 5 runs 3 checks per engine.