Our Methodology
How we measure AI visibility, transparently.
How We Test Each Platform
We call each calibrated platform's native grounded search API, not consumer UI scraping. ChatGPT uses the OpenAI Responses API with web_search_preview. Perplexity uses the Sonar API. Gemini and China-market platforms are enabled for customer reports only after calibration for the engagement scope.
Buyer Prompts
We run prompts that sit inside real purchase decisions (e.g., "What project management tools do B2B SaaS teams use?"), not keyword lookups. Prompts are drawn from our calibration fixture library, seeded from buyer-intent research across verticals (B2B SaaS, e-commerce, manufacturing, consumer health).
What "Cited" Means
A brand is cited when its canonical name, domain, or a registered alias appears in the grounded AI answer. We use weighted signal matching: canonical name (weight 1.0), domain (0.9), alias (0.7), product name (0.5). A combined score ≥ 0.5 registers as cited.
Platform Calibration
Each platform prober is validated quarterly against the equivalent consumer UI. We measure citation-set overlap using the Jaccard index. A platform must reach ≥ 70% overlap before being used in customer reports. Where API and UI diverge beyond the threshold, reports carry an explicit footnote.
Platform Coverage
Default calibrated live probes: ChatGPT (OpenAI Responses API + web_search_preview) and Perplexity (Sonar API). Scoped engagements can add Google Gemini and Chinese-market probes such as Doubao, DeepSeek, Kimi, Qwen, and ERNIE after calibration. Entity and schema optimization benefits broader AI platforms, including Claude and Microsoft Copilot, even where live probing is not available.
What We Don't Measure
Sponsored or paid placements within AI answers. Model training-time memorization (we only measure retrieval-augmented grounding). Mobile app interfaces, which may differ from the API path. We are transparent about these limitations in every report.