AI Visibility Monitor

Track how often your brand appears in AI-generated answers across ChatGPT, Gemini, Perplexity and DeepSeek. As people increasingly ask AI assistants instead of searching Google, being mentioned in AI answers is becoming as important as ranking in search results.

You give it your brand name and product category. It asks each AI platform realistic buyer questions, reads the answers, and reports whether your brand was mentioned, how prominently, and how you compare to competitors.

What it does

• Queries up to four AI platforms with the same set of prompts
• Detects whether your brand (and any aliases) appears in each answer
• Measures prominence - a 1-10 score for how early the brand is mentioned
• Tracks competitors and computes your Share of Voice
• Collects cited source URLs from web-grounded answers
• Produces an overall AI Visibility Score (0-100)
• Supports per-country, per-language tracking across 37 countries and 26+ languages

Web search vs memory - an honest distinction

• ChatGPT, Gemini, Perplexity answer with live web search. A mention here reflects real, current AI search visibility.
• DeepSeek answers from model memory only - its API does not perform web search. A mention here means your brand is well-known enough to be in the model's training data. Every result is tagged grounding: memory so you can read it correctly.

Regional tracking

Brand visibility in AI answers varies by region. A brand that AI mentions as "the best note-taking app" in the US may be invisible in China, irrelevant in Japan, and ranked third in Germany. To capture this:

• Set Country to an ISO code (US, DE, FR, JP, RU, BR, TR, CN, ...) and the actor adds an "in [country]" clause to each default prompt and routes it in the language of that country.
• Set Language to override the country default. For example, Country=DE + Language=en runs English-language prompts about Germany ("What is the best note-taking app in Germany?"), useful for English-speaking expats or B2B buyers.
• Built-in native templates exist for 8 languages (en, de, fr, es, it, ru, pt, ja). Other languages (pl, tr, sv, no, da, fi, zh, ko, id, vi, th, he, ar, uk, cs, el, hu, ro, nl, and more) are translated on the fly via an LLM and cached - the first run in a new language takes ~2 extra seconds, subsequent runs are instant.
• In custom prompts mode the region is recorded in the output but not applied to your prompt text - you write what you want, the country and language fields are saved alongside each result for filtering.

Fan-out (sub-prompts)

Enable Fan-out to expand each main prompt into related "People Also Ask"-style follow-ups in the same language. An LLM generates 1-5 (default 3) more specific questions per main prompt - narrower use cases, audiences, platforms, price tiers. This reveals where your brand is invisible: AI may mention you for "best note-taking app" but skip you for "best free note-taking app for students" - that gap is a content opportunity.

Each sub-prompt is billed as a separate check, so fan-out multiplies cost proportionally (default 3 mains x 4 sub each x 4 platforms = 48 checks instead of 12). Disabled by default. When enabled, sub-prompts are generated once per (category, language, country) combination and cached, so repeat runs do not re-pay for generation. Each check in the output carries a parentPrompt field that points to the main prompt it expanded from, so you can group results in your analysis.

Measuring frequency (repeats)

AI answers are not deterministic - the same question can mention your brand once and skip it the next time. By default the actor runs each prompt once. To measure frequency, increase Repeats per prompt to 3-5: with 3 repeats you see whether your brand appears in 1/3, 2/3 or 3/3 answers.

Input

• Brand name - the brand to track (required)
• Brand aliases - alternate names, abbreviations, domain
• Competitors - competitor names for Share of Voice
• Product category - used to auto-generate realistic prompts
• Prompt source - default (auto-generated from category) or custom (use your own prompts)
• Custom prompts - your own questions, used when Prompt source is custom
• Country - ISO 3166-1 alpha-2 code (US, DE, JP, ...) for regional targeting
• Language - ISO 639-1 code to override the country default (en, de, tr, zh, ...)
• AI platforms - which of the four to query
• Number of prompts - how many prompts per platform when source is default (1-25)
• Repeats per prompt - repeats for frequency measurement (1-10)
• Enable fan-out - off by default; on, an LLM expands each main prompt into follow-up questions
• Sub-prompts per main prompt - how many follow-ups per main when fan-out is on (1-5, default 3)
• Response detail - short excerpt or full answer text

Output

Two record types: check (one per prompt x platform x repeat) and summary (one per run with overall visibility score, mention rate, Share of Voice, per-platform breakdown). Each check carries country, language, effectiveLanguage and parentPrompt fields so you can group results by market or fold sub-prompts back into their main prompt.

Pricing

Paid per event - charged once per check (prompt x platform x repeat). All AI API costs (including the optional LLM template translation for non-built-in languages) are included; you need no API keys of your own.

Use cases

• Generative Engine Optimization (GEO) - measure and improve how AI describes your brand
• Competitive intelligence - see which competitors AI recommends
• Content impact tracking - run before and after publishing
• Multi-market brand audit - compare visibility across US, UK, DE, FR, JP, BR, ... in one report
• Agency reporting - add AI visibility as a KPI

Notes and limitations

• AI answers change over time - schedule regular runs to build a trend.
• DeepSeek results reflect training-data knowledge, not live web presence.
• This actor measures visibility (are you mentioned), not search volume - AI platforms do not publish how many people ask a given question.
• Brand mentions vary heavily by region. 100% visibility in Germany and 0% in China for the same brand is a real, useful signal - not a measurement error. Use it to identify markets where you need to invest in local content and authority.