Does the AI ecosystem know your brand?

AI Citation Scan is a free intelligence platform that tests whether major AI systems know your brand, what they say about it, and how accurately they describe it. It sends your domain through multiple AI model families simultaneously and returns scored, compared, and annotated results — no account required.

The free scan tests four AI model families: Meta AI (powered by Llama 3.3 70B via Groq), Google AI (powered by Gemini 2.0 Flash), Mistral (powered by Llama 3.1 8B via Groq), and Gemma (powered by Gemma 2 9B via Groq). ChatGPT (GPT-4o) and Claude (3.5 Sonnet) require adding your own API keys in Settings. Models are auto-detected and updated automatically when providers change them.

API keys are stored exclusively in your browser's localStorage. They are never transmitted to or stored on our servers. Keys are only used during an active scan to authenticate your requests, then discarded. Use one dedicated browser for this tool and avoid clearing browser data to keep your keys persistent. You can export a backup JSON file from Settings as a safeguard.

The awareness score (0–100) measures how well an AI system knows your brand across four dimensions: Recognition (does it know you exist), Depth (how much detail it knows), Accuracy (is what it knows correct), and Confidence (how certain the engine is). Scores above 75 indicate strong awareness. Below 50 indicates significant gaps.

The consistency score measures how much agreement exists between different AI engines about your brand. A score of 100 means all engines produce identical information. Low scores mean users asking different AI tools get contradictory information about your brand — which erodes trust and creates confused brand perception across the AI ecosystem.

OpenAI and Anthropic do not offer free API access. The four free engines (Meta AI via Llama 3.3 70B, Google AI via Gemini 2.0 Flash, Mistral via Llama 3.1 8B, and Gemma via Gemma 2 9B) run at no cost. To include ChatGPT and Claude in your scan, add your own OpenAI and Anthropic API keys in the Settings panel. Your keys are stored only in your browser and used only for your scans.

Your scan results are not permanently stored anywhere. Results display during your current session only. If you close the tab or clear browser data, results are lost. Your API keys are also stored in localStorage — clearing browser data removes them. Export a key backup from Settings before clearing your browser if you want to restore them quickly.

The full platform adds citation probability scoring (which pages AI will cite when answering questions), chunk survivability analysis (how your content survives AI compression), entity knowledge graph intelligence (your presence in Google Knowledge Graph, Wikidata and DBpedia), and retrieval simulation (live testing of how GPT-4, Claude, Gemini and others actually retrieve your content). Join the waitlist to be notified when it launches.

AI models are non-deterministic — the same question can produce a slightly different answer each time, even with identical inputs. Re-scanning the same domain can shift scores by a few points run to run as engines phrase things differently. This is expected behaviour, not a bug. Treat single scores as a snapshot and look at trends across multiple scans over time rather than treating any one number as exact.

No. Each AI engine answers using only what it already learned during training — there is no live web search or browsing involved in generating a response. This is precisely the point of the tool: it measures what the AI already knows about your brand from memory, which is exactly what happens when a real user asks ChatGPT, Claude, or another AI assistant about you in everyday conversation.

Each engine's response is scored across four weighted dimensions: Recognition (30%) — does the engine know the brand exists; Depth (25%) — how much specific detail it provides; Accuracy (25%) — how many claims are verifiably correct versus conflicting or unverifiable; and Confidence (20%) — how certain the language used is. The four engine scores are averaged into one overall score, which maps to a tier from NOT KNOWN through WELL KNOWN. Tier thresholds are periodically recalibrated against real scans of well-documented global brands to keep verdicts realistic.