RAIVE-engine
Diepgaande analyse van Rankfenders Real AI Visibility Engine — het centrale scanningsysteem.
What is RAIVE?
RAIVE (Real AI Visibility Engine) is the core of the AI Visibility module. It orchestrates the entire process of discovering how AI systems perceive and recommend your brand, from language detection through to response analysis. Every scan follows a deterministic 5-stage pipeline that ensures consistent, reproducible results.
5-Stage Scan Process
Each RAIVE scan moves through five sequential stages. Every stage carries a weight that reflects its contribution to overall scan progress.
| Stage | Weight | What Happens |
|---|---|---|
| 1. Language Detection | 3% | Reads brand language from your Brand Book to ensure prompts and analysis are performed in the correct language. |
| 2. Sitemap Validation | 5% | Verifies that indexed pages exist and are accessible. A valid sitemap is required for the scan to continue. |
| 3. Topics Extraction | 7% | Checks for existing content topics. If none are found, the engine auto-generates relevant topics from your Brand Book and sitemap data. |
| 4. Prompts Discovery | 5% | Generates test prompts using Strategic Intelligence, combining your topics, competitors, and industry context into realistic AI queries. |
| 5. AI Response Collection | 80% | Sends every prompt to each monitored AI system and analyzes the responses for brand mentions, competitor references, and citation quality. |
Real-Time Progress
While a scan is running, a real-time progress modal displays key metrics so you can follow along without guessing. The modal shows:
- Prompts Found: The total number of prompts queued for this scan.
- AI Systems Queried: How many AI platforms have been contacted so far.
- Responses Collected: The count of completed AI responses.
- Mention Rate: The percentage of responses that include your brand.
- Brand Mentions Count: The raw number of times your brand has been mentioned.
- Elapsed Time: How long the scan has been running.
Progress updates are delivered via a real-time database subscription. If the subscription drops, a 3-second polling fallback ensures uninterrupted updates. The estimated remaining time is calculated from the average processing speed observed during the current scan.
Response Analysis Per Prompt
Every prompt goes through a multi-step analysis pipeline that extracts the maximum amount of intelligence from each AI response.
- Cache Check: The engine first checks for existing responses less than 7 days old. If a valid cached response exists, it is reused to save time and API costs.
- Fresh Collection: When no cache hit is found, RAIVE calls the AI API directly. Each call is attempted up to 3 times with exponential backoff to handle transient failures.
- Brand Detection: The response is scanned to determine whether the brand mention is organic (AI mentioned it unprompted) or branded (the prompt explicitly referenced the brand).
- AI-Powered Analysis: GPT-4o-mini evaluates the response to determine whether the brand was actively RECOMMENDED, not just mentioned in passing.
- Visibility Score Calculation: A composite score is computed from response quality, position within the answer, sentiment, presence of citations, and the number of competitor mentions in the same response.
- Citation Extraction: For platforms that provide source links (notably Perplexity), all cited URLs are extracted and stored for later analysis.
Organic vs. Branded Mentions
Understanding the difference between organic and branded mentions is critical for interpreting your visibility data accurately.
- Organic: The AI mentioned your brand even though the prompt did not contain your brand name. This is the most valuable type of mention because it indicates genuine brand authority and recognition in the AI's training data and retrieval sources.
- Branded: The prompt explicitly asked about your brand (for example, "What do you think of Acme Corp?"). While still useful for tracking sentiment, branded mentions carry less weight in scoring because the AI was directly prompted to discuss you.
Running RAIVE
To start a scan, click the "Run RAIVE" button on the AI Visibility dashboard. A typical scan takes between 2 and 5 minutes depending on the number of topics, prompts, and monitored AI systems in your project.
The frequency at which you can run scans depends on your subscription plan:
- Essential: Weekly scans
- Professional: Every 2 days
- Agencies: Daily scans
- Enterprise: Daily with near-real-time monitoring capabilities
To configure which AI models are tracked and adjust scan preferences, see Monitoring Settings.