How Cogvert Scout works

Scout is a four-stage pipeline: collect → analyse → score → recommend. This page walks through what happens at each stage, so the numbers on your dashboard make sense — and you can defend them in front of a CMO.

The four stages

1. Collect — query the real AI engines

   For every prompt you've added, Scout fires three parallel API requests: one to ChatGPT (gpt-4o), one to Gemini (gemini-2.5-flash), and one to Perplexity (sonar). We use the official APIs, not browser scraping — so the data is stable and the queries can't be blocked.

2. Analyse — extract structured signals

   Each raw answer is sent through a second pass with gpt-4o-mini using strict JSON output. This extractor reads the answer like a human and produces a structured record:

   {
     "brand_mentioned": true,
     "recommended": true,
     "mention_count": 3,
     "sentiment": "positive",
     "position_rank": 2,
     "brands_in_response": ["Cogvert", "Competitor A", "Competitor B"],
     "excerpt": "Cogvert is a strong choice for…",
     "competitor_reasons": { "Competitor A": "praised for pricing" }
   }

   Strict JSON is critical: it means every chart in the dashboard is reading from the same well-formed schema, not from regex matches over English text.

3. Score — roll signals up to a single number

   Scout combines those signals into a Visibility Score out of 100, weighted to reflect what actually matters for AI search:

   Component	Weight	What it captures
   Mention rate	30 pts	How often you appear at all.
   Recommendation rate	30 pts	How often the AI actively recommends you.
   Position score	25 pts	Average rank when you do appear (1/rank).
   Sentiment	15 pts	Tone of mentions (positive 1.0, neutral 0.5, negative 0).

   The full math is documented on the Visibility Score page.

4. Recommend — turn data into next actions

   The last stage is the Optimize Hub. Scout looks at where your score is leaking — missing mentions, weak sentiment, lost citations — and produces a prioritised playbook split across five action channels: Content, Technical, Citations, Reddit, and LinkedIn.

Why Scout uses APIs (not browser scraping)

Concretely, the API approach gives you:

• Reproducibility — the same prompt run a week apart returns comparable, model-versioned data.
• Speed — parallel requests, not a queue of headless browsers.
• Stability — UI changes by OpenAI don't break Scout; only API deprecations do, and those are rare and announced.
• Observability — when something fails we get a real error code and can fix it the same day.

What Scout does NOT do

To set expectations:

• Scout does not measure organic Google search ranking — that's traditional SEO. Use a tool like Search Console for that, then connect it to Scout for richer prompt discovery.
• Scout does not measure paid placements inside AI answers (those don't really exist yet — when they do, we'll add them).
• Scout does not automatically post content for you. The Optimize Hub tells you what to write; you (or your content team) write it.

Last updated July 2026 · Suggest an edit