Automate Share of Voice Tracking for AI Search Engines

Professional services firms rely on tools like SEMrush or Ahrefs for share of voice. These platforms were built to track ten blue links on a search results page. Their core metrics, like keyword rank, are becoming obsolete as AI engines provide direct answers and citations instead of just links. They are slow to adapt, often misclassifying AI-generated content or missing citation opportunities entirely.

In practice, this means you are blind to your true visibility. Consider a management consultancy that published a detailed guide on market entry strategy. SEMrush shows them ranking number four, which seems good. But Google's AI Overview and Perplexity both cite a competitor's article for the same query, pushing the consultancy's content completely out of view for most users. There is no report in Ahrefs to show you this. This forces your team into hours of manual, one-off searches that are impossible to scale across 50 or 100 critical service-related questions.

Some firms try to solve this with internal scripts using Python libraries like BeautifulSoup. These custom scrapers are brittle and high-maintenance. AI search engines constantly change their HTML structure, use dynamic JavaScript rendering, and actively block scraping attempts. An internal script that works today will break next week, turning a strategic project into a frustrating and expensive maintenance burden.

The structural problem is that traditional SEO platforms are built on a data model of 'rankings'. AI search engines operate on a model of 'answers' and 'citations'. You cannot accurately measure the new model with tools built for the old one. You need a system designed from the ground up to parse and analyze the specific output of modern answer engines.

The first step is to define what share of voice means for your firm in AI search. This involves mapping your key service lines to a list of 50-100 commercial-intent questions. We then identify which AI engines matter most to your clients (e.g., Google's AI Overviews, Perplexity, Brave Search) and audit your competitors' current citation frequency.

We built our own AEO pipeline to solve this, and the same approach applies here. The system uses a scheduler, running on GitHub Actions, to trigger a Python script that queries search APIs like the Brave Search API. For engines without official APIs, the system uses Playwright for headless browser automation to simulate user queries and capture the full page structure. Raw results are parsed and stored in a Supabase database using a schema designed specifically for citation tracking.

The delivered system is a live dashboard that updates every 24 hours, showing your citation share for each tracked question and flagging competitor movements. It runs in your own cloud environment for under $50 per month, handling up to 5,000 queries per day with data retention for 12 months. You receive the full Python source code and a runbook, giving you direct visibility that off-the-shelf tools cannot provide.