Unpacking AI-Driven Traffic: Insights from a Looker Studio Dashboard for LLM Visits
As businesses increasingly monitor how AI models interact with their content, practical analytics dashboards have become essential. Recently, I explored Steve Toth’s Looker Studio dashboard designed to track traffic generated by large language models (LLMs). The results were eye-opening and offered tangible takeaways for anyone managing a site that attracts AI-driven visits.
What the dashboard reveals
– A roll-up of LLM-driven visits: Over the last 30 days, the dashboard shows approximately 1,500 visits originating from LLMs. This figure highlights how often AI agents are engaging with online content, even if those visits come from automated sources rather than human users.
– Page-level focus: The majority of LLM traffic appears to be landing on the homepage, rather than deep into secondary pages. This pattern can inform how you structure your homepage to accommodate AI-driven discovery and routing.
– Dominant model driving the traffic: ChatGPT accounts for about 99% of the LLM traffic in the dataset. This concentration underscores which AI services are most actively indexing or interacting with the site and can influence how you prioritize content and optimization strategies.
Why these numbers matter
– Understanding AI-initiated engagement helps shape content strategy: If most LLM traffic is arriving at the homepage, it may be worth considering how the homepage communicates value, guides further exploration, and surfaces the most relevant content for AI crawlers and assistants.
– Traffic composition informs optimization priorities: Knowing that a single model type overwhelmingly drives LLM visits can help you tailor meta-data, schema, and page structure to improve how that model interprets and navigates your site.
– Observability of AI activity signals broader trends: Dashboards like this offer a concrete view into how often and where AI models interact with your site, enabling more informed decisions about content updates, accessibility, and user experience considerations for AI-powered automation.
Access and replication notes
– The dashboard in question is attributed to Steve Toth’s Looker Studio project. While I’m not including a direct link here, I’m happy to share access with interested readers upon request. If you’re considering a similar setup, you can replicate the approach by connecting Looker Studio to your data sources (for example, your analytics platform), creating dimensions that identify visits by model type, and building visualizations that track visits by page, model, and time window.
How to build a comparable LLM-traffic dashboard
– Connect Looker Studio to your data source: Use your analytics or log data to identify visits by AI model, source, and destination page.
– Define LLM-origin signals: Create dimensions or calculated fields that tag visits coming from recognized LLMs (e.g., model names, user agents, or other identifiers).
– Create key visuals: Build charts that show total LLM visits over time, distribution by landing page, and the share of traffic by model type.
– Add context with supporting metrics: Include metrics like bounce rate for LLM visits, average session duration, and conversions if applicable to understand engagement quality.
– Keep it accessible: Ensure dashboards are shareable with appropriate permissions so stakeholders can monitor AI-driven traffic without compromising data security.
Final thoughts
Dashboards that illuminate AI-driven traffic offer practical, actionable insights for site owners and marketers. By looking at who’s indexing or interacting with content and where they land first, you can tailor experiences, optimize critical entry points, and plan for evolving AI participation in web traffic. If you’d like to see Steve Toth’s dashboard in question, or discuss how to set up a similar Looker Studio view for your site, feel free to reach out.
