Let’s face it, reporting on AI and LLM visibility is hard.
- Classic Google Search got massively disrupted by AI Overview and AI Mode – but the performance data got lumped into Google Search Console and we can only guess what clicks and impressions come from where.
- There is no “Search Console” for ChatGPT, Perplexity or Gemini. So, a lot of the data that we are used to (impressions, clicks, click through rate) does not exist.
- There is no best-in-class third party monitoring tool… yet. The big SEO tools are constantly evolving and new startups are popping up every day.
- Industry benchmarks are not really available yet but voices and opinions and Linkedin are very loud.
We feel you!
And whilst there are many things that we do not know yet. There are a few things that we can track already. And we want to show you how to use GA4 and Looker Studio with the right formulas to do so.
We generally use Looker Studio to report on GA4 data. The charts below are created using Looker Studio.
General Performance of LLM Referrals
This filter will allow you to report on sessions, engagement rates and conversions from LLM referrals.
- Dimension = page referrer
- Metrics = sessions, engagement rate, conversions, etc. Everything that you want to see.
- Filter = regex contains = ^https?://(?:(?:www\.)?(?:chatgpt\.com|perplexity\.ai|pplx\.ai|claude\.ai|you\.com|phind\.com|poe\.com)|chat\.openai\.com|chat\.mistral\.ai|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com)(?:/|$)
We use it in our Classic Search and LLM side-by-side comparison. It allows you to create charts and insights like this:
Breakdown of LLM referrals and which platform people came from
*This screenshot was taken mid-August, so the last month is not complete yet.
Landing pages that LLM referrals arrive at
We do not know what exactly people promoted – but we understand where they landed. And that allows us to draw conclusions about what topics we have visibility for.
If you want to dig even deeper into LLM visibility, think “social listening” for LLMs, try out Peec.ai or Ziptie.dev. They will provide you with insights about how often you appear within ChatGPT and alike for certain topics.
Key events conversions from LLM referrals
*This only includes referrals that came from LLMs and converted on the spot. It doesn’t reflect people who heard about you in LLMs but used Direct, Google Search etc. to find you. Self-reported attribution on the form can help you find these.
LLM referrals in context of other channels
Looker Studio – Manage Data Sources – Edit GA4 data source – Add Field
Use this CASE condition to break out LLM referrals:
CASE
WHEN REGEXP_CONTAINS(
LOWER(Page referrer),
)
THEN “AI Referral”
ELSE Session default channel group
END
It allows you to create charts and insights like this:
Limitations of LLM/AI Referral Data
There are always limitations to data, and different ways to measure LLM referral traffic. The regex provided above only matches a subset of possible sources. It doesn’t cover every single model or AI search tool available. They are the ones that we find are the most relevant in our work.
Another limitation is that some platforms don’t pass referrer information at all, this can include mobile apps, APIs, copy and pasting links directly into a browser, searching brand in Google, etc. In all of these cases, the session will appear as “direct” or misattributed to organic search, hiding the AI-assisted touchpoint.
These gaps make AI referrals appear less significant than they actually are, potentially causing you to skip AI optimization opportunities.
Self attribution is a great way to fill these data gaps, and can be as simple as asking on lead forms and sales calls. While imperfect, combining multiple attribution methods gives you a clearer picture of AI’s role in brand discovery.
