Attribution gap
The attribution gap is the difference between how much AI answers shape what people buy and how much of it your analytics can show.
The attribution gap is the difference between how much AI answers influence your customers and how much of that influence your analytics can actually show. A lot of what happens in AI search never reaches your reports.
It's why a brand can be named in thousands of AI answers a month and see only a trickle of AI visits in Google Analytics.
Where the influence goes missing
Your analytics sees a visit when someone clicks through from an answer and their browser says where they came from. Plenty of influence happens outside that:
- People read the answer and never click, because it told them enough
- They remember your name and search for you later, which shows up as a branded search
- They type your address straight in, which shows up as direct traffic
- They copy a link into a new tab, which drops where they came from
- They buy in a shop, or on a marketplace, where your analytics can't follow
For example, someone asks Perplexity for the best coffee subscription for an office. The answer names Fernhill Coffee first. Two days later, she searches "Fernhill Coffee" on Google and signs up. Fernhill's analytics credits Google search. The answer that put Fernhill in her head gets nothing.
How to work with it
You can't close the gap entirely, but you can read around it:
- Track how often you're named, as well as how many visits AI sends
- Watch branded searches and direct traffic alongside your AI visibility
- Ask new customers where they heard about you, and include AI assistants as an option
- Compare your numbers before and after a specific change
In AnswerOrbit
We measure the part before the click: how often the platforms name you and which pages they read. What people did after reading an answer is one of the things we can't see, so you'll need your own analytics for that. The AI traffic tab in Website brings in the AI visits Google Analytics does see. And Before and after a date, in Recommendations, shows your numbers either side of a change you made.