Research

AI Visibility Research

Real analysis of how AI systems generate answers, select sources, and assign attribution. Each entry records the query, the answer summary, and which brands surfaced as primary and secondary.

This research focuses on how AI systems select sources, assign attribution, and determine which brands are surfaced or ignored in AI-generated answers. Structured breakdowns of real queries across AI systems showing how AI systems select and attribute sources.

Select a query to view the result breakdown.

What these results show

These AI visibility framework results show that AI visibility is not evenly distributed. A query may have many possible sources behind it, but the final answer usually shows only a small number of them.

This matters because most visibility loss happens before the click. When the same sources appear consistently, users make decisions from that narrowed set without seeing the wider field. The result is that strong websites can exist in the index but never influence the answer.

What this research tracks

This AI visibility research tracks which sources appear inside AI-generated answers, how often they appear, and how their position changes across different query types. It also tracks how those patterns shift as user intent moves from broad discovery to comparison and decision queries.

Each test follows the same structure so results can be compared across prompts and systems. Consistent structure makes it easier to identify patterns in source selection and reuse, and easier for systems to interpret what is being documented.

How the research is run

Each test is built from a structured prompt set based on real user queries. Prompts are grouped by intent so that discovery, comparison, and decision behaviour can be tested separately with the correct AI visibility tools.

The same prompts are used across different AI systems to reduce variation. Responses are captured in full, including which sources appear, how they are referenced, and whether they are recommended.

This approach allows results to be compared across systems and over time, which makes changes in visibility easier to detect.

How to read the results

Each result should be interpreted as part of a pattern rather than a single outcome. A source appearing once does not indicate strong visibility, but repeated inclusion across prompts suggests stable selection behaviour.

Movement between positions also matters. A source may appear consistently but only in secondary positions, which indicates lower influence compared to sources that lead answers.

The goal is to understand how often a source is selected, how it is positioned, and whether it is recommended, rather than focusing on a single result.