# Six surfaces, and the method behind each one

> Provider APIs and a licensed SERP vendor. Never a logged-in user’s personalised result.

Source: https://citetune.com/surfaces

## Measured through provider APIs

GPT, Gemini, Claude and Perplexity are queried through their providers’ own interfaces, so the prompt, the model and the run date are all recorded with the answer.

## Measured through a licensed SERP vendor

Google AI Overviews and Google AI Mode are collected through a licensed search-results vendor, because Google offers no API for either.

## What we cannot see on any surface

No personalisation, no logged-in state, no ad placements, and no guarantee that two runs of the same prompt return the same answer.

Every sample reports how many prompts were attempted and how many completed, so a partial run is visible as a partial run.

## FAQ

### What is a completion rate?

The share of attempted prompts that returned a usable answer on a given surface in a given run. We publish it beside every figure, because a 40 percent result over a 30 percent completion rate is a different claim from the same number over a full run.

### Do you measure what a logged-in user sees?

No, and nobody honestly can at scale. We measure a clean, unauthenticated query through a documented method, which is reproducible; a personalised answer is not.

### Will you add more surfaces?

Only when we can state a method for one. A surface we cannot measure reproducibly is worth less to you than an honest gap, so it stays off the list until the method exists.