---
title: "NUR AI research"
description: "Studies by NUR AI of how AI assistants answer: what they name, how consistent they are and which sources they use."
canonical: "https://discovernur.com/research/"
lang: "en"
last-updated: "2026-10-12"
---

# What the assistants actually say.

Studies from NUR AI’s own measurements: the same questions asked many times across the AI assistants, every answer kept and counted. Each one says what it cannot show.

## [AI cited this site 80 times. It never visited.](https://discovernur.com/research/ai-cited-your-page-did-it-visit/)

145 answers from ChatGPT, Perplexity and Gemini cited one site 80 times. Not once did the assistant ask for the page while it answered: it worked from a copy it already had.

## [Does AI always pick the same name?](https://discovernur.com/research/does-ai-always-pick-the-same-name/)

560 answers across seven categories: every model picked the same CRM in 40 answers out of 40, the assistants split on PR agencies, and in business services half the sources were companies’ own websites.

## [What does AI tell the crowd to buy?](https://discovernur.com/research/what-ai-tells-the-crowd-to-buy/)

121 answers from ChatGPT, Gemini, Claude, Perplexity and Google to two questions about buying a stock: one name in 45% of them, and an app that gave the same first pick ten times out of ten.