ai newsDi., 20. Okt.
EilmeldungThe October 2026 GTM Stack Teardown: Why Multi-Tool AI Bloat Is Draining B2B Budgets

We Asked ChatGPT and Google AI Mode the Same 60 Questions. They Shared 19 of 507 Cited URLs.

We put the same 60 buyer questions to ChatGPT and Google AI Mode on the same night. Question by question, the two engines shared 19 of 507 cited URLs. That's 3.7%. Blend both engines into one "AI visibility" score and you're averaging two different source lists. Everything behind that number is pub

dev.toMontag 23:00 · 19.10.1 Min. Lesezeit
Bild: dev.to

We put the same 60 buyer questions to ChatGPT and Google AI Mode on the same night. Question by question, the two engines shared 19 of 507 cited URLs. That's 3.7%. Blend both engines into one "AI visibility" score and you're averaging two different source lists.

Everything behind that number is public: the questions, all 917 citation rows, the ChatGPT answers and our page-by-page notes. CC BY 4.0, on GitHub and Hugging Face, with a DOI. This post covers how it was collected, how a citation is counted, and how to recount every headline number with about 20 lines of pandas.

What the data turned into: the 44-second film.

Load it in 30 seconds · Recounts and findings · The playbook built on it · 中文完整数据说明

Put the same 60 Singapore buyer questions to ChatGPT and Google AI Mode and they cite almost entirely different pages: 19 of 507 cited URLs shared (3.7%). Ask ChatGPT the same question twice, minutes apart, and only 23.9% of its cited URLs come back. Here is every citation behind those numbers, so you can check them yourself.

Five Singapore industries. 110 buyer questions. ChatGPT (OpenAI API, gpt-5.5 with web search, Singapore location) and, in round 1, Google AI Mode (through a third-party SERP API). This dataset keeps the questions, every cited URL in answer order, the ChatGPT answer texts, our page-by-page annotations of cited pages, and a…

Quelle: dev.to (dev.to). Original öffnen