Contact by email
Back to Blog
AI News

Measuring ChatGPT visibility without fooling yourself

GlasBox Studio · · 3 min read

TL;DR: Key Takeaways

"Are we in ChatGPT?" is the wrong question. Better: on which buying questions does which wording appear - and does it rest on a checkable source?

  • A useful protocol is dull, which is why it repeats.
  • One chat is not a baseline.
  • Local CH prompts belong in the set: "cleaning company Lucerne", not only "best commercial cleaning Switzerland".

"Are we in ChatGPT?" is the wrong question. Better: on which buying questions does which wording appear - and does it rest on a checkable source?

A useful protocol is dull, which is why it repeats. Fixed prompt list (service, place, boundary, comparison). Fixed models and surfaces (ChatGPT, with and without browsing where relevant). Log date, language and account state. Classify the answer: named / misattributed / absent / refused. Note visible sources when the UI shows them.

One chat is not a baseline. Memory, UI experiments and model versions move the result. Run three repeats days apart, not one screenshot for the board.

Local CH prompts belong in the set: "cleaning company Lucerne", not only "best commercial cleaning Switzerland". SOURCE/01 by GlasBox (AI) uses that prompt set. It is an instrument, not a campaign.

---

How to measure AI visibility

Next step

GlasBox measures classic SEO and AI visibility separately. No ranking or mention guarantee. The free short check stays thin on purpose.

Request an introductory call

Frequently Asked Questions

Is 'Are we in ChatGPT?' the right question?
No. Better: on which buying questions does which wording appear - and does it rest on a checkable source?
Is one screenshot enough?
No. One chat is not a baseline. Run three repeats days apart, with a fixed prompt set, model and language.
Do local Swiss prompts belong in the set?
Yes. 'cleaning company Lucerne' belongs in the set, not only 'best commercial cleaning Switzerland'.