How we measure whether ChatGPT recommends a studio

We ask ChatGPT, Perplexity and Google AI the questions a client would ask, in the same words each time. For every answer we note whether the studio is named, who is named instead and which pages the answer cites. Then we report how often the studio is named across the runs, never a rank, because AI answers change from one run to the next.

This is the method behind the monthly report in GEO + SEO Monthly; the free AI check runs its ChatGPT and Perplexity part live. Below it is used on our own studio.

Which questions do we ask?

The questions a client types when looking for a designer: "best interior designer in Milan", "minimalist interior architect in Lisbon", "who designs period homes in London". We agree the list with the studio before the first check: its city, its kind of work, and the places its clients come from.

We keep the wording the same every month. Otherwise a change in the answer could come from a change in the question.

The free AI check adds one instruction to each question: "Answer briefly, as a short ranked list of studio names with one line each." A longer answer could name more studios. The monthly check asks without it.

Which AI assistants do we check?

• ChatGPT, with web search on, the way most people use it.
• Perplexity, which reads more pages for each answer and lists its sources.
• Google AI, the answers Google writes above its results.

Gemini is not in the monthly check yet.

How do we read the answers?

For each answer we write down three things:

1. Named or not. Is the studio in the answer at all?
2. Who is named instead. These are the studios a client sees in its place.
3. Which pages are cited. These show what the assistant read: the studio's own pages, a directory, a magazine, an award page, a list written by someone else.

Answers change from run to run. The same question can name a different studio an hour later. So we never report "you are third". We report how many answers named the studio, out of how many.

A sample report: our own studio

We run the same check on fioel. These are the runs of 28 October 2026, through ChatGPT (with web search) and Perplexity, on questions that do not name us. One Perplexity answer failed and is not counted.

Named in 3 of 23 answers. On 21 October, a week earlier, our first check named us in 0 of 6.

Google AI is checked by hand. On 26 October it named fioel for a question about interior design studio websites in Amsterdam. On 28 October it did not name us for "best interior designer website maker in the world": it listed website builders first.

What did the sample show?

Three things, and they hold for interior studios as much as for us.

• A page about real work decides. Almost every agency ChatGPT recommended was cited from a page about a real, named client. For a studio, that is a project page with the place, the year and what was done.
• Perplexity reads more widely. Directories, award pages and lists written by others all appeared among its sources. Being listed correctly in a few good places counts.
• The language decides the city. The Paris question, asked in French, was answered from French pages only.

For us, the next step is clear: real client sites, shown with each client's written permission. For a studio, the first step is usually its own project pages.

What do we never do?

We never promise a place in an AI answer or on Google. No one can honestly promise that. We do the work that earns it, and every month we show you what changed.

If you want to see where your studio stands today, run the free AI check. It asks ChatGPT and Perplexity what your clients ask, live, and shows you whether you come up.