Do ChatGPT, Gemini and Perplexity name your business when a buyer asks?
This is for owners and independent professionals who sell a service and suspect that buyers now ask an AI assistant who to hire. I work from Istanbul, in English, Turkish and Persian.
I count how often these assistants name your business, tell you at which step you drop out, and count again. I do not promise rankings, traffic, leads or citations.
What goes wrong
A buyer types a question into ChatGPT: who does this, in my city, in my language. The answer names a handful of businesses. If you are not one of them, you never hear about the buyer who chose someone else.
From the outside you cannot tell why. The assistant may never have found you, or it may have found you and passed you over. Each needs a different fix.
And one answer proves nothing. Ask the same question again and the names change. I counted this on my own site (see "What I measured").
How it works
-
I record your site as it is
Before I ask anything, I write down what your site looks like to a machine: its pages, what its robots file (the file that tells crawlers what they may read) allows, its structured data (labels in the page code that describe it), and its language tags. Without this, a later comparison compares nothing.
-
I take the questions from real answers
The question list comes from a database of real AI answers in your category, not from my imagination. Where no such database exists for a language, as with Persian, I say so and mark those questions as assumed.
-
I ask each question seven times
Seven times per question, engine and language. Today I measure ChatGPT in English, Turkish and Persian, tied to a city. Gemini and Perplexity I measure in English only, and without a city, which makes that evidence weaker. I count how often your name appears and how often your own site is cited, and I write both as a count out of a total, for example "3 answers out of 7".
-
I diagnose in a fixed order
Was an answer produced at all? Is your site open to the assistants? Is it found? Was it found and passed over? Only then does the page itself matter. Rewriting a page that was never found is wasted work.
-
You get the fix list, then I measure again
The same questions, word for word, are asked again on day 7, day 14 and day 30, because one early re-check can mislead (see "What a published test found").
What you get
-
One readable report Every rate is a count out of a total. Anything not measured says "not measured", never zero.
-
The names you lose to Who the assistants recommend instead of you, ranked by how often, and which sites they were cited from.
-
A diagnosis Which of the five checks you pass and which you fail.
-
A ranked fix list Each item says who applies it, what it needs, and how strong the evidence behind it is.
-
A re-measurement plan The question list recorded word for word, with the dates for day 7, 14 and 30.
-
Site changes, kept separate I can make them, or your webmaster gets a written spec. We settle which on the call.
What I measured
-
I measured my own site first. My name appeared in 0 of 98 answers and my site was cited in 0 of 98. That is ChatGPT in English, Turkish and Persian, plus Gemini and Perplexity in English. With 98 answers, my real rate could be anywhere from 0% to about 3%.
Date: 12 September 2026 Source: Baseline of erfandigital.com: 98 answers, each question asked 7 times
-
Asking once misleads. A combination is one question put to one engine in one language. I ran 14 of them, seven times each. In 11 of them the list of names was different every one of the seven times. Perplexity returned the same twenty sources all seven times.
Date: 12 September 2026 Source: Baseline of erfandigital.com: 98 answers, each question asked 7 times
-
Some questions name nobody yet. I asked ChatGPT in Turkish whom to consult to get a business name into ChatGPT. In 7 of 7 answers, no name was given.
Date: 12 September 2026 Source: Baseline of erfandigital.com: ChatGPT, Turkish, 7 runs
-
Still to do: the second measurement of my own site, after the fixes. When it is done, the result goes here, whatever it shows.
Date: As of 22 October 2026 Source: Re-measurement plan: day 7, 14 and 30 after the changes go live
What a published test found
-
This is not my result. In a controlled test of 81 new pages tracked daily for 30 days, ChatGPT Search cited 17% of them on day 7, 35% on day 14 and 42% on day 30. Google AI Mode peaked at 56% on day 7 and fell to 26% by day 30. It describes new pages on one high-authority site, not a result for yours.
Date: September 2025 Source: Semrush, 81 new pages on one domain (DR 84), daily tracking for 30 days
Questions
Can you promise ChatGPT will name me?
No. I do not promise it, because the answers change from run to run. What I promise is a count of how often you are named today, a ranked list of fixes, and the same count repeated on day 7, 14 and 30.
Why do you ask each question seven times?
Because one answer is a sample of one, and the lists change between runs, as the measurements above show. Seven is the minimum I work to. It lets me say how often you appear instead of whether you appeared once.
Do you sell llms.txt files or schema markup?
No. An llms.txt file is a plain-text summary of a site for AI tools. Google says such files neither help nor harm visibility, and that no special schema markup is needed to appear in its AI answers. FAQ rich results also stopped in Google Search on 7 May 2026. I do not sell any of it as a way into AI answers.
Can I be named if I work online with no office?
Possibly. In my Istanbul measurement, individual consultants with their own websites were named. The local business panel, built from Google Maps data, may matter too. I read four answers word for word: the 2 that showed the panel named people, and the 2 without it named nobody. Whether the panel causes that is untested.
Google Business Profile, the listing behind it, is only for businesses that meet customers face to face. If you never do, your site and directories come first.
How long before I see a change?
Nobody can say how long, and there may be no change at all. So I look three times, on day 7, 14 and 30, and report whatever is there. The published test above shows why one look misleads: ChatGPT Search kept rising to day 30, while Google AI Mode peaked on day 7 and then fell, so a Google AI Mode reading at day 14 that is lower than at day 7 can be normal.
Is this worth doing for you?
Tell me what you sell, where, and in which languages your buyers ask. We decide together on the call.