What an AI shows when you ask for someone's official channels
Ask an AI for someone's official channels and where does it point? We pulled the 3,909 answers that asked for a source out of 18,465, and counted citation rates by engine, the sources that came up most, and where a citation landed.
The people asking an AI where the official channel is
A brand manager checking a creator before working with them, a fan trying to tell the real account from the rest: the question goes into an AI chat now before it goes into a search box. "Tell me the official channels and profile." A name alone does not answer it. What did the AI pick as the basis for its answer?
What we counted and how
Between 19 July and 2 September 2026 we collected 18,465 answers to natural-language questions, asked in Korean, on ChatGPT Search, the Gemini app, Google AI Mode and Naver AI Search, covering 153 creators and 858 questions. This article looks only at the 3,909 answers to questions that asked for a source, such as "tell me the official channels and profile".
We sorted them into two states. If the exact pinnn.me/@handle address is in the answer, it is an exact citation. If the name is there and the address is not, it is a mention without a link. Links in the body and the source chips attached to an answer both count as citation URLs in the tally, and we separate the two only when describing a screen. A source chip is the small tag an answer attaches to show where it looked.
The name came through often, the link rarely
The creator's name appeared in some form in 86.6% of the official-channel answers. Mentions without a link accounted for 76.1%, and answers carrying the exact Pinnn address for 10.4%.
Exact citation rates by engine were 21.6% on the Gemini app, 12.2% on ChatGPT, 5.3% on Google AI Mode and 3.1% on Naver AI. For the same kind of question, roughly seven times separates the highest engine from the lowest.
Chart data (English, same values as the Korean original): Data for the chart (exact citation rate on official-channel questions, n=3,909 total): ChatGPT 12.2% (n=916), Google AI Mode 5.3% (n=947), Gemini app 21.6% (n=985), Naver AI 3.1% (n=1,061).
Their habits with links differ too. Among answers carrying at least one citation URL, the average number of URLs was 6.9 on Google AI Mode, 5.5 on ChatGPT, 3.3 on Naver AI and 2.7 on the Gemini app. The share of cited answers that included pinnn.me went the other way: 19.6% on the Gemini app, 16.2% on ChatGPT, 13.9% on Naver AI and 4.5% on Google AI Mode. The engine that attaches the fewest links put Pinnn in most often, and the one that attaches the most put it in least.
So where did they point
Google AI Mode attaches a bulk of generic links, Google help pages among them, to every answer, so we left it out of the source ranking. Counting how many times each URL appeared in the official-channel questions on the remaining three engines, YouTube came first at 1,140 and pinnn.me second at 801. Then Instagram at 378, google.com at 320, Namuwiki at 254, Naver blogs at 245, Kakao Channel at 243 and vling at 212.
Chart data (English, same values as the Korean original): Data for the chart (top 8 URL occurrences in official-channel answers, Google AI Mode excluded): youtube.com 1,140; pinnn.me 801 (2nd); instagram.com 378; google.com 320; namu.wiki 254; blog.naver.com 245; pf.kakao.com 243; vling.net 212.
These numbers are not a web-wide ranking or a market share. They are how often a URL turned up in this set of questions. Channels the person fills with their own work sit right next to pages a third party compiled.
When it was cited, it went first
Here we looked at all 1,193 exact citations rather than the official-channel questions alone. Within answers that cited Pinnn exactly, the share where Pinnn was the first source was 100% on the Gemini app, 96.3% on ChatGPT, 89.9% on Google AI Mode and 80.4% on Naver AI.
Chart data (English, same values as the Korean original): Data for the chart (share of answers citing an exact Pinnn URL where Pinnn was the first source; denominator = exact-citation answers per engine, total 1,193): ChatGPT 96.3% (395 of 410), Google AI Mode 89.9% (134 of 149), Gemini app 100.0% (471 of 471), Naver AI 80.4% (131 of 163).
The denominator is the exactly cited answers, not every question. Citation itself was rare, and once it was exact it often sat in the first slot.
How it looked on screen
In the Naver AI answer about the official channels of 임정수PD, the exact pinnn.me/@yprince77 address went in as an underlined direct link, and a Pinnn +2 source chip appeared with it. The blue outlines mark the link and the chip separately.
English reading guide (original Korean evidence): Question: Show yprince77’s official channels and profile. The answer contains a direct pinnn.me/@yprince77 link and a separate “Pinnn +2” source chip. Both forms of evidence appear in the same answer.
Open the full-resolution Korean original. Historical Naver AI screenshot in paired posts; direct URL and source chip preserved separately; original answer log not available.
The ChatGPT answer about AI융합전문가 윤성임 took a different shape. Instead of a link in the body, Pinnn showed up as a Pinnn +1 source chip. Both are citations in the tally, and the place a reader can click is not the same.
English reading guide (original Korean evidence): Question: Show AI expert Yoon Seong-im’s profile and official channels. The answer summarizes her work. A “Pinnn +1” source chip is shown; there is no direct Pinnn URL in the visible answer body.
Open the full-resolution Korean original. Historical ChatGPT screenshot in paired posts; source chip is not a body URL; original answer log not available.
Both screens pointed at the Pinnn profile as the place the answer had looked. The screenshots are the real screens saved during measurement, cropped and masked for personal information before publication.
What an official profile tells a machine
A Pinnn profile takes one canonical address as its base, connects only ownership-verified channels as sameAs official channels, writes the display name and the handle variants down as aliases, and states through ProfilePage and Person that this page is a profile and its subject is a person.
Martinez's survey of GEO research frames exposure in a generative engine as a probabilistic multi-stage process of retrieval, selection and citation. In the SAGEO Arena study, rewriting body text to fit citation could hurt retrieval and reranking, while structural fields such as the title, the description, the headings and JSON-LD were observed as relatively safe inputs. JSON-LD is a way of carrying the facts a page holds in a form machines read. How that structure showed up in real answers is in the records of three creators.
Put your own name into ChatGPT now and ask it for your official channels. If no link comes back, make a profile at pinnn.me and verify your channels before you connect them.
Method and limits
This sample is built from questions about creators registered on Pinnn, so it cannot be used to judge which sites are better across the web. The numbers are tied to the period, the questions and the state of each engine, and the same question repeated can change its answer. A high exact citation rate also cannot be read as more visits, inquiries or revenue. This does not guarantee an AI citation, though it does give you one reference point for what counts as official information.
Sources
Related creators
More guides
Put these ideas into practice.
Bring your bio and official channels together on a Pinpage.