SEO podcast EP 31

How to get cited by ChatGPT: a two-week test with my own blog post

SEO podcast EP 31. Choo Sung-hoon UFC ranking test: ChatGPT cited my website as a source
SEO podcast EP 31 · Aired 23 October 2026 · 5 min · Korean audioWatch on YouTube

How do you get cited by ChatGPT? This is what I did when AI got a fact wrong, and what four AI engines said two weeks later.

In episode 25 I explained why AI search gets the ranking of Yoshihiro Akiyama (known in Korea as Choo Sung-hoon) wrong.

I then published that explanation on my website as a post and a video.

On 23 October 2026, about two weeks later, I asked four AI engines again while logged out.

When I checked for episode 25 (around 1 pm on 7 October), none of them gave his No. 8 record.

This time three of the four gave the No. 8 record: ChatGPT, Google AI Overviews and Gemini.

ChatGPT cited my website post (the episode 25 post) as a source. Google AI Overviews showed my YouTube video first in the video panel inside its answer.

This is a written version of episode 31 of The Insung Song SEO Podcast (4 min 45 s, in Korean).

It is translated from the Korean video and lightly trimmed for reading. The original Korean is in the Korean version of this post.

✦What did I check two weeks after episode 25?

Watch this part of the episode 00:00

Card with the episode 25 thumbnail: "AI got it wrong, world No. 8 is right" (Korean text on screen)
The episode 25 thumbnail shown at the start (video 0:08).

In episode 25 I explained why AI search gave the wrong UFC ranking for Akiyama, and how it could get it right.

Then I published that explanation on my website, as a post and a video.

About two weeks passed. So I asked AI, in Korean: "What is Choo Sung-hoon's highest UFC ranking?"

Then ChatGPT showed my post as a source in its answer.

✦What should you do when AI gets your information wrong?

Watch this part of the episode 00:32

Text card: "Someone asked AI about my name, and the answer was wrong?" (Korean text on screen)
The question this episode answers (video 0:36).

Someone asks AI about your name or your information, and the answer is wrong. What should you do? This episode answers in the order I tried it.

First I looked at the records to see why AI got it wrong. Then I did three things based on that analysis.

Two weeks later I asked again to see whether the answers had changed.

✦What did AI say about the ranking on a TV show?

Watch this part of the episode 00:40

Screenshot of Sherdog's official rankings of 13 November 2007, middleweight, with the line 8. Yoshihiro Akiyama boxed in red
The No. 8 line in Sherdog's middleweight ranking of 13 November 2007 (video 1:12).

On a Korean TV travel show, Akiyama said he had been ranked No. 8 in the world.

The clip is an unaired part of SBS Plus 상남자의 여행법 in 규슈, posted on its official YouTube channel on 6 May 2026.

He and the crew then checked a search engine and AI, and AI said he was not No. 8.

But when I looked at the records, No. 8 was right. It just was not a UFC ranking. The official UFC rankings only began in February 2013.

The No. 8 was the middleweight world ranking published in November 2007 by Sherdog, a US MMA outlet.

✦How did four AI engines answer at the time of episode 25?

Watch this part of the episode 01:14

Card with a large zero: 7 October, 1 pm, logged out, four engines, No. 8 answers: 0
At the time of episode 25, none of the four engines asked while logged out said No. 8 (video 1:24).

When I made episode 25, I asked AI too.

Around 1 pm on 7 October 2026, logged out, I asked four engines: ChatGPT, Google AI Overviews, Perplexity and Gemini.

None of them said No. 8. Google AI Overviews and Perplexity even said he had been No. 10 at welterweight.

When I asked Google again at 2:29 pm the same day, AI Overviews did say No. 8, but it cited the TV show clip (this is in the episode 25 post).

The table in section 7 puts those answers next to the answers two weeks later.

✦Why did AI get the ranking wrong?

Watch this part of the episode 01:31

Text card: "A record has to exist as a sentence for AI to say it" (Korean text on screen)
Explaining why AI got it wrong (video 1:40).

The record existed, but no one had written it down as a sentence.

No page said in one sentence when, where, in which ranking and at what position, like "No. 8 at middleweight in Sherdog's 2007 ranking".

Namuwiki and Wikipedia did not have it either.

His name was split three ways as well: Choo Sung-hoon in Korean, Akiyama Yoshihiro in Japanese and Yoshihiro Akiyama in English.

So the records about him were scattered across many places.

AI builds answers by pulling sentences from here and there. With no clean sentence to pull, it took one of the scattered numbers from articles and posts.

The welterweight No. 10 came from a single line in a Korean news article from May this year (Mydaily, 27 May 2026), and that line gave no source.

Press this and my new posts show up more often in your Google results.

✦What did I do so AI could answer correctly?

Watch this part of the episode 02:24

Screenshot of the episode 25 post: "No. 8 is right, but the UFC had no official rankings back then" (Korean text)
The first thing I did: the sentence written in the episode 25 post (video 2:35).

Based on that analysis, I did three things.

  • I wrote the missing sentence on my own site first. In the episode 25 post I wrote: "No. 8 is right, but the UFC had no official rankings back then."
  • I linked the record pages of the outlets that ranked him: Sherdog's ranking of 13 November 2007 and his Fight Matrix record.
  • I made a blog post on my website and embedded the episode 25 video in it. Every time I record an episode, I publish its content on my website as well.

With the answer and the evidence in one place, it is easier for AI to use them.

✦How did AI answer when I asked again two weeks later?

Watch this part of the episode 03:07

Screenshot of a logged-out ChatGPT answer: No. 5 middleweight on Fight Matrix, No. 8 on Sherdog, and a source chip reading Insung Song on the official ranking line
Logged-out ChatGPT answer on 23 October 2026, with Insung Song as the source chip (video 3:16).

On 23 October 2026 I asked again, logged out.

ChatGPT put the No. 8 record in its answer, and it showed my post as a source.

The line with the source chip was "there was no official ranking system in his early UFC years". The chip read Insung Song +1: my site and one other source.

Google AI Overviews showed the episode 25 video first in the video panel under its answer.

My channel has about 100 subscribers, and when I checked, that video had 99 views.

A Google search for "추성훈 ufc 랭킹" (Choo Sung-hoon UFC ranking) showed the episode 25 post third on the first page.

Gemini also gave the No. 8 record correctly, but showed no source.

In the episode 25 check (around 1 pm on 7 October), none of the four gave the No. 8 record. Two weeks later, three of them did.

The same four AI engines, two weeks apart. All logged out, asked in Korean, once per engine (Gemini twice on 23 October, No. 8 both times). 7 October: the episode 25 check around 1 pm (Google "추성훈 UFC 8위", the others "추성훈 ufc 랭킹"); asked again at 2:29 pm, Google said No. 8, citing a TV show clip. 23 October: "추성훈 UFC 최고 랭킹은 몇 위야?" (What is Choo Sung-hoon's highest UFC ranking?), Google only "추성훈 ufc 랭킹".
AI7 October, around 1 pm (episode 25)23 October (two weeks later)Source shown on 23 October
ChatGPTNo mention of No. 8. Not in the official UFC rankingsNo. 5 middleweight on Fight Matrix (January 2008), No. 8 middleweight on Sherdog (November 2007), no official ranking system in his early UFC yearsInsung Song +1 (my site and one other) on the official ranking line
Google AI OverviewsNo mention of No. 8. Top 10 at UFC welterweightNo. 8 middleweight on Sherdog (November 2007), No. 5 middleweight on Fight Matrix (January 2008), never in the official UFC rankingsEpisode 25 video first in the video panel
GeminiNo mention of No. 8. Never in the official UFC top 10, around No. 12 at middleweight in media rankingsNo. 8 middleweight on Sherdog (November 2007), No. 5 middleweight on Fight Matrix, no official UFC rankings before 2013None shown
PerplexityNo mention of No. 8. Highest No. 10 in the official UFC welterweight rankingsNo. 10 at welterweight according to press reports, No. 5 middleweight on Fight Matrix (2008)A news article (daum) and 10 sources in total

✦Did every AI change its answer?

Watch this part of the episode 03:54

Comparison card of four engines: ChatGPT correct, Google AI Overviews correct, Gemini correct, Perplexity No. 10 at welterweight
On 23 October 2026, three of four engines gave the No. 8 record (video 4:00).

No. On 23 October Perplexity still said No. 10 at welterweight.

On the same day Google AI Overviews in a logged-in session said No. 12 on Sherdog, but Sherdog's November 2007 list has him at No. 8.

So it is not a one-time fix: I keep asking the same question and watch how the answers change.

✦How do a website and SEO affect AI answers?

Watch this part of the episode 04:07

Screenshot of Google results for "추성훈 ufc 랭킹" with the insungsong.com episode 25 post third
The episode 25 post third in Google results for "추성훈 ufc 랭킹" (logged out, 23 October 2026) (video 4:18).

I think my website had an effect too. AI also finds pages on the web to build its answers, and my post was the third result on Google.

In Google Search Console, my site had 23 impressions for the query "추성훈 ufc 랭킹" from 16 to 22 October 2026.

The average position was 3.8 (22 October data not final yet).

So if you want to appear in AI answers, it helps to take care of SEO as well, so your pages rank high on Google.

That said, I asked each engine once, so I cannot pin the change on a single cause.

Being cited as a source is not the same as being recommended. For why AI recommends some brands over others, see episode 23.

✦What does branding look like in the AI era?

Watch this part of the episode 04:24

Text card: growing followers is branding; writing first the answer AI will take is branding in the AI era (Korean text on screen)
Summing up branding in the AI era (video 4:30).

Growing your followers is branding. But AI will also talk about you, and writing down first the answer you want it to use is branding in the AI era too.

If AI is saying something wrong about you or your brand, I hope this episode helps.

✦Wrap-up

When AI got Akiyama's ranking wrong, I did three things.

I wrote the clean sentence on my own site first, linked the record pages of the outlets that set the number, and embedded the video in the post.

Two weeks later the AI engines giving the No. 8 record went from none of four (around 1 pm on 7 October) to three.

ChatGPT cited my website post as a source and Google AI Overviews showed my video first in its video panel.

Perplexity still says No. 10 at welterweight, so I keep watching.

Try this today: ask AI about your name or your brand, and write down the answer and its sources with the date.

When you ask again two weeks later, you can compare what changed.

✦Three-line summary

  • In the episode 25 check (around 1 pm on 7 October 2026), none of four AI engines gave Akiyama's No. 8 record (Sherdog, November 2007). Two weeks later, on 23 October, three did: ChatGPT, Google AI Overviews and Gemini.
  • AI got it wrong because the record existed but no page put it in one sentence, and his name was split three ways. So I put a clean sentence, the ranking outlets as sources and the video on my website.
  • ChatGPT cited my website post as a source, and Google AI Overviews showed a video from a channel with about 100 subscribers first in its video panel. Perplexity still says No. 10 at welterweight, so I keep watching.

Sources

The text is translated from Korean, based on the captions of the published video, with verbal tics removed; the English is a translation of the meaning only. AI answers were checked while logged out, once per engine (Gemini twice), except the logged-in Google answer noted in section 8. Answers can change from day to day. Click "Watch this part of the episode" under a section title to play that part of the video.

✦Common questions

In the Akiyama ranking case, Insung Song did three things. He wrote the correct answer as one sentence on his own site, linked the record pages of the outlets that set the number as sources, and embedded a related video in the post. Two weeks later, ChatGPT showed his site as the source for one line of its answer. One test cannot prove the cause.

On 23 October 2026, logged out, ChatGPT was asked in Korean for Choo Sung-hoon's highest UFC ranking. The line "there was no official ranking system in his early UFC years" showed a source chip with his site name (Insung Song +1). His episode 25 post on that site had the answer sentence and the Sherdog and Fight Matrix sources in one place.

In this test, yes. On 23 October 2026 the episode 25 video from Insung Song's channel, which has about 100 subscribers, appeared first in the video panel of Google AI Overviews. The video had 99 views when checked. This was a single logged-out check.

The record existed but was never written as one sentence. No page said he was No. 8 at middleweight in Sherdog's November 2007 ranking, and his name was split across Choo Sung-hoon, Akiyama Yoshihiro and Yoshihiro Akiyama. The welterweight No. 10 answer came from one unsourced line in a May 2026 news article.

Insung Song thinks so. AI also finds pages on the web to build answers, and this time the episode 25 post was the third Google result for the query. So if you want to appear in AI answers, he says it helps to take care of SEO as well.

Share this post
Insung Song, SEO & AI Search Consultant
Written by
Insung Song
SEO & AI Search Consultant, founder of Hosoo Media

I help companies grow in search, mostly on Google, in Korea and across borders. Since 2021, the field has taught me to look at content and structure together rather than chasing one technical fix.

Get in touch

Tell me where you are stuck

Insung Song, SEO & AI Search Consultant based in Seoul
You do not have to figure out Korea on your own.

If you are planning to launch in Korea, or you are already there and search is not working, send me a note. Tell me about the company and where things stand. I'll look at it myself and tell you whether I can help, and where I would start.

I read every message myself. If I am not the right fit, I will say so and point you somewhere better.

Consent

Prefer email? admin@hosoomedia.com · Based in Seoul (UTC+9) · I reply in English or Korean.

Newsletter

Not ready to talk yet?

I send notes on what is actually happening in Korean search. What changed at Naver, what worked, what did not. Written for people who have to decide about this market.

Consent

Press this and my new posts show up more often in your Google results.