How to check whether your brand appears in ChatGPT's answers

Start by entering the category questions customers would actually ask into ChatGPT yourself. Then add multi-engine comparison, competitor share, citation-source back-tracing, and regular repeat measurement in turn, turning a one-off check into a diagnostic you can re-run the same way every time. The table below shows the order in which an in-house team can start for free and expand to tools or an agency when needed.

1. Enter your category's core prompts into ChatGPT yourself

Pick 10–20 questions customers would actually ask in your category, enter them into ChatGPT as they are, and note whether your name appears in the answers. This is the fastest method and costs nothing; before you even evaluate a separate tool, 30 minutes is enough to get a feel for where you stand.

Customers ask ChatGPT something like "recommend skin-firming treatments for people in their 50s" and choose from the three or four brands that appear in the answer body. Whether your name is on that list decides whether inquiry calls come in. According to Omnia's summary of a G2 survey, 50% of B2B buyers begin their purchase journey on an AI chatbot instead of traditional search. Of those, 47% named ChatGPT as their first choice.

Group the questions into category-entry, comparison, and alternative types. An entry question is something like "recommend popular vegan skincare brands in Korea," a comparison question is "what's the difference between Brand A and Brand B," and an alternative question is "something worth using instead of a premium cleansing oil." Repeat each question three times in a new chat window and check whether the answer is the same each time.

If your brand appears, record the following.

• Citation position: the first paragraph, the middle, or a footnote at the very end
• Every competitor name mentioned alongside you in the same answer
• The wording used to describe you (copied exactly, without editing the sentence)

ChatGPT gives different answers to the same question depending on the session, account, and time, and extra elements such as map widgets or cards don't appear the same way every time either. So use this method for one-off checks, and don't treat the numbers it produces as KPIs.

2. Do the same brands appear in Perplexity, Gemini, and Claude?

ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews each differ in training data, search backend, and how they choose citation sources, so the list of brands that appears for the same question differs by engine. Measuring only one engine means you've seen only half the picture.

The point where Korean in-house teams most often stall is the move from step 1 to step 2. They enter a question or two into ChatGPT, conclude "we don't show up," and stop, but only by going on to multi-engine comparison and share of voice do you get a map of where the gaps are.

Take the question set you built in step 1 and enter it into the other engines as it is. Each engine's characteristics are roughly as follows.

• Perplexity shows the source URLs for every answer right on the screen. It's the fastest way to learn which pages are the source of answers in your category.
• Gemini is based on Google search results, so it tends to cite pages with strong SEO authority first.
• Claude sometimes answers from its trained knowledge alone without searching, so newer brands are left out relatively often.

Don't add the results together; lay them out in a single table. Put questions across the top and engines down the side, and in each cell write "mentioned or not, citation position, competitors that appeared alongside." Use this table to decide which screen to fix first.

If your brand still doesn't appear after steps 1–2, the cause is usually one of five. You can narrow them down one by one in Why your brand doesn't show up in ChatGPT and Gemini.

3. How do you calculate share-of-voice, competitors included?

Share of voice (SoV) is calculated by counting how many times, and with what nuance, you and each competitor are mentioned in the same answers, to show "who holds what % of this category's answer space." If you count only whether you appeared, you can't see where you stand against competitors.

SoV can be compared with last month only if the question set stays fixed. Choose 50–100 questions, give the set a name, and measure with the same set every time. Organize the results of each measurement round into the items below.

• The number of times our brand appeared in the answer body out of the total N questions
• The number of times competitor A appeared
• The number of times competitor B appeared

Re-measure this ratio every four weeks with the same question set and draw a trend line. If a ratio that was 30% last month fell to 18% this month, that can be a sign a competitor published new content in the meantime. According to GoodFirms statistics, 58.5% of searches end without a click. Being cited inside an AI answer is exposure in itself, so if your SoV is 0%, you are effectively a nonexistent brand in that answer space.

4. Back-trace the cited source URLs

If AI doesn't mention you, some other page is currently being used as the source for that answer. Analyze that page and you can see almost exactly what content you need to reinforce.

Back-tracing is easiest in Perplexity. Click the source card below the answer to open the cited URL, and on that page note the following.

• The H1/H2 structure and whether there's an FAQ section
• The schema.org markup type (Article, FAQPage, HowTo)
• External media cited in the body

ChatGPT often doesn't list sources in its answers. In that case, fill the gap with Semrush's ChatGPT clickstream analysis, comparing the domains frequently used in ChatGPT citations across the whole industry.

When you analyze several pages, certain kinds of pages keep showing up as especially frequent answer sources.

• Pages with a definition sentence in the first paragraph
• Pages whose FAQ is marked up with schema
• Pages where citations of authoritative media are woven naturally into the body

The AEO guides Search Engine Land and Moz published in 2026 also point to the same three signals.

5. Automated repeat measurement

Running steps 1–4 once gives you a single snapshot. To use it for decision-making, you need to repeat the same measurement weekly or monthly and accumulate the results in chronological order. Manual measurement can't be sustained once you pass 50 questions. In GoodFirms' 2026 AI SEO statistics roundup, 14% of organizations actually tracked their AI exposure.

For automation, use the tools below.

• Dedicated AI visibility measurement tools: Otterly.AI, Adthena, Profound, Nozzle. These tools are built for this purpose, from per-question citation tracking to automatic source extraction and competitor share calculation.
• Add-on features of existing SEO tools: Semrush AIO, Ahrefs Brand Radar. They let you see search-ranking data and AI exposure together on one screen.

Overseas tools often have shallow coverage of Korean-language questions, so brands with a high share of Korean prompts, such as beauty and consumer goods, should first check whether the tool properly captures their category questions before adopting it.

If you use an agency instead of a tool, PION runs these five steps on the brand's behalf. We ask ChatGPT, Gemini, Google AI Overviews, and AI Mode the same questions on a set schedule to measure per-question citation frequency and position and competitor share against a Korean-language category set, and people take those results through to content planning and publishing. PION handles this recurring measurement and follow-up work.

Don't rush step 5. It's worth the cost only if you adopt it after repeating steps 1–4 enough to decide which question set to keep tracking.

If you want a free diagnostic report instead of checking yourself, distinguish between a report that ends with a single score and one based on actual prompt measurement. The items a report should contain are laid out in Free AI visibility audit reports from GEO agencies: What to check before requesting one.

If it's hard to decide which method to start with, you can first check your site's readiness with the AI answer visibility basics check.