KI-Suche und Entdeckung
Den ChatGPT-Code entschlüsseln: Machen Sie Ihre Marke in der KI-Suche sichtbar
TL;DR: AI models don’t all rely on the same sources. ChatGPT draws heavily from managed third-party listings, while Gemini relies more on brand-owned websites. If ChatGPT visibility matters to your brand, start by making sure the profiles you manage across the web are accurate, complete, and consistent with your website.
When ChatGPT search launched at the end of 2024, “How do I show up in it?” was a genuinely open question. Today, there’s plenty of advice: structure your data, keep your information accurate across the web, write the way customers search, and build useful FAQs.
At the same time, the way people use ChatGPT has changed — as has the platform itself. Information-seeking grew from 14% to 24% of all ChatGPT usage between July 2024 and July 2025, according to an NBER working paper analyzing OpenAI's message data. And Pew Research Center found that six in ten US adults now read AI-generated search summaries, while 42% of chatbot users use them to look something up.
OpenAI has also added product discovery and an agentic commerce protocol, moving ChatGPT beyond answering questions and toward helping customers complete transactions.
So yes, the familiar AI visibility advice still matters. But there’s a problem with treating it as one universal playbook.
Answer engines synthesize information from the sources they retrieve. And different AI models don’t retrieve those sources in the same way.
That difference is one of the most useful things brands can understand about improving visibility in ChatGPT.
AI search visibility and ChatGPT visibility are not the same problem
Yext Research analyzed 17.2 million distinct AI citations gathered across Q4 2025, covering four models and seven sectors.
The first finding is encouraging: 90.6% of citations point to surfaces a brand can manage or influence. Brand-owned websites account for 46.89% of citations, while managed third-party listings account for 43.71%. Reviews and social make up 5.48%, with news, forums, and other sources accounting for the remaining 3.92%.
In other words, brands have a meaningful ability to improve much of the information AI models use to build their answers.
But that’s only part of the story.
Break those citations out by model and the pattern changes. Each model relies on a different mix of sources, and the differences are large enough to change where brands should focus first.
The report sums it up clearly: “Businesses that treat AI search as a single channel to optimize are making an assumption the data doesn't support.”
For marketing teams, that means “improve our AI visibility” isn’t one project. The right starting point depends on which AI experience matters to your brand.
Which surfaces each model actually cites
The percentages below show how much of each model’s citations come from brand-owned websites compared with managed third-party listings.
| Model | Brand-owned website | Managed listing |
|---|---|---|
| Gemini | 51.62% | 39.15% |
| Perplexity | 46.11% | 43.41% |
| Claude | 41.27% | 45.93% |
| ChatGPT | 39.55% | 53.36% |
Yext Research, AI Citation Behavior Across Models (Q4 2025). The study labels the OpenAI model SearchGPT.
Compare Gemini and ChatGPT. Gemini gets the largest share of its citations from brand-owned websites. ChatGPT gets the smallest share from those websites — and the largest share from managed listings.
That distinction matters.
If your AI visibility program has focused primarily on your website, that work aligns well with how Gemini finds information. But for ChatGPT, managed profiles play a larger role.
The pattern also changes by industry. ChatGPT was the most industry-variable model in the study, with the share of citations coming from websites ranging from 28.2% in Food and Beverage to 43.7% in Organizations.
Hospitality showed another notable difference. ChatGPT cited official hotel websites 38.08% of the time, while the other three models ranged from 16.72% to 22.44%.
The implication is simple: there isn’t one ChatGPT visibility strategy for every brand. A bank, restaurant group, and hotel brand may need to prioritize different sources.
And a higher volume of listing citations doesn’t make your website less important.
Listings account for 54.53% of the distinct URLs cited, but websites generate 4.31 citation occurrences per URL compared with 2.46 for listings. Once models find useful first-party content, they return to those pages more often.
Managed listings can help your brand become part of the answer. Your website gives AI models more detailed first-party information to return to.
What ChatGPT can and cannot see
Part of the reason ChatGPT behaves differently comes down to how retrieval works.
ChatGPT’s base model doesn’t browse the web on its own. Search depends on a retrieval layer that finds information for the model to synthesize into an answer.
Yext Research makes an important distinction here: “There is no published evidence that ChatGPT internally scores domains by trust, authority, or E-E-A-T. Any such behavior reflects the upstream retrieval system rather than the model itself.”
That changes how brands should think about visibility.
The goal is to make accurate, consistent information about your brand easy for the retrieval system to find and understand across the sources it uses.
Three points from OpenAI's own crawler documentation are especially relevant:
- OpenAI uses four crawlers with separate controls. OAI-SearchBot helps surface websites in ChatGPT search results. Blocking it can remove your pages from search answers while still allowing them to appear as navigational links. OpenAI says robots.txt changes can take about a day to propagate.
- The customer’s prompt isn’t necessarily the search query. ChatGPT can rewrite a question into several targeted searches before producing an answer. That makes specific content about a location, service, product, or customer need especially useful.
- Retrieval doesn’t automatically mean citation. OpenAI says ChatGPT may consult more sources than it displays. Getting retrieved gives your content a chance to contribute to an answer, but it doesn’t mean that content will appear as a visible citation.
Google takes a somewhat different approach. It says no special markup or AI text file is needed for its generative search features and points brands with physical locations toward their Business Profile for local details.
The broader lesson is consistent: the information brands manage outside their websites matters to AI visibility, too.
Getting cited is not the end of the decision
Getting mentioned in an AI answer is important. But it isn’t the end of the customer journey.
Yext Research's 2026 Consumer Search Behaviors report, based on 3,848 respondents, found that 95% take at least one verification step before acting on an AI recommendation — including customers who report high trust in AI.
They visit the brand’s website. They read reviews. They check profiles.
In financial services, where customers are evaluating high-consideration decisions, that verification step matters even more.
That means the AI citation and the experience customers find afterward are connected. If an AI answer gives one set of branch hours or advisor credentials and your website gives another, customers are left to decide which source is accurate.
Traditional local search visibility also remains part of that journey.
Yext Research's analysis of 21.6 million local search results found that brands with synchronized data ranked 2.71 positions higher within a mile of the searcher. That difference rose to 6.20 positions in markets with more than 100 competing brands.
The results also varied by industry. Banking showed the smallest difference in the study at 0.34 positions. For branch-based brands, the impact depends on the category and competitive market.
AI visibility doesn’t replace local search fundamentals. Customers move between these experiences, and the information they find should move with them.
Where to start
Start with the AI engine that matters most to your brand, then prioritize the sources that engine relies on.
For ChatGPT, the data points toward managed profiles first and brand-owned pages second. But the two need to tell the same story.
At scale, that becomes a data distribution problem. With Yext, the facts live in your Knowledge Graph and can be distributed directly to 200+ publishers without an aggregator in between. When an hour, credential, or service changes, you can update the source rather than fixing individual profiles one at a time.
The other half of the job is knowing where to focus.
Yext Scout tracks how AI models answer questions about your brand at the location level. It shows where and on which models competitors are being cited instead of you, so your team can identify the source data that needs attention.
AI search isn’t one channel with one playbook. Understanding how each model finds and cites information gives you a clearer place to start.
