SE-Ranking Study: Backlinks Drive ChatGPT Citations

An analysis of 129,000 domains shows: The number of referring domains is the most important factor for ChatGPT citations.

If you want to know what really generates seo-glossary/visibility/">visibility in the age of AI, it's worth taking a look at the SE-Ranking study from 2025. The team analyzed around 129,000 domains and their mentions in ChatGPT responses. The result provided one of the first data-driven answers to a question that many marketers are concerned about: What factors determine whether a website is cited in generative responses or not? Spoiler: Backlinks clearly lead the ranking.

Methodology: 129,000 Domains, ChatGPT as a Testing Ground

The study captured a sample of around 129,000 domains from various industries and countries, analyzing how often they appeared as sources in ChatGPT responses. Thousands of topics were queried, and the citation frequency per domain was recorded. At the same time, classic SEO metrics were collected: number of referring domains, organic traffic, trust scores, number of indexed pages, anchor text distribution, domain age, and other factors. These data points were then correlated with the ChatGPT mentions to identify and weigh influencing factors. The methodology thus combines classic SEO audit data with a new metric, namely AI visibility, providing robust correlations between both worlds for the first time.

The statistical robustness of the study is important. With 129,000 domains, individual outliers can no longer be used as an explanation. The patterns that emerge are systemic and reproducible. This makes the investigation one of the most valuable sources for SEO and GEO strategists who are faced with the question of where to invest their budgets. The industry distribution also covers a wide range, from B2B software to e-commerce, financial services, travel, and health, to classic industrial and craft sectors, ensuring that the results are not limited to a single market niche.

The Core Result: Link Profile Beats Everything

The strongest individual factor for the frequency of ChatGPT citations was clearly the number of referring domains. Websites with over 350,000 referring domains received an average of 8.4 citations per topic area. Domains with 50,000 to 350,000 referring sources averaged around 4 to 5 citations. For domains with 10,000 to 50,000 referring sources, the value dropped to about 2.5 to 3 citations, and websites with fewer than 2,500 referring domains only achieved 1.6 to 1.8 citations. Even domains with fewer than 500 referring sources were virtually absent from the responses, supporting the threshold theory: There is a critical mass below which a link profile does not generate measurable AI visibility.

The difference is not subtle; it is drastic. The upper segment is cited more than five times as often as the lower one. So if you are considering how much link building budget to allocate, that is the answer: Enough to get out of the lower corridor. Any shift into a higher segment will measurably impact AI visibility. Particularly noteworthy: The growth curve is not linear but accelerates above certain thresholds. Those who manage to scale from 10,000 to 50,000 referring domains gain proportionally more AI visibility than in the first 10,000 domains.

The Top 3 Factors in Detail

SE-Ranking identified a total of 20 factors that correlate with ChatGPT citations. The top three positions are the most important and provide a clear roadmap. First place: Number of referring domains. As described, the strongest individual factor. Second place: Organic traffic. Domains with high visibility in Google are also cited more often by AI models because they are perceived as reliable, frequently used sources. Here, the reinforcement effect between Google rankings and AI selection that we have already described comes into play.

Third place: Trust scores. This includes domain trust metrics that depict the relationship between authoritative link sources and spam sources. A domain with 10,000 high-quality backlinks will rank higher than a domain with 100,000 backlinks of questionable origin. AI models seem to apply this quality filter very similarly to Google. Only in positions four to twenty do factors like domain age, number of indexed pages, anchor text diversity, technical SEO hygiene, and content freshness follow.

The number of referring domains is the strongest individual factor for ChatGPT citations. Ignoring this means optimizing past the actual lever.

What This Means for Marketing Budgets

The implications for marketing budgets are significant. Many companies have shifted large parts of their SEO budget into content production and on-page optimization over the past two years. This is not wrong, but it is only half the truth. Without parallel investment in domain authority, the best content remains invisible, both in Google and in AI response engines. Those who allocate 80 percent of their budget to content and only 20 percent to link building do not have an optimal allocation in the age of AI.

A balanced distribution often looks like this today: 40 to 50 percent for high-quality content and data assets, 30 to 40 percent for strategic link building including digital PR, and 15 to 25 percent for technical SEO and user experience. Neglecting any of these three pillars forfeits the potential of the other two. Link building is not a discipline that runs alongside; it is its own strategic lever with its own KPIs and investment logic.

Consequences for Your Next Steps

If you take the results of the SE-Ranking study seriously, three immediate steps emerge. First: Have your current link profile professionally audited. How many referring domains do you have today? In which segment of the study do you fall? What proportion is high quality? Which domains are potentially toxic and should be neutralized via disavow? Which competitors have gained the most in the last twelve months and what sources have they used? Second: Set realistic goals for the next 12 to 24 months, measured in additional referring domains per quarter. Third: Connect your link building activities with content assets that naturally attract links, such as your own studies, tools, or industry reports. These three steps only work together; one alone without the others is not sufficient.

What you should not do: Postpone the topic. The study does not show that more backlinks will soon become irrelevant, but that the lever is stronger than ever. Every month without strategic link work is a month in which competitors pass you by. In the age of AI, visibility is a game with a long half-life, and the early bird wins. Those who manage to jump from 2,500 to 10,000 referring domains measurably elevate themselves out of the lower segment, and those who scale from 50,000 to 350,000 enter the league of AI standard sources.

Data does not lie. If you want to be cited in ChatGPT, you need a link profile that can compete with the top domains in your industry.