Data Study
ABSTRACT
Backlink profile diversity is proving to be a key factor in ranking within Large Language Models (LLMs) like ChatGPT, Claude, Grok, and Gemini. Our data shows that sites with a broader mix of backlinks consistently outrank those relying on a single source type. In other words, variety wins.
Read more to see the data behind this and learn how backlink diversity shapes LLM rankings.
Background
Traditional search engines like Google and Bing have long relied on backlinks as a fundamental ranking signal, treating them as digital “votes of confidence” that indicate a website’s authority. This link-based approach has shaped SEO strategies for over two decades. However, with the rapid emergence of Large Language Model (LLM)-powered search experiences; including ChatGPT, Claude, Grok, and Gemini; the established rules of search optimization may be undergoing a profound transformation. These AI platforms generate responses based on patterns learned from training data rather than real-time web crawling, raising questions about how they determine which tools or services to recommend.
Problem Statement
The core challenge facing digital marketers today is the opacity of LLM ranking algorithms. While these models were trained on web data likely influenced by backlink profiles, we don’t know if backlink metrics still determine LLM-generated search outputs. This uncertainty has significant implications: if backlinks remain influential, existing SEO investments retain value; if not, businesses need new optimization strategies focused on different signals that LLMs prioritize.
Objective
This study aims to empirically measure the correlation between traditional backlink metrics; including total volume, domain authority, and referring domain diversity; and ranking positions within LLM recommendations across B2B software categories. We seek to:
Scope
Our analysis encompasses 7 commercial-intent keywords spanning SEO tools and CRM platforms. We evaluated over 20 software tools across four leading LLM platforms, collecting ranking positions and comprehensive backlink metrics including total backlinks, referring domains, and high-authority links (DR 35+ and DR 65+). This scope examines whether “links as votes” continues to influence how AI systems surface and recommend software solutions.
To investigate the relationship between backlink metrics and LLM rankings, we decided to run a test around commercial-intent keywords in the B2B software space.
Tools Evaluated: 20+ major software tools including HubSpot, Semrush, Ahrefs, Pipedrive, etc.
For each keyword, we collected the following data points:
We compute the rank per tool across all LLMs per keyword.
Pearson correlation coefficient between backlink metrics and LLM average rank
Correlation between backlink metrics and normalized LLM rank
Now in Support of the Hypothesis
There’s a strong positive correlation between the number of referring domains and better LLM ranking.
Tools with higher backlink volume and diversity tend to surface higher in ChatGPT, Claude, Grok, and Gemini outputs.
The trend is most clear with referring domains, validating their continued weight in AI-driven search results.
The data indicates a moderate to strong positive correlation between the number of DR 35+ and DR 65+ backlinks and normalized LLM rankings. While high-DR backlinks alone aren’t the strongest signal, they contribute meaningfully when paired with domain diversity. Tools like Ahrefs, Semrush, and HubSpot CRM consistently rank higher and also lead in both high-authority links and referring domain count.
LLMs appear to value domain diversity over elite authority. While high-DR links have influence, outliers like SE Ranking and Screaming Frog show that niche relevance and brand factors also play a role.
ChatGPT and Claude rely most strongly on backlink signals, while Gemini and especially Grok show more variability, suggesting they incorporate additional factors beyond links.
SE Ranking and Screaming Frog rank highly despite weaker backlink profiles, while Moz Pro and Majestic underperform despite strong backlink counts, indicating that factors like freshness, reputation, or relevance also affect outcomes.
Backlinks explain about 60% of the variance in rankings. That means roughly 40% comes from other factors not captured in this study, such as content freshness, brand mentions, or semantic relevance.
The findings suggest the best ROI may come from prioritizing domain diversity and mid-tier backlinks (DR 35–65)
This study provides compelling evidence that the diversity of a website’s backlink profile is a critical driver of visibility within generative AI search outputs. Rather than signaling the end of traditional SEO, our findings show that LLMs continue to reward sites that earn recognition from a broad range of referring domains. The data suggests that generative AI systems still view these varied “votes of confidence” as a strong indicator of authority and trustworthiness. Far from being replaced, the principle of link diversity has carried over into this new search paradigm, underscoring its continued importance for businesses seeking prominence in AI-driven discovery.
Our analysis reveals a clear hierarchy of backlink influence on LLM rankings:
Referring domain diversity emerges as the dominant factor with a correlation coefficient of +0.62, suggesting that LLMs, like traditional search engines, value broad consensus from multiple sources over sheer link volume. This finding indicates that the “votes of confidence” principle remains deeply embedded in how AI systems assess credibility and authority.
Total backlink volume maintains substantial influence (+0.55 correlation), demonstrating that quantity still matters, though not as much as diversity. This suggests LLMs may be interpreting high backlink counts as proxies for content popularity and industry recognition.
Quality signals from high-DR backlinks show meaningful but secondary impact, with DR 35+ links (+0.52) and DR 65+ links (+0.43) both contributing to rankings. The diminishing correlation as DR thresholds increase suggests LLMs balance authority signals with other factors, preventing over-reliance on elite domain endorsements.
These findings have profound implications for businesses navigating the AI-powered search landscape:
Existing SEO investments retain value: Companies with strong backlink profiles built for traditional search engines are well-positioned for LLM visibility, validating past link-building efforts.
Diversification over concentration: The primacy of referring domains suggests strategies should prioritize earning links from many different sources rather than multiple links from few high-authority sites.
Multi-channel optimization is essential: While backlinks correlate strongly with rankings, they explain only part of the variance. Businesses must consider additional factors like brand mentions, content freshness, and semantic relevance to maximize LLM visibility.
Our research uncovered notable differences between LLM platforms. ChatGPT and Claude demonstrate stronger adherence to backlink signals, suggesting their training data or algorithms may weight traditional authority metrics more heavily. Conversely, Grok and Gemini show greater ranking variability, potentially incorporating more diverse or recent signals into their recommendation logic. This platform divergence implies that optimization strategies may need to be tailored to specific LLM ecosystems as they mature.
While our findings are statistically significant, several limitations warrant consideration:
The study focused on B2B software categories; patterns may differ in other industries.
We captured a snapshot in time; LLM algorithms continue to evolve rapidly.
Correlation does not imply causation; backlinks may be proxies for other unmeasured quality signals.
Future research should expand to additional industries, track changes over time, and investigate causative mechanisms through controlled experiments. Additionally, exploring the interplay between backlinks and other potential ranking factors such as content structure, user engagement metrics, and real-time data could provide a more complete picture of LLM ranking dynamics.
As we transition from an era of traditional search to AI-mediated discovery, this study offers reassuring continuity for digital marketers: the fundamental principle that external validation matters remains intact. However, the nuanced differences in how LLMs interpret and weight these signals suggest we’re entering a more complex landscape where success requires both maintaining traditional SEO best practices and adapting to new AI-specific optimization opportunities.
The strong correlation between backlink metrics and LLM rankings indicates that building authentic, diverse link profiles remains a cornerstone of digital visibility. Yet the presence of notable outliers and platform variations reminds us that the age of AI search will reward those who combine time-tested authority building with innovative approaches to content creation and distribution. As LLMs continue to reshape how users discover and evaluate software solutions, businesses that understand and adapt to these evolving ranking dynamics will maintain their competitive edge in the AI-first future.