What 25,000 URLs reveal about citations

Large language models (LLMs) excel at synthesizing enormous amounts of information into personalized responses to plain-language prompts. These responses draw on massive training datasets and are often enhanced with internet searches. The fastest way to influence what LLMs say about your brand is to influence the content they retrieve through those searches.

At Evertune Research, we use the Evertune AI marketing platform to track hundreds of brands across 250 categories across every major LLM. This gives us clear insight into which content AI models cite most often, especially when users ask for brand or product recommendations across industries.

For this analysis, we reviewed the 6,000 most-cited URLs per model across ChatGPT, Copilot, Gemini, Google AI Mode, Google AI Overview, and Perplexity for March and April. We found that these models share a key behavior: they heavily cite listicles.

Half of LLMs’ most-cited URLs are listicles

Of the roughly 25,000 unique URLs we reviewed, half were listicles. Across nearly 400 million citations from all models, 63% pointed to listicles.

Listicles have many qualities that make them ideal for models’ consumption. 

• They’re tightly focused on a single topic, like “best laptops for gamers,” which makes them highly relevant to user prompts. 
• Their structured format also makes the content easy for models to parse and reproduce. 
• For brand-related queries, listicles do much of the work for LLMs by comparing products head-to-head on features, price, materials, and more—a format ChatGPT now features prominently in its shopping widget.

Listicles were pervasive across every model we reviewed. They accounted for 40–65% of the most-cited URLs, with Copilot at the low end and Gemini at the high end.

The vast majority of listicles in our analysis featured ranked lists, such as “Top 5 CRM Tools.” Depending on the model, these made up 71% to 86% of listicles. Unranked lists, such as “7 Ways to Save on Groceries,” were a distant second. Institutional rankings (e.g., data-heavy lists like U.S. News & World Report’s Best Colleges rankings) accounted for just 1.4% to 4.7% of listicles.

Corporate, earned media, and affiliate domains were the top sources of listicles in our analysis. It’s worth noting, however, that individual pages may contain affiliate content even when the broader domain does not. 

• For example, Forbes.com is an earned media domain, but it includes affiliate segments such as Forbes Advisor and Forbes Vetted. It ranked among the top three sources on every model for listicles in our URL dataset.

A word of warning before making listicles the foundation of a GEO strategy: Google has already signaled its intent to crack down on promotional listicles. Simply ranking your own brand No. 1 alongside competitors could also run afoul of a Federal Trade Commission rule that “prohibits a business from misrepresenting that a website or entity it controls provides independent reviews or opinions about a category of products or services that includes its own products or services,” among other prohibitions.

URLs that thrive on multiple models

We reviewed the 6,000 most-cited URLs across six LLMs, which in theory produced a pool of 36,000 URLs. In practice, the dataset contained about 25,000 unique URLs, since many appeared among the most-cited results across multiple models.

Among the models, the three Google Gemini-powered models — Gemini, AI Mode, and AI Overviews — showed the highest overlap. More than half of Google AI Mode’s most-cited URLs also appeared among Google AI Overviews’ most-cited URLs. Gemini likewise shared a large portion of its top-cited URLs with both Google AI Mode and Google AI Overview.

The remaining models also shared the most URLs with Google AI Mode and Google AI Overviews, though the overlap was much smaller. Perplexity shared more than 20% of its URLs with both models, while ChatGPT shared more than 15% with each. 

Given the thousands of URLs models cite on any topic, that still represents meaningful overlap. Copilot, by contrast, shared just 4% to 6% of its URLs with any other model.

The URLs that models cite most deviate for many reasons, including model training, sites’ crawl permissions and other factors. Traditional SEO that moves content higher in search results, no matter if the search is by a bot or a person, also plays a role, especially for Google AI Mode and Google AI Overview.

Page components of heavily cited URLs

Our review of the roughly 25,000 URLs heavily cited by LLMs found that these pages typically ranged from 1,000 to 2,000 words, averaged 18 words per sentence, linked frequently, and used structured headings (H2s and H3s) throughout.

Copilot favored the most concise content, typically citing pages with 964 words and 24 paragraphs. Gemini skewed more verbose, typically citing pages with 1,977 words and 53 paragraphs.

Although there’s no cookie-cutter formula for success in AI visibility, we found that the most-cited pages typically included the following components:

GEO takeaways

Each LLM has its own preferences and quirks, and a strong GEO strategy accounts for them. But our analysis of more than 25,000 URLs suggests that some GEO best practices can improve brand visibility and sentiment across models.

• All LLMs cite large volumes of highly structured, hyper-specific content, which listicles exemplify. Avoid spammy, self-promotional listicles that Google penalizes, but otherwise aim to create and appear in lists where relevant.
• Traditional SEO supports GEO. Pages that perform well in human search results also tend to perform well in bot-driven searches. This is especially true for Gemini-based models.
• Pay attention to the page structures most often cited by the model you want to target. Copilot tends to favor brevity, while Gemini responds better to more expansive content. In general, keep pages under 2,000 words, use frequent links, apply strong structure, and include images and lists when relevant.