25x Growth in LLM Traffic in 3 Months

Between January and April 2025, OpenArt saw a 25x increase in LLM-referred sessions. This jump in LLM-referred traffic wasn’t a lucky break, it was the result of designing content that matched user intent and structuring it in a way that both traditional search engines and AI models could easily interpret, cite, and trust. Here’s how we did it. 

Built for traditional search, but ready for LLMs

In early 2024, OpenArt partnered with daydream to launch an organic growth strategy designed to capture long-tail, high-intent queries related to generative art. We mapped out how people search for creative AI tools and found two major patterns:

• AI [Topic] Generator (e.g., Pixar-style generator, anime character creator)
• AI [Effect or Feature] (e.g., AI background remover, AI hair color changer)

To serve this demand at scale, we built a programmatic content engine with:

• Thousands of product-led pages tailored to specific use cases
• Clear, descriptive copy designed to resolve search intent directly
• Metadata and structured markup optimized for discoverability

This content quickly gained traction on Google, but we also discovered it was performing surprisingly well in LLMs.

Early LLM traction validates our intent-aligned, product-led content thesis

Our thesis was straightforward: if you build structured, product-led content that directly answers high-intent queries, it will perform well across organic channels. 

By January 2025, when daydream began formally tracking LLM performance, OpenArt’s programmatic pages were already generating about 7K monthly sessions from sources like ChatGPT, Perplexity, Claude, and Copilot.

That early signal led to a deeper analysis: Why was this content performing so well in AI search? We identified two forces driving early success:

1. Intent alignment. The content wasn’t just relevant, it answered the user’s question. LLMs prioritize clarity, structure, and utility, especially for task-based queries. Because OpenArt’s pages were product-led and action-oriented, they aligned closely with what LLMs are designed to surface: direct, high-confidence answers. 

These “middle-of-funnel” pages (e.g., guides or tools with clear CTAs) are also more resilient to AI summarization, meaning LLMs are more likely to link to them directly rather than paraphrasing or omitting them.

2. SEO fundamentals. LLM visibility starts with traditional SEO fundamentals. Platforms like ChatGPT, Claude, and Gemini still rely heavily on search engine indices—Bing, Brave, and Google, respectively—to retrieve content. That means technically sound, high-authority content is far more likely to be included in LLM outputs. Because OpenArt’s content ranked well in traditional search, it was more likely to be included in LLM response pipelines.

A spike in AI image generation: predicted and captured 

In mid-March 2025, a sudden wave of interest in Studio Ghibli-style imagery swept across generative art communities. OpenArt was ready. With the “AI [Topic] Generator” and “AI [Effect or Feature]” framework already in place, daydream was able to spin up relevant pages almost instantly to capture the surge in search and prompt activity tied to the trend.

Two pages in particular exploded:

• /features/studio-ghibli-filter → 6.1M sessions
• /generator/studio-ghibli → 1M sessions

Overall, LLM traffic peaked in April 2025 at nearly 170,000 sessions/month, up from just 6,600 in January, a 25X increase. 

Sustained performance after the spike

As the cultural moment passed, traffic naturally leveled off. But OpenArt’s performance across organic channels remained strong. As of May 2025, OpenArt’s long-tail, product-led content strategy continues to drive around 19k LLM-referred sessions per month.

Results Snapshot

The takeaway: Generative engine optimization (GEO) is a layered expansion of SEO

This case proves that SEO and LLM visibility aren’t opposing goals; they’re mutually reinforcing. When you build structured, product-led content that meets real user needs, you’re not just optimizing for one algorithm. You’re building a foundation that performs across every discovery layer: Google, ChatGPT, Perplexity, and whatever comes next. 

If you’re ready to start optimizing your growth strategy for LLMs, let’s chat.