LLM Product Discovery Benchmarks
Introduction
The way consumers discover brands is shifting faster than ever. Large Language Models (LLMs) like ChatGPT, Claude, Gemini, Grok, and Perplexity are no longer just tools for conversation, they are becoming powerful engines for product discovery.
In Q2, Fairing data highlighted a surge of early consumer adoption with more than a tenfold increase in mentions of LLMs in “How Did You Hear About Us?” surveys since January. At that time, about 15% of brands were seeing any LLM mention, and those mentions were still relatively small fractions of attribution.
By Q3, the number of brands with mentions has grown modestly, from 15% to 18%, but within those brands, the LLM attribution share of responses has deepened dramatically. 50% of brands now see at least 0.5% of responses tied to LLMs, and 35% see 1% or more, compared to only 13% reaching the 0.5% LLM response share threshold in Q2.
LLMs are moving from novelty to habit. Consumers who try them for discovery keep returning with August marking the highest peak yet for LLM-driven brand mentions. Bain research confirms the broader trend, with ChatGPT usage up 70% in the first half of 2025 and shopping queries doubling in popularity over the same period. For marketers, the question is no longer if LLMs matter, it’s how to measure and act on them.
Key Findings
In August, LLM-driven discovery hit its highest levels to date, signaling habitual reliance among consumers.
ChatGPT still dominates, but Claude, Gemini, Grok, and Perplexity together account for nearly a third of responses, signaling a more competitive ecosystem.
Category specialization is clear: ChatGPT leads in lifestyle sectors, Claude in Consumer Electronics, Grok in Office/Food & Drug, and Perplexity in Automotive.
The number of brands with mentions grew modestly, but the share of attribution within those brands increased dramatically, proving depth is replacing breadth as the growth driver.
The attribution gaps remain massive, with nearly 75% of LLM-driven transactions lacking last-click data.
The share of brands seeing at least one LLM mention grew from 15% in Q2 to 18% in Q3. The slope is steady, not explosive.
Breadth is growing slowly, but if your brand is already seeing LLM mentions, you’re in a group where adoption is sticking. Don’t expect it to fade, expect it to deepen.
In Q2, only 13% of brands with LLM mentions saw 0.5% share of response attribution to LLMs. By Q3, 50% crossed that line, and 35% saw at least 1% of share responses attributed to LLMs.
The long tail remains, but if your brand sees even a fraction of mentions today, history suggests that share will grow.
Mentions per brand kept rising in Q3, with August setting a new high at ~0.47% per brand. The number may sound small, but the curve is steep, almost doubling in just a quarter.
Consumers who use LLMs for discovery are doing so more often. This is habit formation, not hype.
Weekly responses peaked sharply in August, the strongest signal yet. ChatGPT remained the bulk of the surge, but Claude and Gemini contributed consistent shares, while Grok and Perplexity added meaningful secondary volume.
ChatGPT may dominate, but discovery is multi-platform which means your brand presence must be optimized beyond just one LLM.
Category-level specialization became even clearer in Q3.
ChatGPT: Dominant across all categories with the exception of Automotive.
Claude: Strongest in Consumer Electronics (29%).
Grok: Notable in Food & Drug (16%), Office (15%) and Apparel (13%).
Perplexity: Highly concentrated in Automotive (93%).
LLM discovery is not a monolithic channel. The winning platform for you depends on your vertical and your winning strategy must be category-specific.
Nearly 75% of transactions tied to LLM discovery lack last-click attribution.
The measurement gap is still massive. Even as reliance on LLMs deepens, last-click attribution data remains unknown. Consumers may read recommendations in ChatGPT or Claude but search for the brand later, leaving brands to believe the conversion happened via Google. That makes it difficult to understand true LLM performance without an alternative measurement approach like attribution surveys.
The Maturation of LLM Discovery
Between Q2 and Q3, the story of LLM discovery shifted from proof of concept to reliable channel. Breadth has grown modestly, but depth has increased noticeably. Consumers are using LLMs habitually across platforms and relying on them differently by category.
For marketers, the implications are clear:
Treat LLMs as core discovery channels, not fringe experiments.
Optimize per platform and per category.
Close attribution gaps to properly capture and invest in this growing source of acquisition.
LLMs are no longer a signal on the horizon, they are shaping discovery today, and the brands that adapt first will lead tomorrow.
