How many competitors do LLMs mention in their answers

Based on our study of 32,961 queries, LLMs mention an average of 0.81 to 2.5 competitors per answer, with Gemini quoting more competitors than SearchGPT and Perplexity. This represents a shift from traditional search where ranking is no longer about appearing in the top 10 results, but about being explicitly mentioned as a brand in the LLM's response.

Which LLM gives the longest answers

Gemini provides the longest answers with an average of 2,955 characters (approximately 490-590 words), followed by Perplexity with 2,029 characters (340-410 words), and SearchGPT with 1,686 characters (280-340 words). If you want to be present on SearchGPT, you should be as concise as possible in your content.

What sources do ChatGPT and SearchGPT prefer

SearchGPT most frequently cites Wikipedia, Google services (like Google Maps), French newspapers with ChatGPT partnerships (like Le Monde), and comparator websites like TrustPilot. The study analyzed over 3,557 unique domains quoted by SearchGPT across 32,961 queries.

What websites does Gemini use as sources

Gemini heavily favors Google-owned domains including Google.com services and YouTube as its top sources. It also frequently cites French public service websites (reflecting Google's YMYL focus), Reddit, and independent media sites like lecoindesentrepreneurs.fr and Studyrama. The study found 13,388 unique domains quoted by Gemini.

How to optimize content for LLM ranking

To optimize for LLM ranking, use structured content with bullet points (present in 99% of LLM answers), mirror LLM language patterns like impersonal phrasing and conditional forms, include common expressions like 'by following these steps' (found in 7.61% of SearchGPT answers), and create clear, easily digestible formats. Also consider publishing video content on YouTube, especially for Perplexity visibility.

What is the difference between SearchGPT Gemini and Perplexity sources

Each LLM has distinct source preferences: SearchGPT favors Wikipedia and news sites with partnerships, Gemini prefers Google-owned domains and authoritative public sources, while Perplexity uses more diverse community-driven content with YouTube as its most-cited source. The study found 6,037 common sources between Perplexity and Gemini, but only 1,383 between ChatGPT and Gemini.

Do LLMs prefer recent content or older content

The preference varies by LLM. Perplexity shows a strong preference for fresh content with '2025' being the dominant year in URL slugs, indicating it favors current information. In contrast, ChatGPT URL analysis showed many sources from 2023, suggesting it may rely more on training data from that period.

What phrases do AI chatbots use most in answers

The most common phrases used by LLMs include 'by following these steps' (7.61% of SearchGPT answers), 'it is important to' (most common in Gemini), and cautionary language using conditional forms. SearchGPT uses impersonal forms in 17.8% of answers, incentive verbs in 25.2%, and imperatives in 25.3%. Gemini shows higher percentages with impersonal forms in 38.42% of answers.

NEWQwairy v1.16: Smarter Metrics, Deeper Insightsv1.16: Smarter Metrics

Pricing
Compare
Blog
55

Optimize your visibility in the AI era with advanced Generative Engine Optimization.

Platform

Cockpit
Monitor
Act
Analyze
Optimize
Measure

Product

Pricing
Documentation
API/MCPNew
Changelog
AffiliatesNew

Resources

Blog
GEO Guide
GEO Glossary
AI Crawlers Guide
Help Center

Company

About Us
Book a Demo
Privacy Policy
Terms of Service
Legal Notice

GDPR Compliant

🇪🇺EU Data Hosting

Made with ❤️ in France 🇫🇷

32,961 Queries LLM Study: SearchGPT, Gemini, Perplexity | Qwairy

Home
Blog
32,961 Queries Study - Q2 2025

GEO

SearchGPT

Gemini

Perplexity

AI Search

Q2 2025

We analyzed 32,961 queries on SearchGPT, Gemini and Perplexity - Q2 2025 Study

Luca Fancello•June 16, 2025•Updated Dec 21, 2025•16 min read•

Research

Q2 2025 study on LLM ranking factors based on 32,961 queries across three LLMs. For the latest insights, see our Q3 2025 study with 184,128 queries.

Summarize with AI