I recently discovered something fascinating about how people interact with AI. It turns out most AI chats don’t have any commercial intent! This insight came from a thorough analysis by Dan Petrovic, the director of AI SEO agency Dejan, who scrutinized millions of conversational turns to shed light on actual AI assistant usage.
Why is this important to us? As someone involved in SEO and marketing, I’m often focused on optimizing for AI. However, Petrovic’s research suggests we might be misunderstanding how people genuinely engage with AI assistants. They don’t typically flood AI with purchase queries. Instead, they explore issues and weigh options.
By the numbers, Petrovic dived into 4.4 billion characters across 613 million words and 3.9 million conversation turns. Here’s what that looks like:
- Median chat: Just 2 turns, usually involving a quick question and an immediate response.
- While most interactions are short, there are lengthy sessions when users paste documents for summarization or analysis.
- Median words per session: 430 words.
- Astonishingly, more than 80% of chats contain fewer than 1,000 words.
- Only a small fraction, 4.2%, exceed 2,500 words. These are often complex tasks, like editing, coding, or tutoring.
- Mean words: 732. This statistic is heavily influenced by long document submissions.
- Assistant output: Typically, it’s 1.5 times more than what users contribute.
- Median user contribution: Users make up about 16-17% of the conversation.
In exploring how people utilize AI assistants, Petrovic examined 24,259 sessions across 42 intent categories. Surprisingly, 64.6% of chats didn’t align with any purchase funnel. People used AI for writing, brainstorming, planning, learning, analyzing, or just simply chatting. Here’s the breakdown:
- Other: 25%
- Included are jailbreak attempts, role-playing, and specific requests.
- Brainstorming: 7.7%
- Planning: 6.5%
- Conversation / emotional support: 6.2%
- Analysis: 5.7%
- Learning: 4.7%
- Transformation (summaries, translations): 4.6%
- Creation (writing, code, docs): 3.9%
Only 35.4% of chats showed any commercial intent, and most were in the early stages of the buying process. Other insights:
- Awareness (10%) and consideration (8.5%) combined to form 18.5%, which Petrovic noted as prime territory for product content.
- Post-purchase needs (5.1%) outpaced transactional support (4.8%), discovery (4.1%), and decision support (2.8%), suggesting users seek AI more for ‘How do I use or fix this?’ rather than ‘Should I buy this?’
Bottom line, my takeaway is that AI assistants are utilized far more for creation, cognition, and conversation than for commerce.
If you’re keen to dive deeper into the findings, check out the full report titled How do people use AI assistants?
Inspired by this post on Search Engine Land.
Leave a Reply