Dan Petrovic, director of Dejan, conducted an extensive analysis of 3.9 million conversation turns – covering 613 million words and 4.4 billion characters – to uncover patterns in real-world AI use. The results challenge many assumptions in AI content strategy.
Usage patterns: Most chats are short and task-oriented
- Median chat duration: 2 turns (a single question and answer)
- Median session length: 430 words
- More than 80% of chats contain less than 1,000 words
- Only 4.2% of sessions exceed 2500 words, typically involving editing, summarizing, guiding, coding or data analysis
- Average session length: 732 words, biased by long document submissions
- Assistant output volume: approx 1.5 times that of user input
Petrovic noted that user contributions are usually just responsible 16–17% of the session, meaning the assistant generates most of the content.
Dig deeper: what happens when no one clicks anymore
Intent Analysis: Commercial use is limited
Petrovic analyzed 24,259 secret sessions across 42 intent categories. He found that 64.6% of sessions had no commercial intent. Users mainly turned to AI for tasks such as:
- Brainstorming (7.7%)
- Planning (6.5%)
- Emotional support or conversation (6.2%)
- Analysis (5.7%)
- Learning (4.7%)
- Text transformation, including summaries and translations (4.6%)
- Content creation (3.9%)
Even of the 35.4% of sessions that showed commercial intent, the majority were early in the funnel:
- Consciousness phase: 10%
- Consideration: 8.5%
- Discovery and decision support: 6.9% combined
- Transactional support: 4.8%
- Post-purchase support: 5.1%, including troubleshooting and usage guidance
Key insight: AI conversations are cognitive workflows, not questions
Many marketers and SEOs optimize content with a search-first mentality, assuming that AI chats will mimic keyword-based searches. Petrovic’s findings suggest otherwise. AI assistants are more often used to support multi-step tasks, rather than making direct purchases.
“The use of AI chat is predominantly non-commercial,” Petrovic said. “Users treat assistants as co-pilots, not as salespeople.”
Implications for AI content strategy
Marketers must adapt their approach to AI optimization by:
- Prioritize content that supports awareness and early funnel exploration
- Creating structured, high-context information that agents can reuse in longer workflows
- Avoiding overemphasis on transactional keywords in AI-focused content
Bottom line: The future of AI content visibility isn’t about game intent, it’s about fulfilling it. Assistants become tools for cognition, not just conversion. If your content helps users learn, plan, write, or evaluate, it is much more likely to show up in real AI use.
The report. How do people use AI assistants?
Energize yourself with free marketing insights.
MarTech is owned by Semrush. We remain committed to providing high-quality reporting on marketing topics. Unless otherwise stated, the content of this page was written by an employee or paid contractor of Semrush Inc.
#content #strategies #focus #tasks #transactions #MarTech