Inside ChatGPT's Search Behaviour: What 10,000 Queries Tell Us

After analysing 10,000 queries extracted across dozens of topics and prompt styles, clear patterns emerge about how ChatGPT formulates its web searches. Here's what the data shows.

Finding 1 — Average Query Length is 4.2 Words

ChatGPT favours medium-length queries — not single keywords, not full sentences. The sweet spot is 3–6 words, mirroring how experienced human searchers use Google. These queries have high practical SEO value.

Finding 2 — Four Dominant Intent Types

• 46% Informational — "how does X work", "what is Y"
• 28% Comparative — "X vs Y", "best tools for Z"
• 14% Navigational — brand or site-specific searches
• 12% Commercial — pricing, reviews, recommendations

The high comparative percentage maps directly to high-converting "best of" and "vs" content formats.

Finding 3 — Prompt Framing Shifts Query Type

Prompts starting with "compare" or "which is better" produce 3x more comparative queries. Prompts with "latest statistics" produce informational queries with year references — excellent for featured snippet targeting.

Finding 4 — Niche Depth Increases Query Quality

Generic prompts produce broad, high-competition queries. Specific prompts produce niche, low-competition queries with clear purchase intent. Specificity is a multiplier.

Finding 5 — Optimal Batch Size is 40–60 Prompts

Sessions under 40 prompts don't generate enough diversity. Sessions over 80 show 30%+ query repetition. The 40–60 range gives maximum unique query yield per session.