AI-visibility vendor Profound tested 1,724 identical prompts across Claude and Claude Code between July 13–23, 2026, generating 24,135 total responses, then tracked which pages each product's agent visited most from July 18 to August 18, 2026. Claude Code triggered a live web search in only 13% of its responses versus Claude's 93%, and on identical prompts the two products overlapped on cited brands only about 20% of the time. Search Engine Journal covered the report on August 31, 2026 — operators building or refreshing an H2 AI-visibility tracking plan should fold Claude Code in as its own monitored surface this week, not a subset of "Claude."
Sources:
Search Engine Journal: "Claude Searches Web Differently From Claude Code, Data Shows," by Matt G. Southern, published 2026-08-31 — https://www.searchenginejournal.com/claude-code-rarely-searches-web-compared-claude-data/587584/
Profound: "Claude and Claude Code Are Distinct Answer Engines" — https://www.tryprofound.com/blog/claude-and-claude-code-are-distinct-answer-engines
Credibility: MEDIUM. Single-vendor study — Profound sells AI-visibility monitoring, the exact category this report supports — with a disclosed sample size (1,724 prompts, 24,135 responses) and a checkable method at the source blog, but no independent outlet has replicated the numbers and Anthropic has not confirmed or commented on the underlying cause.
Fact bullets:
- Profound tested 1,724 identical prompts across Claude and Claude Code, July 13–23, 2026, generating 24,135 total responses.
- Claude Code triggered a web search in 13% of responses; Claude triggered one in 93% — a roughly 7x gap.
- On identical prompts, Claude and Claude Code overlapped on cited brands only about 20% of the time; Claude Code cited 6.6 brands per response on average versus Claude's 5.2.
- In a second dataset tracking page visits from July 18–August 18, 2026, 73% of Claude Code's visited pages were docs/informational/pricing pages versus 5% for Claude; 60% of Claude's visited pages were robots.txt/sitemaps/homepages versus 4% for Claude Code.
Stay ahead of AI search
The changes that move your rankings and AI citations — and the exact move to make — before they cost you. Free, three times a week.
For teams working on AI visibility, "Claude" is not one surface to monitor. The consumer-facing Claude chat product and the developer-facing Claude Code tool pull from the live web at very different rates and land on different kinds of pages once they do. A brand tracked only through Claude's citation pattern will badly undercount — or simply mismeasure — its footprint with Claude Code's users, who are reading documentation and pricing pages directly rather than getting broad web-search results.
Recommended actions:
- Ask your AI-visibility tracker (or vendor) to report Claude and Claude Code as separate rows, not a blended "Claude" total, so citation and mention counts aren't mixed across two products with 7x different search rates.
- Have your web team confirm product documentation and pricing pages are indexable and current — Claude Code draws 73% of its visits to exactly those page types, so a stale docs page costs visibility there specifically.
- Record the decision on how much reporting weight Claude Code gets relative to Claude for your account — it's a smaller share of general AI-chat traffic today, but its documentation-heavy visit pattern makes it a distinct, high-intent surface for products with technical buyers.
Search Engine Journal: "Claude Searches Web Differently From Claude Code, Data Shows," by Matt G. Southern, published 2026-08-31 — https://www.searchenginejournal.com/claude-code-rarely-searches-web-compared-claude-data/587584/
Profound: "Claude and Claude Code Are Distinct Answer Engines" — https://www.tryprofound.com/blog/claude-and-claude-code-are-distinct-answer-engines