LAB REPORT / 005 · 2026-10-08
Cloudflare edge requests vs real visitors: a 7-day audit
Verdict: Different metrics, not interchangeable measures of human traffic.
A real seven-day comparison of HTTP requests, browser RUM and AI crawlers on NewLab.si, with sampling and testing caveats.
Question and answer
Can Cloudflare edge requests be treated as qualified human traffic?
No. At approximately 2026-10-08 15:04 UTC, seven-day edge analytics reported 2,714 HTTP requests, while browser Web Analytics reported 47 visits and 49 page loads. The sources count different events. Launch and test traffic is included; neither number establishes unique people, organic search visits or conversions.
Observed measurements
- 2,714 — edge HTTP requests / 7d
- 47 — browser RUM visits / 7d
- 49 — RUM page loads / 7d
- 71 — ClaudeBot requests / 7d
Procedure
- Query two different Cloudflare sources — Read the seven-day edge summary for HTTP requests and the seven-day Web Analytics RUM summary for browser visits and page loads, around 15:04 UTC on 2026-10-08.
- Inspect named crawler traffic — The same seven-day crawler view recorded 71 ClaudeBot requests and six GPTBot requests, plus Googlebot and OAI-SearchBot activity. Bot access does not prove indexing or AI citations.
- Check sampling and internal testing — The 30-day RUM query returned 30 raw visits with sample_interval=10, versus 47 seven-day visits with sample_interval about 1.04. Do not compare raw counts across these differently sampled windows. Our own browser tests may be included.
Findings
Edge requests measure infrastructure activity
Assets, HTML, bots and scanners all produce requests. One browser visit may generate many requests; requests are not unique visitors.
RUM is closer to browsers, not verified humans
Privacy settings can block RUM, while internal QA and automated browsers may still execute it. No independently verified human or organic conversion count was available.
Failure modes
Confusing requests with people
Treating 2,714 as visitors would inflate audience size and reward crawler noise.
Reading sampled windows as a trend
The 30-day raw RUM count being lower than seven days is a sampling artifact, not evidence of audience decline.
Treating crawls as search success
Crawler requests do not prove a page is indexed, ranked, cited, recommended or clicked. Search Console and referrer evidence are needed.
Evidence and limits
- Source
- Cloudflare edge analytics + Web Analytics (RUM)
- Seven-day window
- 2026-10-01 → 2026-10-08 UTC
- Query time
- 2026-10-08 ~15:04 UTC
- Edge total
- 2,714 HTTP requests
- RUM totals
- 47 visits · 49 page loads
- 30-day sampling caveat
- 30 raw visits · sample_interval=10; not comparable
- Verified humans
- Unknown; internal testing included