agentificial
case study · anonymised
·Diagnosis· 2026

64% of apparent email opens were Apple Mail prefetches, and link scanners inflated clicks

Cold-email dashboards at a medical-tourism group showed opens and clicks that looked like engagement. Splitting them by cohort showed that most were machines.

455 of 711
apparent human opens were Apple Mail Privacy Protection (64%)
0.52% vs 8.33%
click-through: machine-opened mail vs identifiable mail clients
< 10
humans behind 437 recorded clicks

The problem

Open rate is the metric most cold-email tools steer by. At the group, raw open rates of 2.3–3.8% looked like a deliverability signal worth optimising.

Apple Mail Privacy Protection fetches images on a recipient's behalf, and security scanners follow links, so a recorded open or click can be a machine.

What we built

  • A cohort analysis of the campaign database that separates opens by user agent: Apple's prefetch proxy (a bare "Mozilla/5.0") against identifiable mail clients.
  • Click-through measured per cohort, plus a filter for link-scanner clicks.
  • A written runbook for recognising bot opens and clicks, so campaign results are read from the cleaned numbers.

Result

  • Of 711 opens that looked genuine, 455 (64%) were Apple Mail Privacy Protection prefetches.
  • Raw open rates of 2.3–3.8% corresponded to a real human rate of 0.6–1.4%.
  • Mail opened by Apple's proxy clicked at 0.52%; identifiable mail clients clicked at 8.33%.
  • Link scanners turned 437 recorded clicks into fewer than 10 humans.

Scope and limits

  • This describes one client's sending programme, from its campaign database. It is not a statement about email in general.
  • We did not measure any revenue effect. The finding is that the dashboard metric overstated engagement.
· Could be you next

Same problem,
different stack.

Tell us the workflow that should already be done by the time your team logs in. We'll come back with a one-page assessment.