Security gateways click every link in every email before humans see them. How scanner traffic pollutes click metrics, breaks one-click flows, and how to filter it.
Apple's MPP taught senders that opens lie. The B2B equivalent has been quietly true for longer: clicks lie too. Corporate security gateways (Microsoft Defender for Office 365, Proofpoint, Mimecast, Barracuda and peers) fetch and often fully execute every link in every inbound message at delivery time, checking destinations for phishing before a human can be harmed by them. The result in your analytics: click events from recipients who never touched the message, conversion pixels firing on scanner visits, and, at B2B-heavy lists, click rates that can run several times the human reality. This article is about seeing through it.
Why scanners click
Modern filtering does not trust a URL's reputation at delivery time alone, because attackers arm links after delivery. So gateways do two things: detonate links in sandboxes when the message arrives, and rewrite URLs so every eventual human click routes through the vendor's checking infrastructure first. Both behaviours touch your tracking domain. The delivery-time detonation is the click you see seconds after the send with no human involved; the rewrite means even genuine clicks arrive at your servers from vendor infrastructure first, complicating naive IP-based filtering in the other direction.
What the pollution does
Three failure classes. Metrics: click rates and click-to-open ratios inflate, A/B tests resolve on scanner behaviour rather than human preference, and engagement-based segmentation promotes dead recipients whose gateway is very diligent. Automation: click-triggered journeys ("clicked pricing, send the follow-up") fire for scanners, mailing humans about interest they never showed. And state: single-use links get consumed, password reset tokens expire before the human arrives, and GET-based unsubscribe links unsubscribe people whose security software merely inspected the mail, the exact bug RFC 8058's POST requirement exists to prevent.
Filtering machine clicks
A click-classification pipeline
- 1
Capture the evidence per click
Log timestamp relative to delivery, user agent, source IP and ASN, and which links were hit. Classification is impossible from a bare count.
- 2
Flag the mechanical patterns
Clicks within seconds of delivery, all-links-in-order patterns, known scanner user agents, and data-center ASNs classify as machine. Vendors publish some ranges; behaviour catches the rest.
- 3
Use a canary link
An invisible or clearly non-human link (styled out of sight, or labelled for bots) that no person would click. Any recipient whose click set includes the canary is a scanner for that message.
- 4
Report both numbers
Raw clicks for continuity, filtered clicks for decisions. Journeys, scoring, and A/B tests consume the filtered stream only.
- 5
Protect destructive endpoints
Anything that changes state (unsubscribes, confirmations, one-time tokens) requires a POST or an explicit human step on the landing page. GETs must be safe to replay, always.
What this means for engagement policy
Between MPP inflating opens and scanners inflating clicks, the naive version of engagement-based sunsetting has lost both its inputs at B2B lists: a recipient can register opens and clicks for years after leaving the company, as long as the mailbox forwards and the gateway scans. The robust policy uses filtered clicks plus downstream evidence (site sessions, conversions, replies) and treats mailbox-only signals with suspicion proportional to the audience's corporate share. Meanwhile the same filtering logic that cleans your metrics also cleans your automations: no journey should fire on a click your classifier calls mechanical.
None of this is the security vendors misbehaving; they are protecting recipients from exactly the link-borne attacks the rest of this publication documents. The scanner traffic is the cost of mailing protected audiences, and like every measurement distortion before it, it punishes only the senders who never look underneath their numbers.
Frequently Asked Questions
Can I just filter clicks arriving within 10 seconds of delivery?
Do scanner clicks affect deliverability or reputation?
Why do my rewritten links (safelinks etc.) complicate things further?
Does this affect one-click unsubscribe?
Key Takeaways
- Security gateways click every link at delivery time, inflating B2B click metrics and firing automations for recipients who never engaged
- Classify clicks with timing, user agent, ASN, all-links patterns, and a canary link no human would touch
- Decisions, journeys, and A/B tests consume filtered clicks; raw clicks remain for continuity only
- State-changing endpoints need POSTs or a human step; GETs must be safe to replay
- With opens and raw clicks both corrupted, B2B engagement policy leans on filtered clicks plus site and conversion evidence
Related articles
Half-Year Review: Email in the AI Inbox Era
Six months that rearranged the reading layer: Gemini in Gmail, Microsoft rejecting outright, DMARC finally a standard. What the first half of 2026 means for senders.
Backscatter: The Bounces You Never Sent
When spammers forge your domain as their envelope sender, the bounces come to you. What backscatter is, why it happens, and how to stop causing and receiving it.
Per-Provider Playbooks: Operating Gmail, Microsoft, and Yahoo Differently
The big three score senders differently, expose different data, and fail in different ways. One operating playbook per provider beats one averaged strategy.