Gmail Image Proxy affects clicks, not just opens

Most marketers know the Gmail Image Proxy story: every external image in a Gmail inbox is fetched throughgoogleusercontent.com, cached, and served to the user. This means the open pixel fires once at delivery time (not when the user opens), and every subsequent open hits Google's cache, not your server. Open rates became nearly useless for Gmail around 2013.

What is less well known is that Google has been expanding the proxy concept beyond images. Link pre-fetching for spam detection, phishing protection, and user experience has crept into the Gmail pipeline. The net effect on click metrics is smaller than SafeLinks, but it is real, and the industry underestimates it.

What the Image Proxy actually does

When Gmail delivers a message, it scans every <img src>tag and rewrites it. The rewritten URL points toci<n>.googleusercontent.com/proxy/... and includes an encrypted pointer to the original. Gmail then fetches the image from your CDN, caches it, and strips headers. This has three effects:

• You see one IP address family (Google's) instead of the user's.
• You see one User-Agent (GoogleImageProxy) instead of the user's mail client.
• The fetch happens at delivery time, often before the user is even on a device. Your "open" is really a "delivered".

The lesser-known link pre-fetch

Gmail's anti-phishing system does crawl some links. Not every link in every email — that would be prohibitive — but a meaningful subset:

1. Links in messages flagged by heuristics — if the spam classifier is uncertain, Gmail may fetch the destination to check final landing page content.
2. Shortened URLs and redirect chains — bit.ly, t.co, tracking wrappers. Gmail resolves them to the final URL to evaluate reputation.
3. Preview cards in the chat and Inbox — when Gmail shows a link preview, it fetches the HEAD or GET of the URL to pull OpenGraph data.
4. Priority Inbox and Smart Reply training — machine-learning pipelines may sample links for classification.

Measuring the noise

The pre-fetch rate on marketing mail is low — we measured between 0.4% and 1.8% of unique recipients, depending on sender reputation and list hygiene. New domains and cold outreach lists hit the higher end because Gmail is more likely to sample them.

That sounds small. But consider: on a 100,000-recipient send with a baseline 4% CTR, an extra 1% of pre-fetched URLs adds 1,000 clicks to your 4,000 real clicks — a 25% relative inflation on Gmail. And you will not notice unless you actively look.

Distinguishing real clicks from proxy fetches

The User-Agent and IP filter from the SafeLinks article works for Gmail too, with different values. Minimum viable filter:

# Block these User-Agent substrings from your CTR:
GoogleImageProxy
Google-Safety
Google-InspectionTool
Google-Read-Aloud
Googlebot
GoogleOther

# And these IP ranges (abbreviated):
66.102.0.0/20
66.249.64.0/19
72.14.192.0/18
74.125.0.0/16
209.85.128.0/17
2607:f8b0::/32

Gmail does not publish a clean JSON service tag list the way Microsoft does, but Google Workspace documentation lists the main mail-related ranges, and reverse DNS on .google.com is a reliable fallback.

Why the image proxy also helps you sometimes

It is not all bad news. The fact that Gmail caches your images aggressively means your CDN load is cheap, and the proxy protects user privacy in ways your brand should respect. When an image strategy is built around the proxy (for example, no per-user pixel fingerprinting), your compliance story with GDPR and CCPA gets simpler.

The lesson is not "fight the proxy" but "stop trusting the metrics it breaks." Open rate and image-based engagement for Gmail are lost. Click rate is partially corrupted. Replace both with signals the proxy cannot reach.

What to measure instead

• Landing-page events — pageviews, scroll depth, form interactions. Fire a GA4 or server-side event from the landing page and reconcile against ESP clicks.
• Reply and forward rate — no proxy replies to your email.
• Verified inbox placement — a delivered message that actually lands in Primary or Promotions tells you more than an inflated CTR.
• Revenue per email — the bottom line is immune to pixel and prefetch noise.

FAQ