Updated 21 September 2026. An earlier version of this article said Gmail's image proxy inflates open counts, and that Outsolvi filters proxy opens out. Both were wrong. This version replaces them.
The short answer
When your tracker says an email was opened "via proxy", or logs a user agent ending in "GoogleImageProxy", Gmail loaded your tracking pixel through Google's servers instead of letting the reader's device fetch it. Gmail does that when someone opens or previews the email with images turned on. So a Gmail proxy open is usually a real person reading. What you lose is detail: the IP address, the location and the device all belong to Google.
Where the proxy came from
Tracking pixels work because opening an email makes the reader's device request a tiny image from the sender's server. That request carries the reader's IP address and browser details, and the sender logs the moment it arrives.
On 12 December 2013, Google changed that for Gmail. Its announcement said Gmail now serves all images "through Google's own secure proxy servers", checking them for known viruses or malware on the way.[1] Google's admin documentation still puts it the same way: "When your users open email messages, Gmail uses Google's secure proxy servers to serve images that might be included in these messages."[2]
The effect on tracking showed up the same day. Filippo Valsorda tested it and found that nothing reached his server when the email arrived. "Then, I opened the email, the image automatically loaded and immediately a request got logged on my server." The request came from a Google address, not from him.[3]
What your tracker actually sees
A Gmail proxy fetch has two tells in a request log:
- A user agent ending in "(via ggpht.com GoogleImageProxy)", with an old Firefox version in front of it
- An IP address that belongs to Google, often in the 66.249.x.x range
A 2019 report on the gophish project's issue tracker recorded the full string: "Mozilla/5.0 (Windows NT 5.1; rv:11.0) Gecko Firefox/11.0 (via ggpht.com GoogleImageProxy)".[4] No ordinary browser sends it, which is why most trackers key on the user agent rather than the IP address. Inside Gmail, the image itself is served from an address on googleusercontent.com, which is where that domain comes from when people inspect a tracked image.
When Gmail fetches the image
On view, not on delivery. Google's own wording says so, Valsorda's test showed it, and it is what we found when we tested it ourselves in July 2026: nothing happens when a message lands in a Gmail inbox, and a fetch arrives when someone opens it.
Two things can make it look otherwise.
Your own Sent folder. If you send from a Gmail address, viewing the message in your Sent folder fires exactly the same proxy request as a recipient opening it, and the pixel alone can't tell them apart. Suped lists this, along with delegates and shared mailboxes, as a common cause of proxy opens that nobody on the recipient's side made.[5] When people report Gmail opens seconds after sending, this is the first thing to rule out.
The reading pane. Gmail can show a message in the reading pane without the reader thinking of it as opening it.[5] The fetch still happens. We count that as an open, because a person had the email in front of them.
Why a proxy open counts as a real open
Because Gmail fetches on view, a proxy fetch means the email was put on someone's screen. That is what an open is supposed to mean.
We learned this the hard way. Until July 2026, Outsolvi treated every Google proxy fetch as a machine and left it out of the count, the way it handles Apple Mail Privacy Protection. Then we opened a test email on three of our own devices and Outsolvi counted zero opens. Every fetch had come through Gmail's proxy, and every one had been filed as a machine. The readers were real. The rule was wrong. Gmail proxy opens now count as most likely human.
That is also the correction to the earlier version of this article, which argued the opposite.
What the proxy hides
Everything on the reader's side of the request is Google's:
- Location. The IP address belongs to a Google server, so any city your tracker shows for a Gmail open is where that server sits, not where your prospect is.
- Device. The user agent is the same old Firefox string for everyone, so you can't tell a phone from a laptop.
- Network. You can't see whether they read it at the office or at home.
If your tracker says a Gmail prospect opened your email in Mountain View on Windows, it is describing Google. Treat location and device on Gmail opens as unknown.
One view, several fetches, or none
Two quirks make Gmail open counts unreliable at the edges, in opposite directions.
Several fetches for one view. In our logs, a single view of one email once produced 10 fetches from Google addresses in the 66.249.84.x range within three minutes. A tracker that counts every fetch turns one read into ten. Outsolvi groups Gmail fetches that arrive within ten minutes of an earlier one into a single open.
No fetch for a later view. Gmail can show a cached copy of an image on later views instead of requesting it again, so a second or third read may never reach your tracker.[5] Valsorda saw no caching at all in 2013, so the behavior has changed over the years. Don't build anything on exact repeat counts.
The practical rule: one open per view, and treat a Gmail recipient's repeat-open count as a minimum, not a total.
Gmail proxy opens vs Apple Mail Privacy Protection
People lump these together because both hide the reader. They behave very differently:
| What happens | Gmail image proxy | Apple Mail Privacy Protection |
| When the image is fetched | When the email is viewed | When the email arrives, read or not |
| Is it a person? | Usually | Not on its own |
| Reader's location and device | Hidden | Hidden |
| How Outsolvi counts it | Most likely human, counted | Labeled Apple Mail, not counted unless they click or reply |
Both matter because of where email gets read. Apple Mail accounted for 62.26% of the opens Litmus measured in July 2026 and Gmail for 27.03%, across all the senders Litmus tracks.[6] Your own list will look different, especially if you mostly email businesses.
Gmail clicks are a different story
Opens through Gmail's proxy are easier to trust than people think. Clicks went the other way. Omnivery, which sells bot detection, found that the bot share of Gmail clicks jumped from about 6% to about 22% in late February 2026 and stayed there week after week, which it puts down to Gmail fetching or scanning links more aggressively.[7] A Gmail click on its own is weaker evidence than it used to be.
How to test it yourself
You don't have to take our word for any of this. With any tracker that shows raw request details:
- Send a tracked email to a Gmail address you control, from an address that isn't Gmail.
- Leave it unopened for ten minutes and check that no fetch arrived.
- Open it once. Note how many fetches arrive and what user agent they carry.
- Open it again an hour later and see whether a new fetch arrives at all.
Step 2 shows Gmail doesn't fetch on delivery. Step 3 shows the user agent and the burst. Step 4 shows how far you can trust repeat opens today.
How to read Gmail opens in your own tracker
- Look for "GoogleImageProxy" or "via ggpht.com" in the user agent. That is Gmail.
- Count one open per view, and group fetches that land within a few minutes of each other.
- Ignore location and device on those opens.
- If you send from Gmail, don't trust an open in the first few minutes after sending. It may be you.
- Treat repeat opens as a minimum.
- Judge clicks separately, and more skeptically.
How Outsolvi handles it
Outsolvi does those six things for you. A Gmail proxy fetch shows up as most likely human, with the reason "opened in Gmail", and it counts. Repeat fetches from the same view are grouped under the first one and labeled as a repeat fetch, so one read is one open. If you send from a Gmail address, a fetch in the first five minutes shows as "could be your own Sent folder" and stays out of the count. Yahoo Mail's proxy opens are counted the same way. Apple Mail Privacy Protection fetches are labeled Apple Mail, with the reason "reached an Apple Mail inbox". They don't count on their own, because Apple loads images whether anyone reads or not, but one becomes a counted read if that person also clicks or replies.
You can see how every open on your own email was graded, and why, free for 14 days. No card needed.