Somebody opens your email in Gmail. Your tracking pixel fires. But here's the twist: the request didn't come from their laptop or phone. It came from Google.
That's a Gmail proxy open. And whether it should count as a real open is one of the most quietly botched questions in email tracking.
What is a Gmail proxy open?
Gmail serves every image in every email through its own image proxy. Your tracking pixel is an image, so it goes through the proxy too. The [glossary entry on the Gmail image proxy](/glossary/gmail-image-proxy) covers the mechanics.
The result: when a Gmail user opens your email, the pixel request your tracker sees comes from Google's infrastructure, not the recipient's device. The direct fingerprint you'd get from other clients isn't there.
So the open happened. Probably. Through a middleman. The question is what to do with that.
Why does Gmail do this?
Privacy and safety, mostly. Routing images through a proxy means senders can't see the recipient's device details directly. It also lets Google scan images before they ever reach the user.
This isn't Gmail being hostile to your business. It's the direction the whole industry has moved. Apple went further with Mail Privacy Protection, which pre-fetches pixels whether the email gets opened or not. Gmail's proxy is milder: it generally fires around real activity, it just anonymizes the route.
That difference is exactly why the two need different treatment.
So should a proxy open count?
Sometimes. And that's the honest answer, even though it's less satisfying than yes or no.
A proxy open needs the full signal picture, not blind counting or blind discarding. Count every proxy request and you'll inflate your numbers, because proxy behavior can include duplicates and automated fetches. Discard every proxy request and you'll throw away real readers, because plenty of proxy opens are genuine humans reading your email in Gmail.
Both mistakes are common. Blind counting is why raw open rates on B2B lists typically run 2 to 3 times inflated, where a reported 70 percent is often 25 to 35 percent real reads. Blind discarding is the overcorrection some tools made when open tracking got harder, and it's arguably worse: you stop following up with people who are actually reading you.
How do you judge a proxy open, then?
You look at everything around it. One request in isolation is ambiguous. A pattern isn't.
Real people open email in patterns. Normal hours. Real devices behind the proxy. Sometimes they come back and re-read the same message a day later. Machines behave differently: they fire instantly after send, often hitting many links at once.
Outsolvi grades every open from Tier 1, high-confidence human, to Tier 5, bot or scanner, using request timing, network origin, and client signature. Opens below a 25 percent confidence floor get excluded. A Gmail proxy open that looks like a person at 2pm on a Tuesday gets treated very differently from a burst of fetches one second after you hit send. The full approach is on the [methodology page](/methodology), and the [confidence scoring feature page](/features/confidence-scoring) explains the tiers.
The model agrees with human-rated classifications 95 to 98 percent of the time, and it keeps learning as providers change behavior. Which they do. Gmail's proxy today doesn't behave exactly like it did two years ago, and a static filter written back then would already be wrong.
How do I know it's not just guessing?
Click the open and look. That's what the [Diagnose view](/features/diagnose-open-evidence) is for.
Every open in Outsolvi shows why it counted or why it was filtered. If a Gmail proxy open got graded Tier 2, you can see the reasoning. No mystery numbers, no trust-us dashboard.
That transparency is worth testing for when you [compare tracking tools](/compare). Ask any vendor one question: "This open right here, why did you count it?" If the product can't answer, the number on the dashboard is a guess with confident formatting.
What does this mean for your open rate?
Your Gmail opens are neither fake nor gospel. They're evidence, and evidence needs weighing.
If your current tracker counts every proxy hit, your open rate is inflated. If it drops Gmail opens entirely, you're blind to a huge chunk of your list. Either way, you're making follow-up decisions on bad data.
If you want to see your Gmail opens graded instead of guessed, Outsolvi is $7 a month billed yearly. There's a 14-day free trial with no credit card, plus a 7-day money-back guarantee. Two weeks of graded opens next to your old dashboard usually settles the question.