Why Gmail open rates stopped meaning anything

Open tracking works by embedding a tiny image whose URL encodes the recipient. Gmail does not fetch that image the way your tracker assumes. It routes image requests through Google's own proxy, caches the result, and may fetch images before a human looks at the message. The consequences: your Gmail open rate is inflated, unreliable per-recipient, and no longer a usable proxy for engagement or placement.

What Gmail's image handling does to your data

• Proxying. The request arrives from Google's infrastructure, not the recipient. You lose IP, geolocation and usually device detail.
• Caching. The pixel is fetched once and served from cache afterwards, so repeat opens undercount while the first "open" may not correspond to a human at all.
• Prefetching. Images can be retrieved when the message is processed rather than when it is read, producing opens for messages nobody saw.
• Scanners and bots. Security products fetch every URL in a message, including your pixel, which is why some opens arrive seconds after delivery in perfectly uniform bursts.

Add privacy features on the client side — Apple's Mail Privacy Protection being the largest — and a cross-provider open rate is a composite of several different phenomena, weighted by which clients your list happens to use.

The deliverability consequence

Two separate mistakes follow from trusting opens:

Mistake one: using opens to measure placement. An open implies the pixel was fetched, which can happen for a message sitting in Spam if a scanner touched it, and will not happen for a message read with images blocked. Opens are not a placement measurement in either direction.

Mistake two: using opens to decide who is engaged. If you suppress "unengaged" recipients based on opens, you will cut people who read every message with images off, and keep people whose scanner fetches everything. The list you end up with is optimised for your tracker, not for revenue — and list-based reputation signals get worse, not better.

What Gmail actually weights

Gmail's own signals are not your pixel. In rough order of strength:

1. Replies. A human writing back is the strongest positive signal available, and it also shifts how subsequent mail in that relationship is treated.
2. Folder rescues. A message dragged out of Spam into the inbox is an explicit correction.
3. Not marking as spam. The absence of complaints across volume, expressed as your spam rate — Google's bulk-sender threshold is 0.30% and the sane target is under 0.10%.
4. Durable behaviours like starring, archiving without reading, and deletion patterns, which you cannot see.

Notice that none of these are visible in your ESP's open column, and that the two strongest are things a warm-up can genuinely generate for a new sender with no history.

Measure the right things instead

On that last row, be clear about what exists: v2 reports spam rate, authentication, delivery errors and compliance status, plus a Deliverability analysis report added in June 2026. The domain and IP reputation dashboards are permanently gone, retired with v1 — senders redirected from 30 September 2025, v1 gone by the end of 2025.

Should you drop open tracking altogether?

For cold outreach, usually yes. The pixel adds a remote image to a message that would otherwise contain none, which pushes it toward the promotional structure the categoriser looks for, and it adds a tracking domain whose own reputation is scored separately from yours. You are paying a real deliverability cost for a number you cannot trust.

For permissioned bulk mail, keep tracking if your reporting depends on it, but treat Gmail opens as directional at best and never as a suppression criterion.

What not to do

• Do not compare this year's open rate to last year's and conclude your deliverability changed. The measurement changed.
• Do not suppress recipients for not opening. You are cutting image-blockers and keeping scanners.
• Do not use a shared or throwaway tracking domain. Link-domain reputation is separate from yours and you inherit whatever it has.
• Do not A/B subject lines on Gmail opens alone. Cache and prefetch noise will outweigh the effect you are testing.
• Do not put a pixel in a cold first-touch email. It is the cheapest structural signal to remove.
• Do not read an open as "not in spam". Scanners open spam-foldered mail too.

How MailPilot helps

MailPilot measures placement with a seed-inbox network that reads folders and Gmail tabs directly, so you get a placement number that does not depend on pixels at all. Warm-up over OAuth or SMTP/IMAP generates the signals Gmail actually weights — real replies and folder rescues between real mailboxes — and per-mailbox reputation is tracked daily alongside SPF, DKIM and DMARC checks.