Google will switch how it counts display ad impressions on February 17, 2027, replacing count-on-download with begin-to-render measurement. That date sits 5.5 months out, easy for an Ad Manager publisher to file and forget. Google’s help document carries a detail worth acting on now: the comparison report it recommends is anchored to a start date after August 12.
What is the difference between count-on-download and begin-to-render impressions?
Count-on-download counts an impression the moment an ad creative starts downloading to a user’s device. Begin-to-render counts it later, only after the creative has loaded and started rendering. Google Ad Manager and AdSense move display inventory on web, mobile web, and CTV banner from the first method to the second on February 17, 2027. Native, app, and video inventory already use or comply with begin-to-render, so this switch touches display only.
The comparison window starts after August 12
Google says Ad Manager publishers can size their own gap before the deadline arrives. The recipe: pull a report comparing existing count-on-download impressions against a new metric, Ad server begin to render impressions, and read the difference. The fourth step of that recipe carries the constraint.
“We advise running a historical report with a start date after August 12, 2026.”
— Google AdSense Help
Google gives no reason for that cut-off. A publisher running it today gets a short, late-summer sample rather than a seasonal average, so last year’s fourth-quarter peak is not part of the comparison. Ad server begin to render impressions is also the only comparison metric live today, and it covers ad-server-delivered inventory only. Three more are listed as coming soon: Ad Exchange, Yield group, and Total begin to render impressions. Programmatic demand sold through an exchange or a yield group has no comparison metric yet, so whatever gap an Ad Manager publisher measures now covers only part of the inventory the February switch will touch.
| What | Count-on-download (until Feb 17, 2027) | Begin-to-render |
|---|---|---|
| Impression counted when | Ad starts downloading to the device | Creative has loaded and started rendering |
| Inventory affected | Display: web, mobile web, CTV banner | Same display inventory from Feb 17; native, app and video already on BTR |
| Comparison metric today | Ad server impressions | Ad server begin to render impressions |
| Coming soon | Not applicable | Ad Exchange, Yield group, Total begin-to-render impressions |
Why the seller’s number is the one that moves
Google states the reason plainly: the change “is designed to reduce discrepancies with the impressions counted by your advertiser partners and third-party measurement providers.” Begin-to-render also “aligns impression measurement methodologies with demand-side platforms,” Google adds. Both statements point the same way: the discrepancy Google wants to close sits between its own counts and the buy side’s. February 17 moves the sell-side number, on a standard Google says the Media Rating Council and the IAB periodically review and update.
It is also the third time in ten days that a vendor has redefined the unit its industry trades on, though this one is an announcement rather than a change already in force. YouTube began counting a public view from the first frame, leaving the number that pays creators untouched, a split covered in how YouTube’s public view count differs from the engaged views it pays creators on. Nielsen revised its Big Data + Panel television currency seven ways in a single day, detailed in Nielsen’s seven changes to its TV ratings currency. Google’s switch fits the same shape: the count changes, the thing it measures does not.
What a smaller impression count does to reported CPM
Google’s own expectation is hedged: publishers “may notice a decrease in total display impressions.” On revenue, earnings and payouts the document says nothing in either direction. One piece of arithmetic follows from the mechanics regardless: impressions sit in the denominator of CPM. If the count falls and revenue holds steady, dividing the same revenue by a smaller number raises the reported CPM. That is arithmetic, not a forecast, and it is ours, not Google’s.
What publishers can still control
Google’s guidance for the transition period targets the creative pipeline, not the calendar. Google’s instruction is direct: “You should always optimize your creatives to improve asset downloading and rendering on your sites.” Begin-to-render only fires once rendering starts. Chrome disables ads that consume too much network or CPU, which can stop a creative from rendering at all. Lazy-loading settings sit on the same lever: raising the rendering margin renders assets earlier, at a cost of heavier demand on the device.
The switch itself does not happen before February 17, 2027; Google describes only “a brief ramp-up of traffic” after the switch, not a phased rollout before it. The measurable part today is narrower still: a few weeks of data, one metric, ad-server inventory. Google also notes that the numbers themselves are still settling, saying “our measurement heuristics will continue to evolve over the coming months, and we will actively roll out optimizations to mitigate miscounting issues.” Any gap measured now is a first reading, not a final one. Google’s AdSense help documentation lays out the report recipe; Search Engine Land reported the September 1 publisher notification.