Google AdSense today told publishers that display advertisements which load but never begin painting to the screen will stop counting as impressions from February 17, 2027, a change the company will apply automatically and one that arrives more than nine years after the industry standard behind it was finalised.

The notification arrived by email from adsense-noreply@google.com at 16:43 on September 1, 2026, sent from Google Ireland Ltd at Gordon House, Barrow Street, Dublin 4, and labelled a mandatory service announcement. It carries no named author. The signature reads simply as the Google AdSense Team. A companion Help Center article, titled Begin-to-render impression counting for display ads, sets out the technical detail and extends the same date to Google Ad Manager.

What the counter will stop counting

The mechanics are narrow and the consequence is arithmetic. Under the current method, which Google calls count-on-download, an impression registers at the moment an advertisement starts to download to a device. From February 17, 2027, the trigger moves later in the chain. According to the AdSense email, an impression will only be counted once the advertisement "has successfully loaded and has started to render on the user's device."

Everything that falls in the gap between those two moments disappears from the count. Google gives the example directly in the email: an advertisement that begins downloading while a reader is already navigating away from the page will no longer produce an impression. The company states that publishers may see a change in total impressions after the transition, without quantifying it.

The Help Center article is more explicit about direction. It describes begin-to-render as a more rigorous measurement standard than count-on-download, notes that the older method counted advertisements which loaded onto a page but were potentially abandoned before fully rendering, and states that publishers may notice a decrease in total display impressions because the count occurs later in the ad serving process.

Scope matters here, and the two documents describe it differently. The email addresses AdSense publishers and refers only to display advertisement impressions. The Help Center article covers both Google Ad Manager and AdSense, and specifies display inventory on web, mobile web, and connected television banner placements. Its methodology table places banner display on mobile web, desktop and CTV in the count-on-download column, with everything else already outside it: non-banner display such as native and interstitial formats is listed as compliant with begin-to-render, application inventory counts on a one-pixel basis, and video counts on the first frame. Both of the latter are described as compliant with begin-to-render rather than as begin-to-render itself, a distinction the documentation does not elaborate.

That leaves banner display as the last category still counting on download, which is the reason Google gives for the change. According to the Help Center article, the shift unifies impression counting methodologies across the platform.

A standard finalised in 2017

The document Google cites as its authority is the Desktop Display Impression Measurement Guidelines, version 7.1, published by the IAB Technology Laboratory and the Media Rating Council in October 2017 under the label MMTF Final Version 1.1. The MMTF is the Modernizing Measurement Task Force, formed by the IAB in 2015 as an advisory body for updating MRC digital measurement guidelines. Google is listed among its participants, and among the governing member companies of the IAB Tech Lab.

The guidelines define an ad impression as a measurement of responses from an ad delivery system to an ad request from the user's browser, filtered for invalid traffic and recorded as late as possible in the delivery process. The operative sentence is blunt: "The ad must be loaded and at minimum begin to render in order to count it as a valid ad impression."

The definitions matter for anyone reconciling counts across vendors. In the guidelines, loaded means the creative file has been transmitted and received at the client side. Render refers to "the process of painting the creative file or adding it to any portion of the Document Object Model." Crucially, the standard does not require the advertisement to be seen. As the document states, "It does not require any portion of the ad to be visible or displayed on screen." Visibility is the province of a separate metric, the viewable impression, governed by its own MRC guidelines.

The guidelines also anticipate variation in exactly when rendering is measured, acknowledging that there will necessarily be slight differences among measurers in the precise moment rendering qualifies. Vendors that cannot measure fully rendered impressions are encouraged to study the difference between render initialisation and completion. Publishers reconciling AdSense figures against a third-party measurement vendor after February 2027 will still be comparing two implementations of the same definition, not two readings of one clock.

Nine years and four months separate the publication of those guidelines from the date Google will apply them to AdSense banner display. Adoption elsewhere in the company came sooner. Display and Video 360 moved display impression counting from on-download to begin-to-render in September 2025, and Campaign Manager 360 followed in November 2025. Buy-side platforms therefore reached the standard roughly seventeen months before the sell-side products that supply them.

The buy-side asymmetry Google names

One of the three benefits the Help Center article lists is that begin-to-render aligns impression measurement methodologies with demand-side platforms. A second is that it eliminates the need for publishers to buffer count-on-download campaigns in order to deliver against begin-to-render terms.

That second item describes a workaround that has been quietly built into publisher operations for years. When a buyer's contract is denominated in rendered impressions and the seller's ad server counts downloads, the two numbers do not match, and the seller over-delivers to close the gap. Removing the mismatch removes the buffer. It also removes whatever slack that buffer provided.

The third benefit is stated as a matter of category rather than argument: begin-to-render is the industry standard and preferred methodology for impression counting.

Google states that Ad Manager will retain MRC accreditation for display, video and rich media creatives across desktop, mobile web and mobile application environments. Accreditation is not permanent by default. Under MRC procedure, a measurement service that changes its methodology submits to a fresh audit, a requirement that has driven measurement changes across the industry, including Integral Ad Science securing accreditation for third-party Amazon DSP measurement in November 2025 and Google's July 2025 revisions to three MRC accredited metrics in Display and Video 360, where invalid begin-to-render impressions were estimated to double.

Reporting history begins on August 12, 2026

The Help Center article offers Ad Manager publishers a method for estimating impact before the switch: build an interactive report with dimensions for demand channel, inventory format, expanded inventory type and device category, then compare the ad server impressions metric against ad server begin to render impressions. The difference between the two is the estimated decrease.

Attached to that method is a constraint that deserves attention. Google advises running the historical report with a start date after August 12, 2026. Anything earlier is not offered as a comparison basis. That places a floor of roughly six months of usable history under an estimate meant to anticipate a permanent change, and it excludes the entirety of 2025 and the first seven months of 2026 from any seasonal comparison a publisher might want to run.

Three further metrics are described as coming soon rather than available: Ad Exchange begin to render impressions, Yield group begin to render impressions, and total begin to render impressions. No date accompanies them. Publishers whose revenue arrives predominantly through Ad Exchange or yield groups therefore cannot currently isolate the affected volume on those channels using the metrics Google names.

AdSense publishers receive no equivalent instruction at all. The reporting workflow in the Help Center article is addressed to Ad Manager, and the interactive reports tool it relies on is an Ad Manager feature. The email sent to AdSense publishers points to the Help Center without noting that the estimation method described there is not available to them.

Rendering conditions the publisher already controls

The Help Center article identifies two site-side factors that will interact with the new counting point.

The first is creative rendering under resource pressure. Google notes that Chrome disables advertisements consuming excessive network or CPU resources under its heavy ad criteria, which can prevent creative assets from downloading or rendering. Under count-on-download, an advertisement blocked at that stage might still have registered. Under begin-to-render, it will not.

The second is lazy loading and lazy rendering. Google states that lazy rendering directly affects begin-to-render measurement because it governs when creative assets start rendering in the browser, and identifies two Google Publisher Tag parameters, renderMarginPercent and fetchMarginPercent, as the controls that move rendering earlier. The article notes that raising them increases demands on the network and the end user's device.

That trade-off connects to a broader tension in publisher engineering. Deferring asset loading improves page performance scores and reduces bandwidth. It also delays the moment an advertisement paints, and after February 2027 that moment is the moment revenue-bearing measurement occurs. Latency in the ad chain, previously a matter of yield and user experience, becomes a factor in whether the impression exists in the ledger at all.

The 2017 guidelines contain a related provision that neither Google document restates. Pre-fetched and pre-rendered content does not qualify as a valid impression unless the advertisement loads and begins to render in response to an actual user request. Speculative loading techniques that Google itself has recommended to AdSense publishers, including the Speculation Rules API promoted in a March 2026 publisher email, sit adjacent to that boundary.

What neither document states

Two omissions stand out.

Neither the email nor the Help Center article addresses publisher earnings. The email says a change in total impressions may follow. The Help Center article says total display impressions may decrease. Neither says whether the money moves with the count.

This is a departure from precedent. When AdSense switched from served impressions to downloaded impressions in May 2018, Google stated at the time that some publishers might see decreases in impression counts and corresponding improvements in impression RPM, CTR and other impression-based metrics, and added that earnings were not expected to be impacted. The 2026 notification contains no comparable sentence. The mechanical effect on ratio metrics is not in dispute: impressions form the denominator of RPM and impression-based click-through rate, so a smaller denominator lifts both at constant revenue and constant clicks. What Google has not said in either document is what happens to the numerator.

The second omission is scale. No figure is offered for the expected reduction, in aggregate or by inventory type. The Display and Video 360 change in September 2025 was similarly unquantified. The July 2025 revision to MRC accredited invalid-impression metrics, by contrast, carried estimates of a 100 percent increase for two metrics and 30 percent for a third. Nothing structurally prevents a similar estimate here, and none is given.

Why this matters for publishers and buyers

Four counting shifts have hit AdSense reporting in under a year, and this is the fourth. Google changed Multiplex ad request counting to a per-slot basis on June 23, 2026, producing a step change in request volumes on the day it was disclosed. It replaced the Auto ads load slider with numerical banner controls in April 2026. It retired the Related Search format from Auto ads on August 6, 2026, thirteen days after notice. Each carried the same assurance that no publisher action was required, and each produced a discontinuity in the reporting series.

For anyone maintaining year-over-year dashboards, February 17, 2027 becomes a break point in the display impression line that reflects definition rather than demand. Comparisons that span it will understate 2027 volume against 2026 by an unknown margin, and any alerting threshold calibrated on count-on-download volumes will fire on the transition.

For the buy and sell relationship, the effect runs the other way. The Help Center article states that the change is designed to reduce discrepancies with impressions counted by advertiser partners and third-party measurement providers. Discrepancy reconciliation between publishers and buyers has been a standing operational cost precisely because the two sides counted at different points in the delivery chain. Closing that gap on banner display removes one source of dispute from monthly billing conversations.

The backdrop is a publisher revenue line under sustained compression. Alphabet's Google Network segment, which contains AdSense, AdMob and Ad Manager, fell 4 percent year over year to 6.97 billion dollars in the first quarter of 2026 and declined 1 percent to 7.30 billion dollars in the second, against Search advertising growth of 17 percent to 63.3 billion dollars in the same quarter. Publisher supply itself has been contracting, with ad supply falling 40 percent in the second quarter as AI answer surfaces absorb queries that previously produced page views.

Into that environment lands a methodology change that will reduce a headline volume metric by an undisclosed amount on a fixed date. It aligns AdSense and Ad Manager with a standard the industry agreed in 2017 and that Google's own buy-side platforms adopted in 2025. It also gives publishers five and a half months, a six-month reporting window, and no estimate of the size of the adjustment.

Timeline

Summary

Who: Google AdSense and Google Ad Manager publishers worldwide, together with the advertisers, demand-side platforms and third-party measurement vendors that reconcile impression counts against them. The notification came from Google Ireland Ltd, signed by the Google AdSense Team, with no individual named. The methodology originates with the IAB Technology Laboratory and the Media Rating Council.

What: Display impression measurement moves from count-on-download, which registers an impression when an advertisement starts downloading, to begin-to-render, which registers it only after the creative has loaded and started to render on the device. Advertisements that download but never begin rendering will no longer be counted. Google states that publishers may see a decrease in total display impressions and offers no estimate of its size, and neither the email nor the Help Center article addresses whether earnings change.

When: Google emailed AdSense publishers on September 1, 2026 at 16:43. The change takes effect on February 17, 2027 and applies automatically. Google advises that historical impact reports in Ad Manager use a start date after August 12, 2026. The underlying IAB and MRC guidelines were published in October 2017.

Where: The change covers banner display inventory on web, mobile web and connected television. Non-banner display, application and video inventory in Ad Manager and AdSense already count on a basis Google describes as compliant with begin-to-render. The Help Center reporting method for estimating impact is available to Ad Manager publishers rather than AdSense publishers.

Why: Google states that the change unifies impression counting across its inventory types, aligns the platform with the IAB and MRC standard and with demand-side platform measurement, removes the need for publishers to buffer count-on-download campaigns against begin-to-render terms, and reduces discrepancies with buyers and third-party measurement providers. Ad Manager is stated to retain MRC accreditation for display, video and rich media across desktop, mobile web and mobile application environments.