Google Ads Now Enforces Your CPA Target: What Changed on August 17, 2026

What actually changed

On August 17, 2026, Google changed how budget-limited campaigns behave under target-based bidding. In Google's own words, campaigns that are limited by budget will now "more consistently perform toward your bid target, including when you make budget adjustments." The Google Ads Help page is short and worth reading directly rather than through anyone's summary.

The scope covers Search, Shopping, Performance Max, Demand Gen, Display, Hotel and Travel campaigns. App campaigns, Video Reach and Video View campaigns are not included. If your account is mostly video, most of this does not apply to you. If it is mostly Search and Shopping on Target CPA or Target ROAS, most of it does.

Google also published the warning plainly: "If you have a campaign that's limited by budget that has historically overachieved its stated target, you may experience performance volatility if you don't make any changes to the campaign before August 17." Read that sentence carefully, because it contains the whole story. Overachieving a target was never supposed to be permanent. It was a gap the system happened to be leaving on the table, and the gap is what is closing.

Why "limited by budget" is the qualifying condition

This is the part that decides whether you care. A campaign carries the "Limited by budget" status when it is spending its full daily budget and the system believes it could spend more at the current bid target. Campaigns that comfortably finish the day under budget are not in that state and are largely unaffected.

You can see the status in the campaigns table. Add the Status column if it is not showing, or hover the status of any campaign that has been flat at its budget cap for a while. Do this per campaign, not per account, because the impact is uneven by design. A single account can hold five campaigns pinned at their budget and fifteen that never come close, and only the five are exposed. That unevenness is exactly what makes the aggregate account view useless for this question. An account-level cost per conversion that moves a little can be hiding five campaigns that moved a lot.

The mechanic: your target was a ceiling, not an instruction

Here is the shape of it, with round illustrative numbers rather than measured ones. Say a campaign carries a Target CPA of ten dollars and has been delivering conversions at five. The campaign is capped by its daily budget, so the system is not trying to buy every available conversion at ten. It is buying the cheapest conversions it can find inside a budget it keeps hitting, and those come in around five.

Nobody set out to run at five. Somebody typed ten into the box, possibly years ago, possibly as a guess, and the campaign quietly beat it every month since. That performance got reported upward as a result. It was really a side effect of a budget cap doing the work of a bid target.

From August 17, the system spends toward the number that was typed. Nothing errors. No setting changed. No alert fires. The campaign simply starts buying the conversions available up to ten dollars instead of stopping at the cheap end, because that is what a target of ten has always meant. The ten and the five here are illustration, not measurement. Your own numbers are the ones that matter, and they are sitting in the account right now.

The one number to pull today

For every campaign showing "Limited by budget" on a target strategy, pull the last 30 days of actual cost per conversion (or ROAS) and put it next to the target that is set on the campaign. The gap between those two figures is your forecast. It is not a precise prediction, but it is the right order of magnitude, and it is directionally honest in a way that a guess is not.

A campaign delivering at five against a target of ten has roughly twice as far to travel as one delivering at nine against the same target. That is a forecast a stakeholder can be given in advance rather than an explanation they have to be given afterwards, and the difference between those two conversations is most of what people mean by trust in reporting.

If the target on a campaign turns out to be a number nobody can account for, you have found something more useful than a forecast. You have found a campaign that has been optimizing toward an arbitrary figure, and the platform is about to start taking that figure seriously.

Volume up, efficiency down, and two honest reports that disagree

Here is where reporting gets interesting. The same underlying change produces opposite conclusions depending on which metric a dashboard leads with. Spend stays flat, because the budget cap has not moved. Conversion volume rises, because the campaign is now buying conversions it previously skipped. Cost per conversion rises, because the ones it skipped were the expensive ones.

A volume-led dashboard shows a campaign that just improved. An efficiency-led dashboard shows a campaign that just got worse. Both are reading the same account correctly. Neither is spinning anything. They are answering different questions, and nobody noticed they were different questions until the answers diverged.

This is worth naming out loud with whoever reads your reports, because the version that reaches a senior stakeholder depends entirely on which chart sits at the top of the page. If the efficiency chart leads, expect a question about what went wrong in August. The answer is that nothing went wrong. It is also, unfortunately, the answer that sounds most like an excuse when it arrives three weeks late.

Telling enforcement apart from a real problem

This matters because a rising cost per conversion has plenty of other causes, and writing off a real regression as "just the August change" is a worse mistake than the reverse. Three checks separate them.

1. Did volume move with cost or against it? Enforcement raises cost per conversion and conversion volume at the same time, because the campaign is buying more of a more expensive supply. A genuine regression usually raises cost while volume falls or stays flat. Two metrics rising together is the signature, and that is the check to run first.
2. Does the timing line up with August 17? Not the week, the date. If the shift starts on the seventeenth or shortly after and there was no site deploy, tracking change, seasonal event or competitor move on that date, the platform change is the likely cause. If it started on the fourth, go and look at what shipped on the fourth.
3. Where did the new cost per conversion settle? Enforcement moves performance toward the stated target and then stabilizes near it. A cost per conversion that climbs past the target and keeps climbing is not the system honoring your number. It is something else, and it needs a real investigation.

All three checks assume the conversion numbers underneath are trustworthy, which is an assumption worth testing before it carries any weight. If conversions are double counted, or a tag has been firing on the wrong page, the diagnostic reads the same as enforcement and you reach a confident wrong conclusion. The patterns in 5 Signs Your Conversion Tracking Is Broken are the ones that most often contaminate this kind of analysis, and a platform-level mismatch between GA4 and Google Ads is its own separate rabbit hole, covered in Why GA4 and Google Ads Never Match.

The Bid Target Adjustment Tool

Google made a Bid Target Adjustment Tool available from July 6, 2026, ahead of the change. It flags campaigns that may be affected and recommends targets based on recent performance, which is a genuinely useful starting point and takes very little time to review.

What it cannot tell you is whether the original target was ever a considered decision. It reads what the campaign has been doing and suggests a number consistent with that. If the campaign has been delivering at five, the tool can propose something close to five, which preserves the status quo and is often the right call. But "keep doing what the budget cap was accidentally doing" is a business decision, not a settings change, and it deserves to be made deliberately by whoever owns the number rather than accepted because a panel suggested it.

Annotate it now, not in September

This is the cheapest thing on the list and the one most likely to get skipped. Put an annotation in your reporting on August 17, 2026, naming the change, linking Google's documentation, and noting which campaigns carried the limited-by-budget status at the time.

An annotation added after somebody asks why September looks different is worth almost nothing. It reads as a rationalization, because that is what it is, and the person asking has no way to tell it apart from one. The same annotation added on the day is the complete answer, and it stays the complete answer in eight months when someone compares year over year and finds a step change nobody remembers.

That generalizes well beyond this change. Every platform change with a published date deserves a dated note in the reporting layer, for the same reason a retired API deserves one. We wrote about that failure class in The API Retired Today. Your Dashboard Didn't Notice. The common thread is that vendors publish these dates and dashboards do not read them.

The part that outlasts this change

Strip out the specifics and a general rule is left over: any automated system handed a number will eventually use all of it. The gap between a target and actual delivery is not a safety margin. It is unclaimed room, and platforms get better at claiming it over time. A number set casually becomes a commitment the moment the platform starts taking it literally, and you do not get to choose that moment.

So the question worth asking this week is not really about August 17. It is this: who owns each bid target in this account, and when was that number last an actual decision rather than something that had to go in the box before the campaign would save? On most accounts nobody revisits a target, because the gap never had to be explained to anyone. This is the change that makes it get explained.