Google Ads: what the bidding change of 17 August really changed

Google has changed the way its algorithms bid. The best-kept accounts are the most exposed: they used to beat their target, and that is precisely the advantage that has just disappeared.

On 17 August 2026, Google changed the way its algorithms bid. For many advertisers, nothing appeared to move. For those with the best-kept accounts, results got worse — above all at the moment of raising the budget.

Two months on, the picture is becoming clear. Here is what changed, what we are seeing in our accounts, and why we did not overturn everything overnight.

The change, plainly

In Google Ads you can tell the algorithm: “I want to pay 50 CHF for a conversion on average.” That is the target CPA.

Before 17 August, a tight budget meant the algorithm could not buy everything. So it took the cheapest conversions first. The result: you asked for 50 CHF and often paid 30 CHF. A pleasant surprise, and one that well-built accounts got almost every time.

Since 17 August, the algorithm aims at the target rather than under it. You asked for 50 CHF, so it will go and find conversions all the way up to 50 CHF. Same budget, same target, but dearer conversions, and therefore fewer of them. Google is not touching your targets or your budgets: revisiting them is up to you.

The same mechanism applies to target ROAS: the algorithm no longer tries to beat the return you asked for, it converges on it.

The change covers Search, Shopping, Performance Max and Demand Gen campaigns. Display campaigns already worked this way, and App, Video reach and Video view campaigns are out of scope.

Why the best-kept accounts are the most exposed

That is the paradox of this update. It does not punish badly run accounts. It neutralises the advantage of the accounts that were beating their target.

A well-structured account — the right keywords on exact match, ads that have been worked on, bids that have been set — almost always beat its target. At a constant budget, a cost per conversion climbing back towards the target mechanically means fewer conversions.

Take a Search campaign at 3,000 CHF a month, a target CPA of 50 CHF and an actual CPA of 30 CHF. Before 17 August: 100 conversions. If nothing is touched and the CPA converges on 50 CHF: 60 conversions. Forty per cent less volume, for exactly the same spend.

Google presents the update as a gain in predictability. For advertisers who were doing their job properly, it is mostly a margin of efficiency that disappears.

What we are seeing in our accounts

The accounts hit hardest are the ones that were well built and regularly beat their target.

The moment it hurts is the budget increase. Before, when a client wanted to accelerate, we raised the budget and performance followed. Today, every extra franc costs more: the CPA rises and results fall. That is exactly what marginal ROI measures — the return on the last franc invested rather than the average of all the others.

Let us be clear: we do not like this update. It penalises the advertisers who made the effort to structure their accounts properly, and it makes budget increases more expensive. But it is here, and it is not going back. Our job is to adapt to it, not to put up with it — as with Performance Max yesterday and AI Max today: automation advances, and what is left in the advertiser’s hands moves.

Three possible settings, and what each one changes for you

Take the same campaign again: 3,000 CHF a month, a target CPA of 50 CHF, an actual CPA of 30 CHF before 17 August, so roughly 100 conversions.

The message is simple: the target CPA has to become your steering lever, not the budget. Set the target at the price you are genuinely prepared to pay for a conversion — which means starting from your margin rather than from an ROI objective — then give the campaign enough budget to stop being limited. A deliberately tight budget under a high target now means paying more to get less.

Why we changed nothing overnight

When Google modifies its algorithms, the worst reaction is to change everything in the same breath. Every change of target or strategy restarts a learning phase. Reacting in the heat of the moment means stacking disturbances and no longer knowing what produces what.

On Google Ads, a serious observation period lasts about six weeks. That is how long the algorithm takes to settle, and how long it takes to tell a genuine change of trend from a passing fluctuation. So we are only now entering the phase where real decisions can be made — and even then, carefully.

What we are doing now

1. Revisiting every target from what the client can afford. The right CPA follows from margin and from what a customer is worth over time, not from a figure set when the campaign launched.
2. Comparing target against reality, campaign by campaign. The campaigns where the actual CPA sat far below the target are the ones hit hardest. They are the ones we treat first.
3. Choosing between the three settings. Keeping the target if it really matches what you accept to pay. Lowering it to recover efficiency, at the cost of some volume. Or settling on something in between.
4. Checking what feeds the bidding. An algorithm aiming at a target is only worth as much as the conversions it is sent: the quality of the collection and the quality of the leads become decisive.
5. Raising budgets in steps. No more flooring the accelerator: gradual increases, measuring the cost of each step.
6. Giving ourselves time to judge again. Every adjustment needs three to six weeks before it can be assessed.

If your results have slipped since the end of the summer

If your Google Ads results have fallen since the end of the summer, above all after a budget increase, the problem is probably not your account. It is this change.

The observation period is behind us. Now is the moment to revisit targets and the way budgets are raised. Calmly, campaign by campaign, and giving ourselves time to measure again.