Why did my campaigns change when nobody touched them?

Both platforms will edit your account on their own

Google describes the behaviour plainly: "When you turn on 'Automatically apply recommendations', the recommendations will apply regularly."

Microsoft goes further, because its version acts on silence. It says you can review, edit, apply or dismiss a suggestion, and then adds the part that matters: "If you don't take action, we'll automatically apply them after the review period of 7 days."

Read that carefully. On Microsoft, doing nothing is a decision. A recommendation you never opened is applied a week later, and the only way to prevent it is to have acted.

Microsoft is also clear about the scope once you opt in: "Opting in to auto-apply recommendations means that all of your ad groups are watched for areas with potential performance improvement." Not the campaign you were worried about. All of them.

What can change without you is broader than most people expect

The useful question is not whether automation is on. It is what it is allowed to do.

Google's list of recommendations available for automatic application includes "Add broad match keywords", "Use targeting expansion", "Expand your reach with Google search partners" and "Remove conflicting negative keywords".

Microsoft's list is similar in kind: "Create Dynamic Search ads, remove negative keywords conflicts, and add new keywords, broad match keywords, and trending queries as keywords."

Those are not cosmetic. Adding broad match keywords and expanding reach change which searches you pay for. That is the setting most owners think of as the core of the campaign. Removing a negative keyword conflict sounds like tidying, and it changes what you are excluded from.

Google also flags a consequence worth knowing if your ads insert the search term into the headline: "If you use Dynamic Keyword Insertion (DKI), the search terms not included as negative keywords could appear in your ad text."

So an automated keyword addition can end up visible in the ad a customer reads.

The score that creates the pressure is not a performance measure

This explains why the feature feels like an obligation, and it is the part worth being sceptical about.

Google defines the number: "Optimization score is an estimate of how well your Google Ads account is set to perform. Scores run from 0-100%, with 100% meaning that your account can perform at its full potential."

Now the sentence that changes how you should read it: "Your account can reach an optimization score of 100% by applying or dismissing all recommendations."

Dismissing counts. A perfect score is available to any account that clears the list, whether or not a single change was made. So the score measures how few recommendations are outstanding, not how well the account is doing. It is a queue length wearing the clothes of a grade.

That does not make it useless. It makes it a prompt rather than a verdict, and a poor thing to report as a performance number. Dismissal is not a gap in the account either, and Google notes the queue refills: "A dismissed recommendation can reappear after a certain period if it becomes relevant then."

The arithmetic of the score points the same way. Google says each recommendation carries a score uplift "which is a value from <0.1% to 100%", and that the individual figures can sum to more than 100, because applying some recommendations invalidates others. A grade whose parts add up to more than the whole is not really a grade.

One more habit worth bringing to all of this. Read the evidence attached to the claims. On Microsoft's own page, the listed benefits of auto-apply are each followed by the same footnote: "based on Microsoft Advertising internal data". That is not a reason to dismiss them. It is the vendor measuring its own feature. Check the effect in your own account rather than accepting the summary.

Microsoft is straightforward about the control, to be fair to it: "You have control over which recommendations are auto-applied to your account." The control exists. It is opt-in by category, and worth setting on purpose.

What the automation will not do is the thing that usually limits you

There is a gap in the coverage, and Google names it twice.

"Auto-applying recommendations won't increase your budget, so continue to review the 'Recommendations' page to ensure your budget isn't limiting your performance." Elsewhere on the same page: "Budget raising isn't included in this iteration of auto-apply recommendations."

That is a sensible boundary and it leaves an odd shape. The automation will widen what you match against. It will add ads and adjust targets. It will not touch the constraint that most often caps a small account. So an account can be automatically expanded into a wider set of searches while the same budget stretches across all of them.

Google is explicitly telling you that this half stays manual. It is worth hearing as an instruction rather than a footnote.

What to check, in order

Five checks, and all of them are free.

• Find the auto-apply settings on each platform and write down what is currently enabled. On Microsoft it is the lightning bolt icon on the Recommendations page.
• Read the change history rather than guessing. Microsoft is specific: "From the Change history tab, filter the records by setting the value of the Change by attribute to Recommendations auto apply."
• Note how far back that record goes. Microsoft says you can "read the applied history up to the most recent six months".
• On Microsoft, check which ads were created this way by adding the Automated ad column to the ads grid and filtering its value to Yes.
• Decide deliberately which categories stay on. Automation that adds ad variations is a different risk from automation that adds keywords.

Now the arithmetic that makes the review cadence matter, because this is where accounts drift quietly. Microsoft applies an unreviewed recommendation after seven days. If recommendations arrive roughly weekly and nobody opens the account for a month, four batches can apply without anyone seeing them.

Then put that against the six month history window. An agency or owner who reviews the account once a year can see the most recent six months of automated changes and cannot see the first six at all. The record has already rolled off. A quarterly look at change history costs a few minutes and keeps every automated change inside the window while you can still read it. The numbers are an illustration, but the mismatch between a yearly habit and a six month record is the part to fix.

Deciding which of these settings should be on for a particular account, and reading the change history when something moves, is part of what we do on paid search.

Terms used on this page

• Negative keyword

Sources

• About applying recommendations automatically (Google Ads Help)
• About optimization score (Google Ads Help)
• Auto-apply recommendation settings (Microsoft Advertising, Microsoft Learn)

Last reviewed 2026-09-12.