How recommendations work
Google Ads inspects an account’s settings, performance and history, compares them with patterns it sees across many advertisers, and produces a list of suggested changes. Each card carries an estimated effect and a button that applies it in one click. The list is grouped into themes such as bidding, keywords and targeting, ads and assets, repairs, and measurement.
The same list drives Optimization Score, the figure Google shows at the top of the account. Dismissing a card moves the score as well as applying one does, which tells you what the score is really measuring: how many of Google’s suggestions you have responded to, not how well the account performs.
Some recommendations can be set to apply themselves without anyone looking. Auto-apply is a separate screen of switches, and it is worth opening once on every account you touch, because a setting applied that way still appears as an edit in your change history.
Why recommendations matter
A minority of them are genuinely valuable, and those are the ones reporting something broken rather than proposing something new: a conversion action that has stopped recording, an ad group with no active ads, a disapproved ad, a final URL that no longer resolves, a payment method that has failed. Those are the account telling you it is faulty, and they deserve attention the day they appear.
The rest are suggestions, and their worth depends entirely on whether Google’s assumption about your goal matches your real one. Google assumes more volume is better. If your constraint is a cost per lead, a margin, or the capacity of a small sales team, more volume can be the wrong outcome.
Where recommendations go wrong
The pattern I see most is broad match proposed for a tightly built search campaign, or a budget increase recommended on a campaign already sitting at its target cost. Both read reasonably on the card and both spend more money. Applying them unexamined is how a small account in a thin auction like Nepal’s drifts away from the searches that were working, because a handful of loosely related queries can absorb a day’s budget before anyone checks.
The other trap is managing towards the score. It is easy to run an account towards a high Optimization Score and away from profit, particularly when a client report puts the score on the first page. It measures compliance with Google’s advice, not results, and no customer has ever paid for it.
What to do about them
Read the whole list on a weekly rhythm. Act immediately on anything in the repairs and measurement themes, and treat every other card as a hypothesis rather than an instruction. If a suggestion looks worth trying, run it as a campaign experiment so you can see what it actually did instead of applying it across the account and guessing afterwards.
Check the auto-apply switches on any account you inherit and turn off anything you would not have approved yourself. And when the same card keeps returning after you have decided against it, dismiss it, so the list stays short enough that the useful ones are still visible.