How to check a Google Ads recommendation before you apply it
TL;DR
Each Google Ads recommendation states its own reason, in terms of traffic and conversions. None of the 88 reasons we read mentions profit. Three questions to ask before applying one, and why dismissing a recommendation raises the optimisation score as well.
Google Ads shows every account a score out of 100% and a list of recommendations, each with an Apply button. Some fit the business and some don't, and from the list alone they look the same.
Google's own help pages describe what the score is and what each recommendation is based on. Reading them gives three questions that sort most recommendations in a couple of minutes each.
What the optimisation score measures
Google's definition is that the "optimisation score is an estimate of how well your Google Ads account is set to perform", and that 100% means the account "can perform at its full potential". Each recommendation carries a share of the score, and the score goes up when the recommendation is dealt with.
Dealt with can mean either answer. The same page says: "Your account can reach an optimisation score of 100% by applying or dismissing all recommendations." A recommendation that was read, judged wrong for the business and dismissed counts towards the score too. Google adds that a dismissed recommendation "can reappear after a certain period if it becomes relevant then", and that the score "is not used by your Quality Score".
Whether a high score goes with better results has been looked at once at scale. Optmyzr's study of 17,380 accounts (Navah Hopkins, August 2024) found "a correlation between higher OptiScores (80+) and improved CPA" and return on ad spend. It also found that "19% of accounts achieved an Optiscore of 90+ without accepting Google recommendations". This is a vendor's study and a correlation: an account that is looked after may score well and perform well for the same reason. It supports reviewing recommendations. It doesn't settle whether to apply them.
Each recommendation states its reason
Google's list of recommendation types gives a reason for each one, usually in a sentence that starts "Recommended because". A few, word for word:
- Raise your budgets: "Recommended because you missed out on 5% or more of your potential traffic last week."
- Add new keywords: "Recommended because you're not targeting searches that could be relevant to your business."
- Adjust your CPA targets: "Recommended because our simulations show your campaigns could receive more conversions with a smaller relative increase in targets."
- Add broad match keywords: "Recommended because using broad match type keywords in these Smart Bidding campaigns can help you effectively grow conversions within your performance targets."
We counted 88 different reasons on that page on 15 November 2026. They are written in terms of traffic, conversions, features the account isn't using, and Google's simulations. None mentions profit or margin. Five mention value, meaning the conversion value an account reports to Google. The estimate attached to a recommendation is built the same way: Google's developer documentation lists its measures as impressions, clicks, cost and conversions.
That is a fair description of what Google can see. It counts what happens in the account, and what a sale leaves behind after costs is known only to the business. So each recommendation arrives with part of the sum done.
Three questions before applying
What did it count, and over how long? The reason says. The budget recommendation counts traffic missed "last week", which is a short record for a business with a busy season or a quiet month. A reason that cites "our simulations" is a forecast, and the account's own recent figures can be set beside it.
What will it change in what the account spends? A higher daily budget is the plainest case: Google's page on budgets says a campaign "might spend up to twice your average daily budget" on a single day and up to 30.4 times it over a month, and since June 2026 the pacing behind that monthly figure has changed for scheduled campaigns. An adjusted target CPA is less obvious. "A smaller relative increase in targets" means agreeing to pay more for each conversion in exchange for more of them.
Keyword recommendations change which searches the ads appear on. For broad match, a finding from the US government's case against Google is relevant. The court wrote: "Because broader matching enters more advertisers into an auction, it leads to thicker auctions (i.e., more auction participants), which creates upward pricing pressure" (finding 278, August 2024). The finding is about how matching affects prices in general. It says nothing about recommendations, and an account's own cost per click before and after is the evidence that counts.
Is the extra worth it to the business? This is the part the recommendation leaves open. The sum is short: what a conversion is worth in profit, set against what it would cost. The working for a single click is the same one used to set a bid. Where the extra conversions would cost more than they bring in, dismissing is the right call, and the score rises all the same.
A worked example
The figures are invented.
An account spending £30 a day shows "Raise your budgets", suggesting £45. Over the last 30 days it produced 20 enquiries at £45 each.
- What it counted: traffic missed in one week. The last 30 days in the account show whether that week was typical.
- What it changes: £15 a day more is up to £456 more over a month, and up to £90 on a single day.
- Whether it's worth it: if an enquiry is worth £60 in profit, each one currently leaves £15, and £456 would buy about ten more if the cost per enquiry holds. That is a reason to apply it and read the next month's figures. If an enquiry is worth £40, each one already costs more than it returns, and a bigger budget buys more of the same. That is a reason to dismiss it.
The recommendation is identical in both cases. What differs is the value of an enquiry, which only the business knows.
Recommendations that apply themselves
Google Ads can apply recommendations without anyone pressing the button. Its help page says that with "Automatically apply recommendations" turned on, "the recommendations will apply regularly". The setting is made for the whole account, and the "History" tab on the Recommendations page lists what has been applied. Opening that tab shows whether the setting is on and which types it covers. Change history keeps two years of changes, with the previous values.
The same three questions apply when an AI assistant proposes a change to the account.
When the recommendation comes through someone else
Many people meet recommendations second-hand, as a suggestion from an agency or a consultant. The three questions work as well asked aloud: what did this count, what will it change in spend, and what figure for the value of a conversion was it judged against?
One piece of background helps here. Google's partner programme asks agencies to keep "a minimum optimisation score of 70%" across the accounts they manage, and its wording is explicit that they are "in control, with the ability to apply or dismiss recommendations based on your assessment and your clients' goals". An agency that reviews each recommendation and dismisses the ones that don't fit meets the requirement as fully as one that applies them.
What we still don't know
The one large dataset on this is a correlation from 2024, and a correlation can't separate the effect of applying recommendations from the effect of looking after an account. The wording inside an account may differ from Google's help page. Google's forecasts for a single recommendation aren't published in a form anyone outside can test.
That leaves the account's own record as the check. A recommendation applied with a note of the date and the figures before it can be judged a month later. One applied without a note can't.
Sources
- Google Ads Help: About optimisation score; Types of recommendations; About applying recommendations automatically; About overdelivery and your average daily budget; How to become a Google Partner or Premier Partner; About change history. Read on 14 and 15 November 2026.
- Google Ads API documentation, Optimization score and recommendations.
- United States v. Google LLC, No. 20-cv-3010 (D.D.C.), memorandum opinion of 5 August 2024, finding of fact 278.
- Navah Hopkins, "Does Optimization Score Affect Google Ads Performance: A Study of 17,380 Ad Accounts", Optmyzr, 23 August 2024. A vendor study; correlational.