Should you apply every Google Ads recommendation?

Google Ads recommendations run from “add more ad text” to “widen delivery with broad match keywords or Search partners.” Every one of them has an Apply button, but the effects are not the same.

Do not apply all of them. Split them into “apply now,” “try after measurement,” and “hold.” The split depends on whether delivery widens, whether you can measure the result, and whether you can revert and compare.

This article uses a campaign with a daily budget of ¥3,000. I checked the recommendations there, added ad assets, and held 28 broad match keywords and Search partners. At the end there is a checklist you can use while looking at the interface, and a prompt you can paste into ChatGPT or Gemini.

Do not aim for an optimization score of 100%

In Google’s official explanation of the optimization score, the score is an estimate of how well the account is set up to perform. Recommendations and their effect on the score are shown from campaign settings, results, and trends in Google Search.

What matters here is that the score changes both when you apply a recommendation and when you dismiss one you do not need. Getting closer to 100% is not the same thing as increasing revenue or profit.

A recommendation to “enable Search partners,” for example, can increase the chances for the ad to serve. If conversion tracking is not working, you cannot tell whether the extra clicks led to a signup. The score can go up while the information you need to decide goes down.

When you open recommendations, check these four points before the points they would add to the score.

What I applied immediately was more ad assets

On our search ads, the responsive search ad had 7 headlines and 2 descriptions, and ad strength was shown as “Poor.” I added the following assets, staying within facts about the product.

• Added 3 headlines, from 7 to 10
• Added 1 description, from 2 to 3
• Added 4 callouts, including “14 days free” and “No card required”
• Added 4 sitelinks, to improvement recommendations, AI chat, pricing, and the report feature

After that, ad strength went from “Poor” to “Average.”

That does not mean click-through rate or conversion rate improved. Google’s explanation of ad strength says ad strength is feedback at creation time about the number, relevance, and combinations of assets. It is not a metric that directly decides ad rank or who wins the auction.

In the 7 days before I checked, impressions were 0 and clicks were 0. What I knew was only that “there are more ad combinations to try.” Whether results improved is a separate check, after delivery, of impressions, click-through rate, conversion rate, and CPA.

Ad assets can add message options without widening who the ad is aimed at by much. If the content matches the facts and you can check the destination, this is a relatively easy change to try among the recommendations.

I held the 28 broad match keywords

The same screen also recommended adding 28 broad match keywords. I did not apply that.

In Google’s explanation of keyword match types, broad match is wider than phrase match or exact match, and the ad may show on related searches that do not contain the keyword itself. Google says it is important to combine this with Smart Bidding, which can use signals for each search.

Broad match is not bad. It can widen search terms and find demand you did not expect. On the other hand, adding 28 at once in the following state makes it hard to tell what caused delivery or cost to change.

• Search terms and results for the existing keywords have not accumulated yet
• There is no standard yet for negative keywords
• You have not confirmed that the conversion tag is receiving hits
• You have not checked the search intent of each candidate

First, look at search terms for the existing keywords and separate words close to a signup from words that are out of scope. If you try broad match, do not add the candidates as a batch. Narrow them to a unit whose intent you can explain, and leave the change date and the condition for reviewing it.

Try Search partners after measurement

Enabling Google Search partners extends the chance to serve ads to partner sites other than Google Search. In Google Ads’ definition, placements include not only search results pages but also site search, directories, and product detail pages.

Reach can increase, and you might get cheaper clicks or conversions than on Google Search. Search behavior and placements are not necessarily the same as Google Search. Splitting a daily budget of ¥3,000 across several networks can make it hard to collect enough volume to judge which side is better or worse.

Before you enable it, check at least these three points.

1. The conversion tag is receiving hits normally
2. You have a baseline for cost, conversions, and CPA from Google Search alone
3. You have decided the period for evaluating Search partners, and the condition for turning them back off

After you enable them, you can compare Google Search and Search partners in the interface under “Network (includes search partners).” Google also documents how to segment by network. Placements can be checked in the Search partner placements report, but what to look at first is cost, conversions, CPA, and quality after signup, by network.

In this campaign, the tags for 2 conversion actions were unverified. There was also no baseline for Google Search alone, so I placed Search partners in “try after measurement,” not in “dismiss.”

If conversions are unverified, do not widen delivery

The premise for judging a recommendation is that you can measure the result. If you change delivery scope or bids while the tag is unverified, more clicks still do not tell you whether the change was good. The result data passed to automated bidding also stays unclear.

From Google Ads conversion-tracking troubleshooting, check the conversion action status and diagnostics. Also confirm the following with a real signup:

• The tag fires at the correct completion point
• A page view alone is not counted as a signup
• The same signup is not counted twice
• You can see the hit in the Google Ads interface

Until measurement is fixed, prioritize changes with a limited effect, such as preparing ad assets. Try broad match, Search partners, and changes to automated bidding after you can compare results.

A checklist that splits recommendations into three kinds

When a recommendation appears in the interface, place it in this table.

When you are unsure, write one sentence: “If I apply this, what increases?”

• Adding ad assets → more message combinations you can try
• Adding broad match → more search terms the ad can match
• Search partners → more placements where the ad can show
• Raising the budget → a higher ceiling on spend you can use

If what increases is different, the measurement you need and the risk are different too. It matters that they are not all treated as the same “optimization.”

A decision prompt to paste into ChatGPT or Gemini

While looking at the recommendations screen, enter the recommendation name, the description, the current settings, and recent results. Do not paste an account ID or personal information.

You help judge Google Ads account decisions.
Evaluate the recommendations below one by one. Do not apply them as a batch.

Goal:
- Campaign goal:
- Definition of a conversion:
- Daily budget:

Measurement:
- Conversion tag status:
- Recent impressions:
- Clicks:
- Conversions:
- CPA:

Recommendations on screen:
- Recommendation name:
- Description from the Google Ads screen:
- Setting that changes if applied:

Rules:
- Do not invent results or industry averages that are not in the input
- Do not use a rise in optimization score as the only reason to adopt something
- For each recommendation, separate the effect on delivery scope, cost, search intent, and measurement
- If it cannot be measured, list what to check before widening delivery
- Do not recommend several changes at once. Split them into units whose outcome you can judge later

Output:
1. Facts that could be confirmed
2. A table classifying each item as "apply now," "try after measurement," or "hold"
3. The reason for each class
4. Screens and metrics to check before applying
5. Metrics to watch after applying, and the condition for reverting

If the AI’s answer says “CPA will improve” or “conversions will increase,” and the input has no results, that is only a prediction. Adopt an answer that separates confirmed facts from the next thing to check.

Shorten the check you do every time

To judge a recommendation safely, you need cost, clicks, conversions, and CPA in Google Ads for the same period, lined up with the change history. How to break down a number when it moves is also covered in A prompt for investigating a worse Google Ads CPA.

Shoryaku AI connects to Google Ads and can use ad spend, conversions, CPA, ROAS, daily trends, and results by campaign as material for analysis. It does not change ads on its own. Pausing, adding, or excluding a keyword is sent to Google Ads only when you check the operation Shoryaku AI showed and approve it one at a time, and you can undo it later. Budget, bidding strategy, and approving recommendations are done in the Google Ads interface.

Google Ads analysis can be tried for 14 days, with no card at the start.

Summary

Google Ads recommendations are a useful checklist. They are not an instruction to apply everything.

1. Start with recommendations that do not change delivery scope much, such as ad assets
2. Do not rush to widen delivery until conversion measurement is working
3. Try broad match after you have search terms and a standard for negatives
4. Build a Google Search baseline for Search partners, then evaluate them by network
5. Judge by cost, conversions, CPA, and quality after acquisition, not by the optimization score

First, rewrite each recommendation on screen into one sentence: “If I apply this, what increases?” Whether that answer is ad assets, search terms, or placements is what separates a change you can try now from a change that waits on measurement.

• Marketing
• AI