For most managed accounts, turn off Google Ads auto-apply recommendations that can change bids, budgets, targeting, keywords, creative or conversion settings without a buyer reviewing them. Keep auto-apply only for narrowly defined maintenance changes that cannot materially alter spend, reach, measurement or the client offer.

The recommendations themselves can still be useful. Treat them as inputs, compare them with the client’s actual targets and constraints, then approve only the changes that fit.

Start with a rule for the account, not Google’s score

Google sees the activity inside Google Ads. Your team sees the rest of the operating picture: client margins, lead quality, sales capacity, geographic limits, expiring offers, platform mix and what happened after the click.

That difference matters. A recommendation can be reasonable within Google’s model and still be wrong for the account.

Before reviewing auto applied recommendations, write down the decision hierarchy for the client:

  1. Business constraints, including total budget, geography, inventory and offer dates.
  2. Measurement rules, including which conversions count and whether offline outcomes are available.
  3. Account and campaign targets, such as CPA or ROAS.
  4. Platform recommendations, forecasts and optimization score.

A recommendation should not outrank the first three items.

Optimization score is best used as a prompt to inspect the account. It should not become the account objective. If a recommendation improves the score but conflicts with the client’s budget or target, the score can stay where it is.

Use this starting position when you review the settings:

Recommendation type Starting position Reason
Bidding and target changes Turn off auto-apply Can materially change how the account enters auctions
Budget increases or reallocations Turn off auto-apply Changes spend and may move money away from the media plan
Keyword and match type expansion Turn off auto-apply Can widen query coverage and change traffic quality
Audience or targeting expansion Turn off auto-apply Can change who sees the ads
Ads, copy and asset changes Turn off auto-apply Can create offer, compliance and approval problems
Conversion and measurement changes Turn off auto-apply Can change the signal used for bidding and reporting
Narrow maintenance changes Consider case by case Some cleanup tasks may be low-risk under strict rules

Google can change recommendation names and how they are grouped. Review the meaning of each setting, not just its label.

Turn off changes that can alter spend, reach or measurement

The highest-risk Google Ads auto apply recommendations are the ones that can materially change the account before a buyer has checked the context.

Bidding strategies and targets

Keep automatic bidding strategy changes off. The same applies to recommendations that adjust a target CPA, target ROAS or another bidding control.

These changes can affect which auctions the campaign enters and what it is willing to pay. Before approving one, check:

  • Is the conversion action reliable?
  • Does the target reflect qualified leads or all submitted forms?
  • Is conversion lag distorting the recent period?
  • Is the recommendation based on a short-lived promotion?
  • Does the proposed change fit the client’s account and campaign targets?
  • Is another platform producing better marginal results?

Example: Say a campaign has a $50 target CPA. Google recommends loosening the target because the campaign could capture more conversions. That may be sensible if lead quality is stable and the client wants more volume, but it is not sensible if sales capacity is full or recent conversions include low-value actions.

The recommendation is a scenario to evaluate, not a standing instruction.

Budget increases and reallocations

Turn off automatic budget increases. Also require review for any recommendation that shifts budget between campaigns.

Google may identify a campaign as budget constrained. That does not answer whether the next dollar belongs in that campaign, another Google campaign, Meta, TikTok or nowhere at all.

Before raising a budget, check:

  • The client’s total approved media budget
  • Marginal CPA or ROAS, not just the campaign average
  • Lead quality and closed sales where available
  • Geographic and operational capacity
  • Conversion lag and recent tracking changes
  • Performance across every paid platform

Example: A campaign can be limited by budget and still be a poor place for the next $500 of illustrative spend. Another campaign or platform may have the stronger marginal opportunity. The budget recommendation is useful evidence, but it is not the full allocation decision.

Broad match, new keywords and targeting expansion

Turn off automatic additions that widen reach. That includes new keywords, broader matching and audience expansion.

These settings can expose an account to queries or users that were not part of the original plan. They deserve a search term review, negative keyword check and landing page check before approval.

For each proposed expansion, ask:

  • Which new searches or audiences could enter?
  • Does the landing page answer that intent?
  • Are brand, competitor and generic traffic separated correctly?
  • Could the change cross a geographic or regulatory boundary?
  • Is there enough clean conversion data to guide bidding?
  • How will the team judge the test?

A good keyword idea can still belong in a controlled test rather than an automatic account-wide change.

Conversion measurement changes

Do not let auto-apply redefine success.

A change to conversion settings can affect bidding, optimization score and client reporting at the same time. Review primary and secondary actions, attribution settings, duplicate events and offline imports before touching the measurement layer.

If the recommendation says tracking is incomplete, investigate it. Do not assume the proposed fix understands the client’s sales process.

Keep creative and offer changes under human approval

A Google Ads auto applied ad suggestion can look like a small copy change. It can still introduce a claim the client did not approve, refer to an expired promotion or send the wrong message to a sensitive audience.

Keep automatic ad and asset creation off when:

  • Copy requires legal or client approval
  • Prices, dates or availability can change
  • Campaigns use tightly controlled brand language
  • Different locations carry different offers
  • The landing page contains language that should not become ad copy
  • Creative testing follows a planned matrix

Review suggested copy against the live landing page, current offer and campaign intent. Check the final combination as an ad, not as isolated headlines and descriptions.

This does not mean the suggestion has no value. It can expose missing message coverage or a weak connection between the query, ad and landing page. The buyer should decide whether to edit it, test it or dismiss it.

The same principle applies to assets. A proposed sitelink, callout or image association can change what the user expects after the click. Check destination, offer, geography and approval status before applying it.

For teams establishing an approval policy, five questions to ask before AI touches client ad accounts provides a useful control framework.

Auto-apply only narrowly defined maintenance changes

There is no setting that is safe for every account. The lowest-risk candidates are maintenance changes that do not affect spend, bidding, reach, measurement, creative or the client offer.

Even then, define the boundary before enabling anything.

A maintenance recommendation may be a candidate only if all of these are true:

  • The change is reversible.
  • Its scope is clear before it runs.
  • It cannot increase the budget.
  • It cannot widen targeting or query coverage.
  • It cannot change a bidding target or conversion action.
  • It cannot publish new client-facing copy.
  • The team can confirm afterward exactly what changed.
  • Exceptions can be recorded for that account.

Removing redundant keywords is a common example of something that appears low-risk. It still needs scrutiny when match types, negatives, reporting groups or campaign structures carry meaning for the buyer.

Resolving a conflicting negative keyword may also look like cleanup. Before applying it, confirm that the negative was not intentional. A deliberate block can appear to be a configuration mistake when the platform does not know the client’s business rule.

The practical default is simple: leave automatic application off until the team can state, in plain language, what a recommendation may change and what it must never touch.

Build a review workflow your buyers can repeat

The harder problem is not finding recommendations. It is reviewing them consistently across many client accounts.

A useful queue should show:

  • Client and account
  • Campaign or asset affected
  • Proposed change
  • Reason for the recommendation
  • Numbers supporting it
  • Client target and applicable constraint
  • Expected scope
  • Approval status
  • Result after implementation
  • Dismissal reason, when rejected

Review material changes first. Budget, bidding, measurement and targeting decisions deserve attention before minor cleanup suggestions.

Then use the same sequence for each recommendation:

Check the source numbers

Confirm date range, conversion actions, attribution, lag and any recent account changes. A recommendation built on incomplete or recently changed data should wait.

Compare it with the client target

Use the target set for that campaign where one exists. Fall back to the account target only when a more specific rule has not been defined.

Check cross-platform context

A Google recommendation is about Google. The media plan may include Meta and TikTok too. Compare marginal performance and current constraints before moving budget.

Define the test and rollback

Record what will change, what success means and how the original state can be restored. For a meaningful structural change, isolate it when possible and give it enough data before making a call. If there is not enough evidence, mark it inconclusive instead of forcing a verdict.

Confirm the change landed

Approval is not the last step. Check the account after execution, confirm the intended setting changed and verify that nothing else moved with it.

Agencies building this process across a portfolio can also review Ads Arsenal for agencies.

Frequently asked questions

Should you turn off auto-apply recommendations in Google Ads?

For most managed accounts, turn off any auto-apply setting that can change bids, budgets, targeting, keywords, creative or conversion measurement. Recommendations can still be reviewed and applied manually when they fit the client’s targets and constraints.

Which Google Ads recommendations are safe to auto-apply?

There is no universal safe list because account structure, tracking and client rules differ. The lowest-risk candidates are narrow maintenance changes that do not alter spend, bidding, reach, creative, conversion definitions or the offer.

Why does Google recommend raising budgets?

Google may recommend a higher budget when its forecast suggests the campaign is constrained and could capture more traffic or conversions. The recommendation still needs to be checked against marginal performance, lead quality, tracking accuracy and the client’s total budget.

Does optimization score affect performance?

Optimization score is a diagnostic score, not a direct input that determines auction performance. Applying a recommendation can affect results, but raising the score by itself is not a reason to make a change.

How do you review recommendations across many client accounts?

Use one queue that records the account, recommendation, expected effect, supporting numbers and applicable client rule. Prioritize material changes first, require approval before execution and check the account afterward to confirm the change landed and performed as expected.

Keep the recommendation, keep the decision

Ads Arsenal reads Google, Meta and TikTok as one budget, with combined metrics and MER at the top of the Overview. It does not sell media, so platform recommendations remain inputs to evaluate against the targets and constraints set by the buyer.

Nothing changes in an account until someone approves it. Once approved, Ads Arsenal writes the change, checks a few minutes later that it landed and tracks whether it worked, with one-click rollback available.

Ads Arsenal is in a free design-partner beta for agencies and in-house teams running real spend on Google, Meta and TikTok. Become a design partner.