Google Ads auto-apply recommendations: helpful or risky for contractors?

Contractors should not enable every Google Ads auto-apply recommendation by default. Review suggestions individually against the jobs you want, the territory you serve, the conversions you trust and the capacity you can handle. Recommendations that change bidding, keywords, networks, conversion goals or reach deserve deliberate approval.

This is not an argument that every recommendation is bad. A suggestion may expose a missing asset, broken conversion setup or genuine opportunity. The risk is context. Google Ads can see campaign data and forecast models; it cannot see that an HVAC team is booked for two weeks, a roofer refuses small repairs, or a plumbing company loses money beyond a particular postcode.

The owner’s question is therefore not “How do we reach 100%?” It is “Which recommendation improves the business outcome this campaign exists to produce?”

What does auto-apply actually do?

Google lets advertisers select recommendation types that may then be applied automatically when the system considers them relevant. This is an account-level setting. The Recommendations page includes a queue, while the Manage and History areas show enrolled types and recent applications.

As of October 6, 2026, Google lists auto-applicable recommendations across ads and assets, bidding, keywords and targeting, and measurement. Examples include adjusting CPA or ROAS targets, switching bid strategies, adding broad-match keywords, expanding to Search Partners, adding Dynamic Search Ads, removing certain keywords and upgrading conversion tracking.

Google’s current documentation also makes two limits clear. Auto-apply does not increase budgets, and budget raising is not included in the available auto-apply set. That does not make the remaining changes harmless: a broader keyword, different target or additional network can change where and how the existing budget is spent.

Does a 100% Optimization Score mean the account is healthy?

No. Google describes Optimization Score as an estimate of how well an account is set to perform. It runs from 0% to 100% and changes with campaign statistics, settings, status, recommendations and recent recommendation history.

Most importantly, Google states that an account can reach 100% by applying or dismissing all recommendations. That makes the score useful as a checklist of unresolved suggestions, but not proof of strong lead quality, profitability or operational fit.

Optimization Score is also separate from Quality Score. It does not tell a contractor whether calls were answered, whether forms came from homeowners, whether appointments were available or whether estimates became revenue. A 72% account producing profitable replacement work is commercially stronger than a 100% account filled with cheap, irrelevant enquiries.

The recommendation risk map for home-service campaigns

The following classifications are Convirtly Trades’ recommendations, not Google rules. They reflect the amount of business context and testing normally required before a local contractor should accept the change.

Why broad-match and network recommendations need restraint

Additional reach can be valuable, but it should solve a diagnosed constraint. A roofing campaign limited by a small exact-match list may benefit from a controlled broad-match test. The same change can waste money if the account already attracts DIY searches, material shoppers, employment queries or work outside the dispatch area.

Similarly, Search Partners are not automatically good or poor. They are a separate source of inventory that should be compared with Google Search using connected calls, qualified opportunities and booked jobs. Our contractor Search Partners audit explains the available network reporting and its limitations.

Before adding keywords automatically, review the actual queries already triggering ads. A negative-keyword process is not a one-time cleanup; it is the feedback loop that tells you whether broader matching still respects the work you sell. Use the contractor search-term audit to classify useful, ambiguous and clearly irrelevant demand.

Why bidding recommendations can look sensible and still be wrong

A suggested target CPA may be mathematically plausible inside Google Ads while being commercially poor. The platform may be counting every form, a short phone call or a secondary service with a lower margin. It may not know that only half the enquiries reach an estimator or that the team cannot serve the recommended growth.

Before accepting a bid-strategy or target change, answer four questions:

• Which exact conversion actions will the strategy optimize toward?
• How many of those conversions were genuine qualified opportunities?
• What complete acquisition cost can the service support?
• Can the office and field team handle more of this specific work?

Our primary versus secondary conversion guide shows how weak actions can dominate the learning signal. The related bidding guide explains why a strategy should follow the most reliable outcome the account can measure.

Which recommendations are worth reviewing first?

Start with repairs and measurement. Broken destinations, disapproved assets and conversion-tracking gaps can obstruct evidence before any strategic expansion begins. Even here, confirm that the proposed tracking improvement does not count the same call or form twice.

Next, review ad assets and copy. Missing sitelinks, callouts or service-specific information may deserve attention, but generated wording still needs a human check. “24/7 emergency service,” “free estimate” or “same-day repair” must not appear unless the business can honour the claim.

Then examine reach and bidding changes with a test plan. State the hypothesis, affected campaigns, success measure, guardrails and review date. “Google recommended it” is not a hypothesis.

A 15-minute auto-apply audit

1. Open Recommendations and the auto-apply controls. Record every enrolled recommendation type rather than relying on memory.
2. Separate review from automation. Turn off automatic application for any category that can materially change bids, targets, query reach, networks, pages or conversion goals until it has an owner-approved rule.
3. Inspect the History tab. Note what was applied, when and to which campaigns.
4. Open Change History. Google documents this as a two-year record of account, campaign and ad-group changes. Filter around unexpected performance shifts.
5. Compare outcomes before and after. Use qualified lead rate, booking rate, won work and acquisition cost—not only clicks or platform conversions.
6. Assign an owner. Decide who reviews new recommendations weekly and who can approve high-impact changes.

Do not reverse several changes together unless the account is in immediate danger. If one recommendation introduced broad match while another changed bidding and a third enabled a network, restore control in a documented sequence so the next review remains interpretable.

What should an agency report?

An agency should disclose which auto-apply categories are enabled, which recommendations it accepted or dismissed, what changed in the account and why the decision fits the contractor’s economics. The report should connect each material change to a measurable hypothesis.

Useful questions include:

• Which recommendations can apply without my approval?
• Who reviews the recommendation queue and how often?
• What happened in Change History during the performance shift?
• Was a recommendation tested or rolled across every campaign?
• Did lead quality, bookings and won work improve after the change?
• Can you explain a dismissed recommendation without saying only that it was “not relevant”?

A lower Optimization Score is not evidence that an agency is neglecting the account. An unexplained change history, unclear conversion setup or inability to connect changes with business outcomes is more concerning.

US, Canada and UK considerations

The platform controls are broadly consistent, but the underlying business constraints are local. US campaigns may cross state licensing or metro boundaries. Canadian contractors may need province, municipality, weather and travel-distance checks. UK campaigns often rely on postcode coverage and trade-specific terms such as boiler, heating engineer or drainage.

Auto-applied expansion cannot understand every licensing boundary, service exclusion, call-handling schedule or local phrase. Keep these constraints explicit in campaign structure, negatives, location settings, landing pages and the review checklist.

Frequently asked questions

When a recommendation is important but uncertain, use our Google Ads experiments guide for contractors to turn it into a controlled hypothesis with traffic, timing and business-level decision rules.