AI for Google Ads: Make Decisions on Profit, Not Reported ROAS

Google Ads already uses a lot of automation. Smart Bidding sets your bids, Performance Max chooses where ads appear, and AI Max widens which searches you match. So the real question isn't whether to use AI for Google Ads. It's who checks the AI, and against which numbers.

Google's automation optimises toward the conversions it can see. It does not know your margins, it counts brand searches as wins, and it will happily spend more when tracking is quietly broken. This guide covers where an independent AI layer helps, what to watch out for, and a simple routine you can follow.

Start with the honest number: ex-brand ROAS

Reported ROAS in Google Ads mixes two kinds of traffic:

• Brand searches. People typing your name. Most of them were already coming to buy.
• New demand. People searching for the problem you solve, who had not chosen you yet.

Brand searches usually show a very high ROAS, which lifts the account average and hides weak campaigns. Strip them out and you get ex-brand ROAS, the return your ads actually bought you.

Compare ex-brand ROAS to your break-even ROAS. If ex-brand ROAS sits below break-even, the account is losing money on new customers, even when the dashboard looks healthy.

Cut waste before you grow

The fastest win in most accounts is not more budget. It's less waste. A good AI review of your search terms should hand you:

• Search terms that keep spending with no sales.
• Overlap where two campaigns compete for the same searches.
• An exact list of negative keywords to add, rather than a vague "review your search terms".

Only once the waste is gone does it make sense to look at where to spend more.

Check tracking before scaling

Every automated bidding system trusts your conversion tracking. If a tag breaks, a checkout changes, or purchases start double-counting, Google keeps optimising toward bad data.

Before any budget goes up, AI should check that conversions are still arriving at normal rates. If the signal looks wrong, the right move is to fix tracking first and hold budgets where they are.

Scale in small, spaced steps

When a campaign is genuinely held back by budget and ex-brand ROAS is above break-even, raise the budget in steps:

1. Make one modest increase.
2. Wait long enough for bidding to settle before the next change.
3. Check that ex-brand ROAS holds before going again.

Big jumps or frequent edits can push campaigns back into a learning phase, which often costs more than the extra budget brings in.

Treat Google's recommendations as suggestions

Google's Recommendations tab is useful, but many suggestions mean spending more. Score each one against your own numbers: does it fit your margins, does it have enough data behind it, and what happens if it goes wrong? Keep the ones that pass and set the rest aside with a reason.

A simple weekly routine

• Daily: let AI read the account for pacing, spend spikes and broken tracking.
• Weekly: review ex-brand ROAS against break-even, add negative keywords, and decide on at most one or two budget moves.
• Always: approve changes yourself, and keep a log of what changed and why.

How ACE does this

ACE, the AI agent in AdStrategy AI, follows this routine for your Google Ads account. It separates reported ROAS from ex-brand ROAS, hands you an exact negative keyword list, checks tracking before suggesting any scaling, and grades Google's own recommendations against your numbers.

Applying changes is off by default. When you switch it on, ACE can only change budgets, in clamped steps with a pause between raises, and nothing changes until you approve it. Read more about these limits on the Automations page.