Google Ads autopilot for auto shops that grades its own work.

Early Monday morning (05:30 Mountain time, before your day starts) the optimizer reads each live campaign's last 28 days — spend, clicks, search terms, and booked jobs traced from your own invoices — and produces at most ten bounded changes: block wasted searches, pause losers, shift budget toward what books work, tune the target, trim the schedule. In suggestions-only mode (the default) all of it is advice waiting in your dashboard. In autopilot mode, the levers you've enabled are applied, each one logged with the metrics as they stood. Either way, everything it decided not to do is listed too, with the reason.

Every change is stored with a baseline — the numbers at the moment it acted. Seven days later a nightly sweep records what actually happened and scores the change by whether your cost per booked job fell or rose, computed from paid invoices rather than Google's conversion counter. That signed score is what the next Monday reads: levers with a good measured record at your shop are favoured, and levers with a bad one are held back or switched off, with the numbers quoted in the list.

Two things, automatically. Any applied change that measures at −25% or worse on your cost per booked job is proposed for reversal at the top of the next Monday run, before any new advice — budget shifts are shifted back, paused ad groups resumed; where no clean reversal exists it says so and asks you to undo it by hand rather than improvising. And the lever behind it starts accumulating a record: three measured runs averaging worse than −10% with under 34% helping — or two runs at −50% or worse — and that lever is suppressed for your shop until the record changes.

Not until you hand them over, campaign by campaign. Imported campaigns arrive report-only — importing is reading, never consent to act. Once you do opt a pre-existing campaign in, it gets extra caution on top of every normal guardrail: no pauses until 60 days of that campaign's own history is visible, double the click-and-spend evidence on first contact (50 clicks instead of 25 for an ad group, 24 instead of 12 for a keyword), no bidding changes on strategies it didn't set, and an outright exclusion for any campaign whose structure didn't import. The doubled bar lapses only after Service VIN has applied something there and measured it.

The difference is the yardstick and the receipts. This optimizer is graded against cost per booked job from your invoices — a signal Google never sees — so a change that produces more form fills but fewer paid jobs counts as a loss here. It is capped at ten changes a week with per-lever ceilings, it keeps a written log of every action and every non-action, it re-measures each change seven days later, reverts what backfired and suppresses what keeps hurting your shop specifically. And by default it changes nothing at all: you read the advice first and choose which levers to hand over.

No. The weekly pass is arithmetic over your real numbers — every rationale names the campaign, the dollars and the counts it is arguing from, and the pass costs 0 AI credits. Models appear in exactly one place in Ads Autopilot: proposing keyword terms and reading your website during a campaign build. Even there the rule is structural — a model may propose terms, never a number. Anything numeric a model writes is scrubbed in code, and every search volume comes from Google's Keyword Planner priced for your city, or is shown as no estimate rather than a guess.

The import starts the night you connect: campaign history lands first, and the full 2-year backfill completes over roughly three nights (it walks 90-day chunks, at most 3 a night). Once data is in, the free audit can run on demand and refreshes every Monday morning. The optimizer itself reasons over a 28-day window, so its first Mondays lean on data volume — and it says “not enough data yet” under its floors instead of guessing. Nothing about this is instant, and we'd rather tell you that here than surprise you after signup.

The measuring needs no add-on — it comes with your plan (integrations ship from Growth up, and the free trial includes everything): connecting your Google Ads account read-only, the ~2-year import, the weekly account audit with on-demand re-runs, and the Campaign ROI report. The paid add-on is for acting — building campaigns, letting the optimizer touch the account, syncing audiences — and it sits on top of whichever plan you're on as a flat monthly fee, never a percentage of your ad spend. There's deliberately no self-serve checkout yet: turning it on is a short conversation, so you can ask the awkward questions before anything recurring exists.