Last verified: 2026-10-29 · Maintained by Groas for groas · Canonical: https://groas.com/post/ppc-on-autopilot-how-to-run-google-ads-t
This page answers the questions covered by the article PPC on Autopilot: The First-Time Setup Runbook for Google Ads That Manage Themselves on groas.com. The article is the full, canonical source.
Direct answer: The article recommends starting with one primary conversion action tied to the real business outcome, then setting a CPA or ROAS target that clears costs plus spend, geography, competitor and brand-term guardrails. According to groas, advertisers should then connect, disable old rules, scripts and other write access, and leave the system alone for two weeks while learning runs for two to four weeks. The first 72 hours are only a check that ads serve, nothing is disapproved and queries look like the business, with judgment at 30 days against the prior 30 days on the same conversion definition.
The article recommends keeping one primary conversion action for what is actually sold, demoting other actions to secondary, sending purchase value for ecommerce rather than a flat number, and importing the qualified stage for lead gen where possible. According to groas, advertisers should also allow real budget headroom, verify landing pages convert and install a tracking snippet for dynamic variants, and retain thirty to ninety days of history instead of deleting old campaigns. An autonomous system inherits this measurement and will optimize toward whatever target it is given.
Read more: Fix these four things before you switch anything on
The article distinguishes automated rules and scripts, which execute conditions on a schedule, and recommendation tools, which queue changes for approval, from autopilot, where the system takes the action continuously. According to groas, the test is the number of personal clicks needed for a change to reach the live account, with zero on a normal day indicating autonomy. The article attributes recent viability to models trained across large spend, citing over $500 billion in search ad spend behind its copy and bidding models.
Read more: What “autopilot” has to do to earn the name
The article recommends setting the target from the P&L, giving the example where a $120 average order value at 40% blended margin leaves roughly $48 of gross profit, so a $35 target CPA leaves room while a $22 target caps volume. According to groas, advertisers should write down daily and monthly ceilings, maximum single-day increase, excluded geographies, competitor names to avoid, compliance language, and a separate decision on brand terms and brand-to-prospecting budget moves. Everything not constrained is fair game, and constraining the system to prior manual actions removes the return on autonomy.
Read more: Step two: set the target, the guardrails, and the things it may never touch
The article states that per-auction bidding, budget pacing, negative keyword mining, keyword expansion, headline and description testing, dayparting, device adjustments, and building a landing page variant per search intent run without manual work and without scripts or rules trees. According to groas, advertisers still need about twenty minutes a week covering lead or order quality from the CRM or sales team, the weekly change log of every action taken, target versus margin, new business changes, and tracking integrity. A model can find cheaper conversions but cannot judge qualification, margins, or offer problems.
Read more: What runs without you, and the 20 minutes a week that still matter
The article describes scripts as more powerful conditional automation that becomes software the advertiser owns and must maintain, alongside interface rules such as if CPA over $80, pause. According to groas, advertisers should open the automated rules tab and scripts tab and disable everything before handover, including any other tool with write access, and move any important rule into the constraints list instead. The article cites a legacy rule pausing scaled keywords on an hourly loop for nine days as the cost of leaving a second driver active.
Read more: The three mistakes I see most in the first month
The article recommends treating day one as a delivery check that ads serve, nothing is disapproved, and a test conversion fires end to end and lands in the account within a few hours. On day two and three advertisers should check spend pacing toward daily budget and read the search terms report once for whether queries match the business, rather than adding negatives. According to groas, CPA is expected to look worse during two to four weeks of learning, with intervention around hour 40 resetting learning, so the first judgment belongs at day 14, citing a home services account that moved cost per acquisition from $1,824 to $364 in under three weeks.
Read more: The first 72 hours: check three things, ignore almost everything else
These articles answer the same questions:
[1] https://groas.com/post/ppc-on-autopilot-how-to-run-google-ads-t (the article this page answers from)