Google Ads scripts and automated rules: what to automate
Automate the monitoring, not the decisions. Automated rules in Google Ads are scheduled checks that run on a timetable and compare a metric against a threshold you set. Scripts do the same thing, but with more data. They are useful for one thing: telling you that something has happened. For example, a campaign that has stopped, a disapproved ad, an exhausted daily budget, a broken feed or conversions that suddenly disappear. They become dangerous the moment you let them pause campaigns, move budgets or change bidding without a person involved. That is why the difference between a useful rule and a dangerous one comes down to a single line: a useful one sends a notification and leaves the decision to you, while a dangerous one acts on the account while you sleep.
My name is Hristiyan Panov — a performance marketing consultant with over 5 years’ experience and over €2 000 000 managed ad budget for more than 90 companies on Google and Meta. Below is what is worth automating and where automation does more harm.
Which checks to automate?
- A stopped conversion tag. A notification if no conversion has been recorded in the last 24 hours while there is active spend. This is the most useful rule in the whole account.
- A campaign with no impressions. This usually means a disapproved ad, an exhausted budget or a feed problem.
- A sharp jump in cost per click. An early signal of a new competitor or a change in bidding.
- Products disapproved in Merchant Center. See how to read the diagnostics.
- Budget exhausted before the end of the day when a campaign goes over your threshold.
Each of these is a notification, not an action. A person decides what happens next.
Which decisions should NOT be automated?
Pausing a campaign, changing a budget and switching a bidding strategy. The reason is not one of principle: these changes put the campaign back into a learning phase, and the rule has no idea whether a weak week is seasonality, a competitor, broken tracking or chance.
The classic scenario: a rule pauses a campaign because of three weak days, someone turns it back on a week later, learning starts from scratch — and this repeats month after month.
What makes the difference between a useful rule and a dangerous one?
- A long enough period. A rule based on three days of data at low volume measures noise.
- A minimum threshold. “Pause a keyword with spend above X and zero conversions” is reasonable; without a threshold it will cut keywords with two clicks.
- Notification instead of action, at least until you are convinced the rule is not getting it wrong.
- A log. A rule that acts silently will confuse you during the next analysis — you will look for a human reason behind a change that nobody made.
Scripts or rules?
Rules cover most needs and do not require code. Scripts become necessary when you need to combine data from outside — stock levels, margins, real orders — or report on something the interface does not show.
In practice, the difference is maintenance. A rule lives in the interface: anyone with access can see it, its condition can be read in seconds, and its runs stay in the account. A script is code that depends on an external source — a spreadsheet, a feed or an export. A renamed column, a moved file or a changed date format will break it, often without any visible error.
So the order is this: first describe the check as a rule. Move to a script only when the rule cannot reach the data.
Before you write a script, answer three questions:
- Who maintains it, when its author is not available.
- Where the data comes from and who is responsible for it.
- What happens when the source is missing — does it stop, or carry on with empty values.
If nobody on your team maintains the script, it turns into a liability. Start with rules.
What should you never automate?
A rule that touches the account at the height of the season. That is exactly when the cost of a mistake is highest and behaviour is most atypical.
Outside the season there are other cases where automatic action does more harm than good:
- Irreversible changes. Deleting keywords, ads or lists cannot be undone with one click. A pause is reversible; a deletion is not.
- Decisions based on an incomplete day. Some conversions are recorded with a delay. A rule that reads yesterday early in the morning sees fewer conversions than there really are.
- A metric you cannot verify. If you do not reconcile conversions against real orders, you are automating a decision based on a number you do not trust.
- Changes to tracking. The conversion tag is edited manually and checked immediately afterwards.
Before your peak period, open the list of rules and switch the active ones to notification only. Write down when you did it, so that they do not stay that way until the next season.
How do you check that the rules are not hurting you?
In the account change history. If it shows changes that nobody remembers making, you now know where they come from. Check it before every analysis of a decline — it saves hours of looking for the wrong cause.
The check runs in this order:
- Open the change history for the period of the decline and filter by author. Changes made by an automated rule do not carry a person’s name.
- Put the dates of those changes on the chart of the metric that dropped. If they match, you have found the cause.
- Open the log of the rule itself and see when it was triggered and how many items it affected.
A rule that has never triggered is also a signal — its threshold is wrong or its condition does not describe what you intended. Launch a new rule as a notification first, and a week later compare what it would have done with what you did.
FAQ — Frequently asked questions about automated rules
How many rules are reasonable?
A few that you understand, rather than many that you have forgotten. Every rule you do not understand is a hidden risk.
Do they work with Performance Max?
Control is more limited. They are useful mainly for spend notifications and feed problems.
Do they replace daily work?
No. They save you checks, not judgement. A rule does not read the search terms report.
How do I stop a rule that has made a mistake?
Turn it off and reverse the changes manually. That is why you keep a log — otherwise you do not know exactly what was changed.
What about Google’s own automations?
The recommendations in the account are not mandatory. Applying them automatically is a setting that I recommend keeping switched off.
Conclusion
Good automation tells you that something has happened. Bad automation decides for you, based on data it does not understand. The difference is usually only visible once a working campaign has already been taken apart.
If your account contains changes that nobody is responsible for, start with an independent review.
Request a quote for a digital audit →
Related articles: campaign scaling · disapproved items · Google Ads account audit
