Your First AI Google Ads Review: Build an Evidence Sheet First

The most useful first request to an advertising assistant is a bounded investigation: identify the account, explain the goal and show the evidence before proposing changes. A confident paragraph is easy to produce. A recommendation you can check needs dates, definitions, account context and a visible connection to your business.

This guide focuses on the first session after connecting. It complements a full Google Ads audit checklist by helping you judge whether an AI-generated finding deserves to enter your decision process.

Establish the account and reporting window

Ask the assistant to state the customer account, currency, time zone, selected period and comparison period. Check the account name against the business you intend to review. If you work with multiple locations, distinguish the manager account from the individual advertising account.

Write down whether the report includes incomplete days. Comparing Monday through Wednesday this week with an entire prior week makes almost any trend look alarming. Use matched weekdays when useful, and note promotions, closures, holidays and stock changes that make periods different.

Ask which requested data was unavailable. A tool that returns no data has not demonstrated zero sales or zero spend. The answer may reflect permissions, the selected account, a missing connection or the query's scope. The assistant should make those limitations visible before it draws a conclusion.

Translate the conversion label into a business event

List the conversion actions the campaign is trying to achieve. For each, record what physically happens: a payment clears, a quote is requested, a call reaches a duration threshold, or a confirmation page loads. Then ask whether that event is counted once per person, per interaction or per transaction in your implementation.

Google's primary actions can influence bidding when their standard goal is selected. Secondary actions are normally for observation, but inclusion in a custom goal is an important exception. Check the campaign's goal configuration as well as the action label. About primary and secondary conversion actions.

If a report calls every form a customer, stop there. An efficient-looking campaign may simply be good at producing a low-value event. The guide to primary and secondary conversions provides additional context, but your team's definition of a paid customer must remain explicit.

Recalculate one recommendation by hand

Imagine a hypothetical report showing $900 spend and 30 reported conversions. The assistant presents a $30 cost per conversion and recommends increasing budget. The arithmetic is correct: $900 divided by 30 equals $30.

Now the owner checks the records. Six events were repeated form submissions, eight were irrelevant inquiries and the remaining 16 were legitimate leads. Four of those leads paid. The relevant costs are $56.25 per legitimate lead and $225 per paid customer, before software or staff costs.

Those figures do not automatically justify cutting the campaign either. If each customer contributes $400 before acquisition, the initial result may be commercially useful. If each contributes $150, the economics require investigation. The point is that the recommendation changes when the denominator represents a business outcome.

Mark the example's 30, 16 and four as different measures. Do not overwrite the platform count with a CRM figure and imply the two systems have identical attribution or timing.

Require evidence for the proposed explanation

Ask the assistant to separate three statements: what the report shows, what might explain it, and what would distinguish competing explanations. For example, lower conversions could reflect a broken form, fewer qualified visitors, delayed reporting or genuinely weaker demand.

A useful next step might be to inspect the landing page or reconcile a small sample of leads. It need not be a budget change. Ask what finding would make the assistant withdraw its original recommendation. That question often reveals whether the recommendation is testable.

Review recent account changes before treating a performance shift as spontaneous. Google provides change history, including API and automated-rule changes, with details useful for investigating timing. Review your account history. Use the change-history walkthrough to organize that investigation.

End the session with a small decision queue

Limit the first review to three categories: measurement problems to verify, commercial questions for the owner, and candidate account changes. Give each item an evidence link or report reference, a responsible person and a next step.

Do not allow a long list of suggestions to masquerade as a prioritized plan. A tracking fault that invalidates all conversion data should usually be investigated before a marginal ad-copy variation. A stock shortage may be more important than a low clickthrough rate. Your business context should determine the sequence.

First-session worksheet

Save the evidence sheet. At the next review, you can see which uncertainties were resolved and which recommendations survived contact with the facts. A read-only session is successful when it reduces uncertainty, even if no campaign setting changes.

AdvisorPPC's free offer is intended for Google Ads reads; other providers and account writes require the appropriate paid product. Check current plans and the tools available in your chosen client before upgrading. If access fails, use support or [email protected] and include any reference code displayed. Do not share tokens or credentials in the request.