Operational analysis · Google Ads
What an approval can and cannot prove
Ask Advisor can move from conversation to account action. The useful control is a record that survives after the chat ends.
- Save the proposal before approvalPreserve the affected scope, old value, suggested value and reason. Approval records intent only when the proposed state is still available for comparison.
- Reconcile the action in Change HistoryUse the timestamp, actor and applied diff to establish what reached the account. Do not assume the final mutation is identical to the conversational summary.
- Date the outcome review in advanceCost and conversion movement after a change is evidence of consequence, not automatic proof of cause. Protect a baseline and avoid overlapping interventions.
The most consequential button in Google Ads may soon sit at the end of a conversation. Ask Advisor can answer account questions, diagnose issues, create assets and implement suggested changes after the advertiser approves them. That shortens the distance between an explanation and a live account mutation.
The approval step is real and important. It confirms that a person authorized an action. It does not, by itself, preserve the state that person saw, show every field that ultimately changed, or establish whether later performance movement was caused by that action. Those are three different evidentiary jobs.
A practical Ask Advisor audit therefore needs three dated records: the proposal before approval, the applied change in Google Ads Change History, and the outcome window chosen before results arrive. Together they form a change receipt. Without one, a team may know that somebody clicked approve while remaining unable to reconstruct exactly what was accepted or fairly judge what happened next.
Approval should preserve the proposal before it changes the account
Google describes Ask Advisor as an agentic assistant that can use account context and saved information to personalize its help. It can suggest optimizations and, with approval, implement changes. Google also places responsibility on advertisers to decide whether those suggestions are accurate, relevant and appropriate for their business.
That responsibility is difficult to exercise retrospectively if the approval record contains only a success message. A reviewer needs the object that was considered: which campaigns or assets were in scope, what their previous values were, what new values were proposed, why the change was suggested and which performance interval informed the recommendation.
Three records answer three different questions
What did the human agree to?
Scope · old value · proposed value · rationaleWhat reached the account?
Actor · timestamp · object · applied diffWhat changed in performance?
Cost · conversions · value · confoundersThe distinction is not bureaucracy for its own sake. A conversational answer is optimized to help a person decide. An audit record is optimized to help another person reconstruct the decision later. One can be concise and contextual; the other must be stable and comparable.
What Google's approval promise actually establishes
Google says Ask Advisor asks for approval before taking action in an account. Its April 2026 safety update repeats that principle while describing policy reviews, troubleshooting, security monitoring and certifications. This is evidence of a human authorization boundary, not evidence that the product silently edits an account.
The boundary has two useful consequences. First, the advertiser can refuse a recommendation. Second, the organization can attach ownership to the decision rather than treating it as an anonymous machine event. The danger begins when teams treat that same approval as proof of scope, execution and business effect.
Ask Advisor remains a beta feature, and Google says it is not currently available in manager accounts. That matters to agencies and multi-account teams. A workflow designed around an MCC-level control room cannot assume the same assistant, context or evidence surface exists at every level. The receipt should name the customer account in which the proposal appeared and the user who approved it.
Change History can prove execution without proving causality
Google Ads Change History is the natural execution record. Google says it retains changes from the past two years, can be filtered by date, campaign, ad group, user and change type, and can place changes on the same timeline as performance metrics. It also distinguishes changes made through the Google Ads interface, automated systems and the API.
This answers a narrow but essential question: what changed in the account, when and under which available attribution. It does not tell the reviewer what the approver believed would happen, whether every part of the conversational proposal is preserved, or whether an observed conversion shift resulted from the change rather than auctions, seasonality, tracking or a simultaneous edit.
What each record can support
| Record | Strongest claim | What it cannot prove alone |
|---|---|---|
| Saved proposal | The scope and state presented for approval | That every proposed field was applied |
| Change History | The recorded mutation, time and available actor | Why the approver accepted it or what caused later results |
| Performance window | The account outcome after the event | Causality when other conditions also changed |
Change History has documented limits. Some account-level settings and changes made by Google representatives may not appear, and details can be limited. Google also recommends reviewing smaller batches, ideally fewer than 2,000 entities, when change sets are large. A blank or abbreviated row is therefore a prompt to investigate, not permission to reconstruct a convenient story.
Build the receipt before opening the approval dialog
The receipt should be created while the proposal is still visible. It can be a controlled ticket, experiment record or account annotation; the format matters less than the fields. For every action, capture the customer ID, affected campaign or asset, current state, proposed state, Ask Advisor rationale, approver and approval time.
- Export or screenshot the complete proposal rather than copying only its headline.
- Record the baseline period and exclude the most recent conversion-delay interval when relevant.
- Name the primary outcome and a guardrail before approval, such as conversions with cost per conversion.
- Declare which other account changes must remain frozen during the review window.
- Set the first reconciliation time and the later performance-review date.
This extends the measurement discipline from our Google Ads AI insights receipt. That article asks whether an AI explanation can be reproduced from its prompt, metric window and data freshness. The Ask Advisor receipt begins one step later: once an explanation becomes an approved account action, can the team connect intent, execution and consequence without collapsing them into one event?
A worked review of an approved recommendation
Imagine Ask Advisor presents a campaign optimization and offers to implement it. Before approval, the operator records the affected campaign, the existing setting or asset state, the proposed replacement, the date range used in the explanation and the intended outcome. The operator also notes that no budget, conversion action or targeting change will be made until the first review.
Immediately after approval, the operator opens Change History for the relevant account and a narrow time window. The applied objects and values are compared with the saved proposal. A match closes the execution check. A missing, partial or additional mutation keeps the change in investigation even if the interface displayed a success message.
The outcome review happens later, at a time chosen from the account's conversion cycle rather than from impatience. The operator compares cost, conversions and conversion value with the protected baseline and records material auction, tracking or seasonal changes. A better result supports continued use; it does not prove that the assistant alone caused the improvement. A worse result triggers a deliberate follow-up, not an automatic reversal Google has not documented.
Prevent overlapping automation from erasing the evidence
An Ask Advisor action rarely occurs in an empty account. Smart Bidding continues to react to auctions, auto-apply settings may operate, other users can edit campaigns and APIs can write changes. The audit must separate the event under review from this surrounding activity.
Use Change History filters to isolate the object, user and time range, then widen the view to find concurrent changes. If text controls are also being revised, follow the versioning method in our Google Ads text-guidelines change-control guide. If delivery changes after approval, do not confuse account mutation with eligibility; our Limited Ad Serving analysis explains why an approved ad can still face an advertiser-level reach constraint.
The cleanest experiment changes one decision surface at a time. When that is impossible, list every overlapping intervention in the outcome record and lower the strength of the conclusion. Operational honesty is more useful than a neat causal claim the account cannot support.
The control should make fast approval safer, not slower
A good receipt takes minutes because most fields already exist on the decision surface. Its purpose is not to reopen every recommendation in committee. It creates a compact handoff between the person who approves, the account record that executes and the analyst who later judges performance.
Google's own guidance for working with its agentic advisors emphasizes human expertise and review before major changes. The most mature response is therefore neither blind adoption nor refusal to use the assistant. It is to preserve enough evidence that speed remains reversible in practice, even when the product does not promise a dedicated rollback.
Ask Advisor compresses the time required to move from question to action. The change receipt protects the time required to understand what that action meant. Approval is the beginning of that record, not its end.
Frequently asked questions
Can Ask Advisor change a Google Ads account without approval?
Google says Ask Advisor requests approval before taking an account action. The article's control starts from that documented approval boundary and adds records for execution and outcome.
Does Change History identify an Ask Advisor action by name?
Google documents user, automated-system and API attribution in Change History, but its public Ask Advisor material does not guarantee a dedicated Ask Advisor label for every action. Verify the actual attribution in the account.
Is a performance increase proof that the approved change worked?
No. It is post-change evidence. Auction conditions, seasonality, tracking and simultaneous edits can also move performance, so the review should preserve a baseline and list confounders.
Can agencies use Ask Advisor from a manager account?
Google currently says the beta is unavailable in manager accounts. Agencies should identify the customer account in which the proposal and approval occurred and check current eligibility before designing an MCC workflow.
