---
title: "One Campaign, One Hour, One Change: A Claude Google Ads Playbook"
description: "Give Claude one campaign, 60 minutes and at most one approved change. Check tracking first, investigate one waste source, and leave with an action or assigned blocker."
url: "https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/47f1eddd-e1dc-47c0-9b54-bcdf03327cef.png"
published: "2026-11-20T16:45:32.776Z"
modified: "2026-11-20T16:45:32.873Z"
---

November 20, 2026 · 10 min read

# One Campaign, One Hour, One Change: A Claude Google Ads Playbook

[](https://groas.com/author/alexander-perelman)

[Alexander Perelman](https://groas.com/author/alexander-perelman)Personally Tested

Head Of Product @ groas

[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![A thick audit binder stripped of its sticky tabs except one bright orange one, held shut by a kitchen timer set to sixty minutes, as discarded tabs litter the desk.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/47f1eddd-e1dc-47c0-9b54-bcdf03327cef.png)

In this article

1. [1. Before minute zero: connect once and choose a campaign worth diagnosing](#1-before-minute-zero-connect-once-and-choose-a-campaign-worth-diagnosing)
2. [2. Minutes 0–10: name the conversion that earns the budget](#2-minutes-010-name-the-conversion-that-earns-the-budget)
3. [3. Minutes 10–25: decide whether the conversion data deserves your trust](#3-minutes-1025-decide-whether-the-conversion-data-deserves-your-trust)
4. [4. Minutes 25–40: investigate one pile of apparent waste](#4-minutes-2540-investigate-one-pile-of-apparent-waste)
5. [5. Minutes 40–50: rank the actions and name the downside](#5-minutes-4050-rank-the-actions-and-name-the-downside)
6. [6. Minutes 50–60: approve once, then verify the exact change](#6-minutes-5060-approve-once-then-verify-the-exact-change)
7. [7. At the hour’s end: close the decision, not the whole account](#7-at-the-hours-end-close-the-decision-not-the-whole-account)
8. [8. If blocked: assign the missing proof within the same hour](#8-if-blocked-assign-the-missing-proof-within-the-same-hour)
9. [9. When the constraint lifts: make the next hour earn itself](#9-when-the-constraint-lifts-make-the-next-hour-earn-itself)

You get one campaign, one hour, and at most one approved account change. A 20-page audit grade does not get a vote.

I have run those free graders. They score every business against the same fixed metrics and hand back a letter grade plus a list of next steps. Useful if you want a pat on the head. Not much help when a campaign has 490 clicks and zero conversions, and you need to know whether tracking failed, targeting failed, or the conversions simply have not arrived yet.

I used to tell clients the audit was the work. I was wrong. Too often, the audit is procrastination with charts.

Here is the replacement: pick your worst important campaign, connect Claude to your Google Ads data through groas, and force a decision before the timer ends. **Success is one defensible next action, logged and reversible, or one clearly assigned blocker.** If conversion measurement cannot be trusted, verification _is_ the deliverable. You do not earn an optimization just because you booked an hour.

## 1. Before minute zero: connect once and choose a campaign worth diagnosing

**Cost: setup before the clock. Buys: one live connection and a scope worth spending 60 minutes on.**

This is not a free grader. [groas](https://groas.com/) is the infrastructure that connects Claude to your actual ad accounts and lets it act within guardrails you set. It costs $19 a month, flat. That connection makes the hour possible: Claude can inspect the account, and, only with your approval, make the selected change.

In Claude, add a custom connector named groas with `https://mcp.groas.ai/mcp`, then sign in and connect your Google Ads account. Do this before starting the timer. The hour is for a decision, not account setup.

Choose one campaign that matters to the business. Skip this optimization hour if it spent under $500 in the last 30 days or launched less than two weeks ago. That is my screening rule, not a magic threshold. With little data, Claude can produce a very tidy ranking of noise. Let traffic accrue and return when there is enough evidence to judge it. Tracking concerns still need verification; low spend does not make broken measurement acceptable.

Leave autopilot off. Then start the timer and paste:

> Inspect campaign \[ID]. Make no changes. Return evidence, uncertainty and the first decision I need to make.

**Read-only first. No account changes before minute 50.** [Claude through groas](https://groas.com/claude) can inspect and act, which is precisely why the instruction needs both halves.

## 2. Minutes 0–10: name the conversion that earns the budget

**Cost: 10 minutes. Buys: a shared definition of what this campaign must produce.**

Claude will happily analyze your whole account into uselessness if you let it. Don’t. Give it:

- One campaign ID and the account ID.
- The account currency and reporting date range.
- The conversion action that decides budget: a booked job, a closed-won meeting, a purchase with margin, whatever actually pays.
- The business context it needs to interpret that conversion.

Then ask: “What changed in the last 30 days, and what context is missing?”

The second question matters as much as the metrics. Someone may have swapped a landing page, paused keywords or changed a conversion action. [Google Ads Change History keeps two years of edits](https://support.google.com/google-ads/answer/2454137?hl=en\&ref_topic=2795214), but account history is not the whole business history. Ask about changes outside the account, too.

If Claude cannot name the conversion that matters or establish what materially changed, the diagnosis is not ready. Resolve that within this block or record the missing context as a blocker. Do not quietly expand the hour.

What gets cut: the Quality Score lecture, the account-structure grade and the best-versus-worst ad beauty contest. Those are easy to compute. They do not settle whether this campaign deserves another dollar tomorrow.

Leave this block with one campaign, one business outcome and the context needed to connect them.

## 3. Minutes 10–25: decide whether the conversion data deserves your trust

**Cost: 15 minutes. Buys: permission to diagnose performance, or a reason to stop.**

Ask Claude to list the campaign’s conversion actions, their statuses, when they last fired and their counts over the last 30 days. Then check both ends of the journey separately:

- **Business end:** did the lead or order reach the form, inbox, call log, booking tool or backend?
- **Reporting end:** did the corresponding conversion reach Google Ads under Goals, then Summary?

A status such as Unverified, Tag inactive or Needs attention calls for verification, not a confident explanation of why the ads failed. Use [Tag Assistant to investigate the tracking](https://bytown.co/guide/google-ads-conversion-tracking-not-working) before treating the zeros as performance evidence.

A campaign with 490 clicks and no recorded conversions can invite a remarkably confident targeting diagnosis. But if nobody can tell whether a real lead reaches both the business and the reporting system, confidence is decoration.

**Stop rule: if measurement is unresolved, measurement is the task.** No bid change. No negative keyword. No pause disguised as caution.

The deliverable becomes a verification assignment: a human test of the conversion journey, a backend check, a Google Ads check and a named owner for any tracking fix. Where click attribution is part of the problem, include that in the verification. A page refresh alone does not establish that the ad-to-conversion journey works.

If the result cannot be confirmed within this block, do not fill the remaining time with optimization. Use it to make the assignment executable, following step 8. You finish with a blocker, not an account change.

## 4. Minutes 25–40: investigate one pile of apparent waste

**Cost: 15 minutes. Buys: three candidates, not a sack of automatic negatives.**

Only enter this block if measurement is usable and the necessary business context is available.

Ask Claude for reported search terms in this campaign with spend and zero recorded conversions over the last 30 days, sorted by cost. Say you spend $20k a month: a review filter such as $75 with no conversions gives you a manageable starting point. Look at the top 10.

That filter is a desk to put papers on, not a machine that adds negatives.

For each term, require:

- Spend and clicks.
- Search intent in plain English.
- Whether that intent fits what the business sells.
- What remains unknown.

The [hands-on search-term walkthrough](https://groas.com/post/find-wasted-google-ads-spend-in-one-prompt-a-search-terms) shows this pass inside Claude. Its value is the narrowness: one pile to inspect instead of a 40-tab audit.

Two traps still need attention.

**First, timing.** Conversions can arrive after the click. Judging last Tuesday’s CPA before the campaign’s conversion lag has cleared is judging mail that has not arrived yet. Check the campaign’s actual conversion window and lag rather than treating every recent zero as final.

**Second, visibility.** [Google does not show every search query](https://support.google.com/google-ads/answer/2472708?hl=en). Low-volume terms can remain hidden for privacy, so spend can be real while the query behind it never appears in the report.

Treat this as a review of _apparent waste in reported terms_, not a complete inventory of wasted spend. Never auto-block from this list. Leave with three candidates and a reason to distrust each one. If the evidence supports fewer than three, do not manufacture the rest.

![Kitchen timer beside a laptop displaying Google Ads search terms with spend and no conversions](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/98b79a02-6632-454e-ad96-bef06bf95a9f.png)

## 5. Minutes 40–50: rank the actions and name the downside

**Cost: 10 minutes. Buys: one decision with a mechanism, not another recommendation list.**

This is where audits go to die. Run it like triage.

Ask Claude to rank the candidates by evidence strength, business relevance and reversibility. A generic grader’s checklist does not become business judgment just because the output includes a score. Your orders, margin and target CPA matter. If that context is missing, Claude needs to say so.

For each candidate, require four things:

1. **The exact action and entity.** For example: pause keyword X, add negative Y as phrase match, or reduce campaign Z’s budget by 30%.
2. **The mechanism before the expected effect.** For example: the query brings job seekers rather than buyers, so excluding it should reduce spend on that intent.
3. **The downside if the diagnosis is wrong.** Which useful traffic or volume could disappear?
4. **The missing evidence.** What would make the decision firmer, and can the decision wait for it?

My rule: if Claude cannot name the downside, the candidate is disqualified. “Improve efficiency” is not a mechanism. “Low risk” is not a downside analysis. Make it use nouns.

Then pick one. Not one negative plus a small budget adjustment. Not one recommendation with three implementation details that happen to change three things. **One approved account change is the ceiling, not the target.**

Picking none is valid when the evidence does not clear the bar. Write down why and what would change the decision. That is restraint, not an incomplete audit.

## 6. Minutes 50–60: approve once, then verify the exact change

**Cost: 10 minutes. Buys: proof that the approved change happened where you intended.**

Before Claude touches anything, state the exact entity and value: campaign name, ad group name, keyword text and match type, or budget amount and currency. Approve that one change and nothing else.

I leave autopilot off for the entire hour. Autopilot is for running a system you trust. You do not trust this campaign yet. That is why you are here.

After Claude reports completion, check two places:

1. The groas action log, including the reason recorded for the change.
2. The account itself, confirming that the status, budget or negative matches your approval.

Then record how to reverse the change. Do not confuse a logged action with a verified one, or assume you will remember the old value later. The undo note belongs beside the decision while the reasoning is still fresh.

Set a follow-up based on the campaign’s conversion lag. Not tomorrow morning because the calendar has a gap. Not an arbitrary date that arrives before the conversions do. The purpose of the check-in is to judge the change after enough time has passed to make the comparison useful.

Your handoff is short:

- One approved change.
- One verified log entry.
- One undo note.
- One dated follow-up.

If you want the wider sweep later, the [full 2026 Google Ads audit checklist](https://groas.com/post/google-ads-account-audit-checklist-2026-red-flags-fixes) is there when the timer is off. It does not get smuggled into these last 10 minutes.

## 7. At the hour’s end: close the decision, not the whole account

**Cost: no extra time. Buys: a usable handoff instead of an unfinished investigation.**

The hour ends with either the verified change above or an explicit reason not to make one. No overtime to polish the report. No second campaign because Claude spotted something interesting.

If you chose no change despite usable measurement, save the decision in plain language: what you reviewed, why the evidence was insufficient and what you need before reconsidering it. That keeps the next session from repeating the same analysis with a slightly different prompt.

**The constraint applies to the deliverable, too.** You do not need 20 pages explaining why you made one decision. You need enough evidence that someone can understand it, challenge it and reverse it if necessary.

## 8. If blocked: assign the missing proof within the same hour

**Cost: the unused portion of the original hour. Buys: one owner, one due date and a restart condition.**

I used to treat a blocked hour as failure. It is not. An unresolved tracking problem does not improve because Claude finds something else to optimize.

For a measurement blocker, write the task plainly: verify that the test conversion reaches both the backend and Google Ads, and identify or fix the break if it does not. Assign a person and a date. “Marketing to investigate” is not an owner.

Missing business inputs get the same treatment. Claude cannot sensibly judge your target CPA without the relevant margin, close-rate and qualification context. Send the question to whoever owns those numbers, with a 48-hour deadline.

Save the exact next prompt, too:

> Re-run the 30-day search-term review for campaign X now that purchase tracking is verified. Make no changes.

**Leave autopilot off. Save the question instead of generating another audit.** The blocker is complete when someone knows exactly what to prove, by when, and what happens after they reply.

## 9. When the constraint lifts: make the next hour earn itself

**Cost: a separate, deliberately approved session. Buys: another bounded decision, not permission to change everything.**

When more time becomes available, I do not start a bigger audit. I run the same hour on the next campaign that matters, using the first hour’s log as context.

Suppose the first campaign eventually shows that one phrase-match negative cut $380 a month of job-seeker spend. The next hour starts with that evidence and asks where else the pattern appears. It does not assume the same negative belongs everywhere.

Only after two consecutive sessions produce approved, reversible changes that hold up after the conversion-lag window would I consider handing that narrow job to autopilot, inside a budget cap I set. That is a gate, not a reward for finishing the timer.

**A grade tells you where you stand. A constrained hour tells you what to do next, including when not to optimize.** Start with one campaign, keep the timer visible, and let the second hour earn itself.

## Frequently Asked Questions

### What is the one-hour Google Ads playbook with Claude?

You pick your worst important campaign, connect Claude to your Google Ads data through groas, and force a decision within 60 minutes and at most one approved account change. Success is one defensible, logged, reversible next action, or one clearly assigned blocker. If conversion measurement cannot be trusted, verification is the deliverable instead of an optimization.

### How do I set up groas and which campaigns qualify for the optimization hour?

In Claude, add a custom connector named groas with https\://mcp.groas.ai/mcp, sign in, and connect your Google Ads account before starting the timer. groas costs $19 a month, flat. Choose one campaign that matters to the business, and skip the hour if it spent under $500 in the last 30 days or launched less than two weeks ago.

### What context should I give Claude before it diagnoses the campaign?

Give it one campaign ID and account ID, the account currency and reporting date range, the conversion action that decides budget such as a booked job or purchase with margin, and the business context needed to interpret that conversion. Then ask what changed in the last 30 days and what context is missing. Account history is not the whole business history, so ask about changes outside the account too.

### Why does a campaign with lots of clicks and no conversions not automatically point to targeting?

If nobody can tell whether a real lead reaches both the business and the reporting system, a confident targeting diagnosis is decoration. Check the business end, such as the form or booking tool, and the reporting end under Google Ads Goals, then Summary. Statuses like Unverified, Tag inactive or Needs attention call for verification with Tag Assistant before treating the zeros as performance evidence.

### How do I find wasted spend in search terms without adding bad negatives?

Ask Claude for reported search terms with spend and zero recorded conversions over the last 30 days, sorted by cost, and review the top 10 for spend, clicks, intent, fit and what remains unknown. Remember conversions can arrive after the click, so check the campaign's conversion window before judging recent zeros. Google also hides low-volume queries, so spend can be real while the query never appears. Never auto-block from the list; leave with three candidates and a reason to distrust each.

### How should I choose which Google Ads change to approve?

Ask Claude to rank candidates by evidence strength, business relevance and reversibility, and require the exact action and entity, the mechanism before the expected effect, the downside if the diagnosis is wrong, and the missing evidence. If Claude cannot name the downside, the candidate is disqualified. Pick one change or none; picking none is valid when the evidence does not clear the bar.

### What do I check after Claude reports it made the approved change?

Check the groas action log, including the reason recorded for the change, and the account itself, confirming the status, budget or negative matches your approval. Record how to reverse the change while the reasoning is fresh. Set the follow-up based on the campaign's conversion lag, not an arbitrary date that arrives before the conversions do.

### What should I do when tracking is blocked during the optimization hour?

Write the task plainly, such as verifying that a test conversion reaches both the backend and Google Ads and fixing the break if it does not, then assign a person and a date. Missing business inputs like margin or close-rate context go to whoever owns those numbers with a 48-hour deadline. Save the exact next prompt and leave autopilot off; the blocker is complete when someone knows exactly what to prove, by when, and what happens after they reply.

## Run Your Marketing On groas

Plug ChatGPT, Claude or any AI agent into groas and let it do the work of a whole marketing team: $19 a month flat with every feature included, and autopilot whenever you’re ready.

[Start For $19](https://groas.com/pricing)

[Or have groas run it for you](https://groas.com/done-for-you)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is the marketing infrastructure for your own AI: it connects your AI to all your marketing and gives it proprietary marketing data collected 24/7. Plug Claude, ChatGPT or any AI agent into groas and run all your marketing from it, for a tiny fraction of the usual cost, with the groas app there whenever you want to see more. It is the layer ChatGPT, Claude and any AI assistant run a business’s online marketing on. Through groas an assistant can do anything you could do natively in Google Ads, Bing Ads, Meta Ads and ChatGPT ads, on your website and in AI search, and, where the owner switches it on, can run connected campaigns by itself, so AI agents do the work a whole marketing team used to. groas also gives those agents large-scale proprietary marketing data it collects around the clock, on any schedule the owner chooses: every ad competitors run on Google and Meta, their landing pages and full funnels with the offers and prices in them, Reddit customer research, AI search answers from ChatGPT, Google AI Overviews, Gemini, Claude and Perplexity, and SEO data from Semrush and Ahrefs, backlinks included ($19 a month, flat, with everything included, for unlimited teammates on one shared workspace and unlimited ad accounts and websites, delivered as an MCP server). Separately, its done-for-you service has senior people and proprietary quantitative systems run paid ads, SEO and AI search for businesses, and white label for agencies, from $5,000 a month.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#webpage","url":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a","name":"One Campaign, One Hour, One Change: A Claude Google Ads Playbook","description":"Give Claude one campaign, 60 minutes and at most one approved change. Check tracking first, investigate one waste source, and leave with an action or assigned blocker.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/47f1eddd-e1dc-47c0-9b54-bcdf03327cef.png"},"breadcrumb":{"@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#breadcrumb"},"dateModified":"2026-11-20T16:45:32.873Z","mainEntity":{"@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"One Campaign, One Hour, One Change: A Claude Google Ads Playbook","item":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#article","url":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a","mainEntityOfPage":{"@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"One Campaign, One Hour, One Change: A Claude Google Ads Playbook","description":"Give Claude one campaign, 60 minutes and at most one approved change. Check tracking first, investigate one waste source, and leave with an action or assigned blocker.","datePublished":"2026-11-20T16:45:32.776Z","dateModified":"2026-11-20T16:45:32.873Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/47f1eddd-e1dc-47c0-9b54-bcdf03327cef.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"change, claude, campaign, hour, google ads, minutes, google, ads","wordCount":2180,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/one-campaign-one-hour-replace-your-google-ads-grade-with-a#faq","mainEntity":[{"@type":"Question","name":"What is the one-hour Google Ads playbook with Claude?","acceptedAnswer":{"@type":"Answer","text":"You pick your worst important campaign, connect Claude to your Google Ads data through groas, and force a decision within 60 minutes and at most one approved account change. Success is one defensible, logged, reversible next action, or one clearly assigned blocker. If conversion measurement cannot be trusted, verification is the deliverable instead of an optimization."}},{"@type":"Question","name":"How do I set up groas and which campaigns qualify for the optimization hour?","acceptedAnswer":{"@type":"Answer","text":"In Claude, add a custom connector named groas with https://mcp.groas.ai/mcp, sign in, and connect your Google Ads account before starting the timer. groas costs $19 a month, flat. Choose one campaign that matters to the business, and skip the hour if it spent under $500 in the last 30 days or launched less than two weeks ago."}},{"@type":"Question","name":"What context should I give Claude before it diagnoses the campaign?","acceptedAnswer":{"@type":"Answer","text":"Give it one campaign ID and account ID, the account currency and reporting date range, the conversion action that decides budget such as a booked job or purchase with margin, and the business context needed to interpret that conversion. Then ask what changed in the last 30 days and what context is missing. Account history is not the whole business history, so ask about changes outside the account too."}},{"@type":"Question","name":"Why does a campaign with lots of clicks and no conversions not automatically point to targeting?","acceptedAnswer":{"@type":"Answer","text":"If nobody can tell whether a real lead reaches both the business and the reporting system, a confident targeting diagnosis is decoration. Check the business end, such as the form or booking tool, and the reporting end under Google Ads Goals, then Summary. Statuses like Unverified, Tag inactive or Needs attention call for verification with Tag Assistant before treating the zeros as performance evidence."}},{"@type":"Question","name":"How do I find wasted spend in search terms without adding bad negatives?","acceptedAnswer":{"@type":"Answer","text":"Ask Claude for reported search terms with spend and zero recorded conversions over the last 30 days, sorted by cost, and review the top 10 for spend, clicks, intent, fit and what remains unknown. Remember conversions can arrive after the click, so check the campaign's conversion window before judging recent zeros. Google also hides low-volume queries, so spend can be real while the query never appears. Never auto-block from the list; leave with three candidates and a reason to distrust each."}},{"@type":"Question","name":"How should I choose which Google Ads change to approve?","acceptedAnswer":{"@type":"Answer","text":"Ask Claude to rank candidates by evidence strength, business relevance and reversibility, and require the exact action and entity, the mechanism before the expected effect, the downside if the diagnosis is wrong, and the missing evidence. If Claude cannot name the downside, the candidate is disqualified. Pick one change or none; picking none is valid when the evidence does not clear the bar."}},{"@type":"Question","name":"What do I check after Claude reports it made the approved change?","acceptedAnswer":{"@type":"Answer","text":"Check the groas action log, including the reason recorded for the change, and the account itself, confirming the status, budget or negative matches your approval. Record how to reverse the change while the reasoning is fresh. Set the follow-up based on the campaign's conversion lag, not an arbitrary date that arrives before the conversions do."}},{"@type":"Question","name":"What should I do when tracking is blocked during the optimization hour?","acceptedAnswer":{"@type":"Answer","text":"Write the task plainly, such as verifying that a test conversion reaches both the backend and Google Ads and fixing the break if it does not, then assign a person and a date. Missing business inputs like margin or close-rate context go to whoever owns those numbers with a 48-hour deadline. Save the exact next prompt and leave autopilot off; the blocker is complete when someone knows exactly what to prove, by when, and what happens after they reply."}}]}]}
```
