The task
The agency says it has carried out hundreds of Google Ads audits and wanted them to be more thorough and less likely to miss things. A manual audit of an account usually gets an hour or two, which does not leave time to read through thousands of negative keywords looking for gaps.
What they did
The agency built the tool with Claude Code. It started with a long text file of its own best practices, so that recommendations follow the agency's way of working instead of generic advice. The tool pulls account data directly from the Google Ads API: keywords, ad copy, extensions, locations, settings, devices, search terms, campaigns, and ad groups. Plotly turns that data into tables and charts. The audit has 17 sections, from keyword strategy to ad copy and bid strategies. Each section has its own prompt, which calls the Anthropic API with the best-practice file and the account data, knows what the other sections cover, and ends with a recommendation of one or two sentences. An 18th prompt reads the section texts and writes the account-wide summary. The agency compared Claude Sonnet with Claude Opus 4.7 and chose Opus. The finished report is styled from the agency's design file and exported to PDF with WeasyPrint and Kaleido. After each audit the team writes feedback and adds it to the best-practice file.
What the source reports
According to the source, each audit has 17 sections plus an account-wide summary and costs $5 - 10 in Claude tokens. The source gives one example of what the tool catches: an accidental negative keyword such as 'kit' that could block most of a kitchen renovation company's relevant searches, which a short manual audit may not reach.
Where the approval sits
According to the source, each section of the AI-written audit can be edited by a person. The team uses the output as a template, checks the account itself for further issues and opportunities, and reviews the result as a whole. The article does not say who signs off before the PDF goes to a client. It describes the tool reading account data and writing a report, and it does not describe the tool changing anything in the Google Ads account.
Worth knowing
The source is the agency's own blog post on its marketing site, so every claim is the agency's. It reports a cost per audit, $5 - 10 in tokens, without stating a currency, but no time saved, no error rate, and no before-and-after results. The post says the aim was a better audit, not a faster one. The tool is a custom build that needs API access to the Google Ads account and to a model provider, and its quality depends on the best-practice file, which the agency keeps revising after each audit.
Source: toogooddigital.co.nz · May 2026
Drafted with AI assistance and checked automatically. A person is responsible for the collection. How these entries are made
Prompt for your agent
Sort search terms from a report
Sort the search terms in a pasted report into keep, negative candidates and watch, with a reason you can check for each term.
More in this category
- Media and ad operations
Google Ads Ask Advisor explains performance drops and applies changes with approval
Google documents an assistant inside Google Ads that answers questions about campaign performance, ad disapprovals, and billing, and suggests changes. The advertiser approves a change before it is applied.
Read the entry - Media and ad operations
The Trade Desk describes Koa agents that review media plans and troubleshoot campaigns
The Trade Desk describes conversational Koa agents in its Kokai platform that help build a campaign from an uploaded media plan, flag setup problems, and troubleshoot campaigns that are not spending. The feature is in closed beta, and the buyer decides what to apply.
Read the entry
Ninety minutes live on your tools. One automation running before we hang up.