What the Bot can read in Google Ads
Through MCP Ads, the Bot reads the same reporting you would open in the Google Ads interface: account totals and a daily trend; campaign, ad group, keyword and ad performance; search terms; breakdowns by device, location, hour and demographics; quality scores with their three components; budget pacing; impression share; conversion actions; Google’s own recommendations; and the account’s change history.
Manager (MCC) accounts are covered: the Bot can list client accounts under a manager and find an account by name, which matters when one Bot checks several clients. More on the tool set is on the Google Ads MCP server page.
Connect it in a minute
Before you paste it, open the MCP Ads dashboard, go to Connect MCP → API keys and create a key. It is shown once, so copy it then. Make one key per Bot, so you can revoke one without breaking the others. Grok Bot's sign-in flow for custom servers does not complete today, which is why the Bot uses the key header; in Grok chat or Grok Build you can instead add a custom connector at grok.com/connectors with the same URL and sign in with the browser.
Then paste this into Grok Bot (connect Google Ads in the MCP Ads dashboard first):
Connectors in Grok are account-wide, so once MCP Ads is added every Bot you make can use it. The Bot sees only the platforms and ad accounts you connected in MCP Ads, under the same workspace scope as the web app. MCP Ads is a free trial, then $49/month flat; Grok bills its own usage separately. The full setup is on the Grok Bot page and in the quickstart.
- Add a custom MCP connector:
- - Name: MCP Ads
- - URL: https://mcp-ads.com/mcp-slim
- - Header: Authorization: Bearer <paste your mcp_live_ key>
- Then list my Google Ads accounts and tell me which campaign spent the most last week with the fewest conversions.
The checks worth putting on a schedule
Search terms. The query report shows what people actually typed. Terms with spend and no conversions over 30 days are the usual first source of waste. Ask the Bot to group them into themes (jobs, free, a competitor’s name, the wrong city) rather than listing 200 rows, and to leave out terms with too few clicks to judge.
Budget pacing. Pacing shows each active campaign’s daily budget, what it spent and the share used. A campaign that spends its whole budget by early afternoon every day is missing the evening; one that uses a fraction may have a targeting or bid problem.
Impression share. Google reports the share of eligible auctions you won, and the share you lost because of budget or because of rank. They point at different fixes, which is why a Bot should report them separately.
Change history. A weekly look at what was changed, when and by whom catches the edit a colleague or an automated rule made on Tuesday that explains Thursday’s drop.
| Check | Good cadence | What to act on |
|---|---|---|
| Search terms with spend, no conversions | Weekly | Negative keyword themes you approve |
| Budget pacing | Daily | Campaigns capped early or badly underspending |
| Impression share lost to budget vs rank | Weekly | Budget or targeting (budget); bids, ads, pages (rank) |
| Quality score components | Monthly | Below-average expected CTR, ad relevance or landing page |
| Change history | Weekly | Edits that line up with a change in results |
Prompts to give the Bot
Scheduled prompts work best when they say what to leave out. "Only the three biggest changes" gets you a message you read; "a full report" gets you one you skim.
- “Every Monday at 8, compare last week with the week before in Google Ads and message me only the three biggest changes in spend, conversions or cost per conversion.”
- “Every day at 14:00, check budget pacing and tell me any campaign that has already used more than 90% of its daily budget. Say nothing otherwise.”
- “Every Friday, list search terms with spend and no conversions in the last 30 days, grouped into negative keyword themes. Propose which to add to a shared negative list. Do not add them until I say which.”
- “Which campaigns lost the most impression share to budget last month, and which to rank?”
- “Show keywords with a quality score of 4 or less and which component is below average.”
- “What changed in the account in the last 7 days, and did anything line up with the drop in conversions on Thursday?”
- “List Google’s open recommendations and tell me which you would dismiss and why.”
What it will not do without your approval
The Bot can prepare changes: negative keywords, a shared negative list, a budget change, a paused keyword, a new responsive search ad, a whole new campaign. Every write goes through the approval and workspace controls in MCP Ads, and goes live when you approve it. New campaigns are created paused, so even an approved build spends nothing until you enable it.
Two habits help. Ask for the change as a list with the reason for each line, so approving is a read rather than an investigation. And for anything touching several campaigns at once, such as pausing everything, ask the Bot to name the campaigns and their last-7-day spend first.
Honest limits
Performance Max campaigns report search categories, not exact queries, and without spend: that is a Google API limitation, so a Bot cannot build a PMax wasted-spend list the way it can for Search.
Auction Insights competitor domains, overlap rate and outranking share are not in the Google Ads API; they exist only in the interface. The Bot reports the impression-share figures the API does provide.
Conversion data is only as good as your tracking. If conversions are misfiring, every recommendation built on them is wrong; the conversion tracking guide and the audit checklist are the place to start before trusting a weekly report.
