Review Microsoft Ads search terms with AI
Find the searches that cost money and never convert, add them as negatives, and keep the ones worth a keyword of their own.
Microsoft Advertising is often a second account, checked less often than Google, so a broad or phrase keyword can match searches you never meant to buy without anyone noticing. This guide runs a full search terms review in Claude or ChatGPT, one prompt at a time. It assumes your Microsoft Advertising accounts are connected through Adplane, so every answer comes from your live data. The example is Fernwood Outdoor, an outdoor gear store running three Search campaigns on Microsoft Advertising.
Step 1: the search terms report
Fernwood Outdoor on Microsoft Advertising, September 11 to October 10: 412 search terms, 1,840 dollars and 64 conversions across Generic tents, Hiking boots and Brand search. The three that spent most, a broad match that went wrong, and your brand term:
| Search term | Keyword | Served as | Spend | Conversions |
|---|---|---|---|---|
| 2 person backpacking tent | 2 person tent | Phrase | 96.40 | 5 |
| waterproof hiking boots women | waterproof hiking boots | Phrase | 88.10 | 4 |
| ultralight tent 2 person | lightweight tent | Broad | 58.70 | 3 |
| tent rental near me | camping tent | Broad | 44.30 | 0 |
| fernwood outdoor | fernwood outdoor | Exact | 31.50 | 19 |
Each row is one search paired with the keyword that matched it. "Served as" is the match type that actually served the ad. It can be tighter than the keyword's own, never looser: a broad keyword serves as exact when the search is the keyword word for word. Broad rows are where waste hides: tent rental near me came in through the broad keyword camping tent.
The report is Microsoft's search query report, and it names the exact dates it covers. Microsoft builds reports in the background, so a large one can take a moment; the assistant collects it when it is ready. You should see your own top terms, each next to the keyword that let it in.
Step 2: find the waste
10 search terms spent more than 15 dollars without a conversion, 318 dollars in all, about 17% of the account's spend:
| Theme | Example term | Terms | Spend |
|---|---|---|---|
| Rentals | tent rental near me | 4 | 112.60 |
| Free and DIY | diy tent stakes | 3 | 84.30 |
| Used gear | used hiking boots | 2 | 74.20 |
| Other | tent footprint | 1 | 46.90 |
Rentals and Free and DIY came from Generic tents, Used gear from Hiking boots. You sell new gear only, so those three themes are not your customers. Other is different: tent footprint is something you sell, but it matched the 4 person tents ad group, whose ads only talk about tents. Your Tent accessories ad group has the right ad for it.
Look for themes, not single terms, so one negative stops a whole family of searches. Check the volume too: a term with three clicks and no conversions is not evidence yet. You should end this step knowing which themes to block and where each one came from.
Step 3: add negatives at the right level
Added 6 negative keywords. Read back from Microsoft Advertising:
| Negative | Match type | Added to |
|---|---|---|
| rental, rentals, free, diy | Phrase | Generic tents (campaign) |
| used | Phrase | Hiking boots (campaign) |
| tent footprint | Exact | 4 person tents (ad group) |
The first five block those words anywhere in a search across the whole campaign. tent footprint is exact match on one ad group, so the same search can still reach Tent accessories.
A campaign negative is for a term that is wrong everywhere in that campaign; an ad group negative steers a search to the ad group that should have it. Microsoft Advertising negatives are exact or phrase match, never broad, and they apply as soon as they are added. If Microsoft refuses one, for example a negative that is already there, the reply names it and says why, and the rest are still added. You should see each negative read back with its match type and where it went.
Step 4: promote the terms that convert
Three search terms converted through a looser keyword:
| Search term | Keyword | Spend | Conversions |
|---|---|---|---|
| 2 person backpacking tent | 2 person tent (phrase) | 96.40 | 5 |
| waterproof hiking boots women | waterproof hiking boots (phrase) | 88.10 | 4 |
| ultralight tent 2 person | lightweight tent (broad) | 58.70 | 3 |
Want me to add them as exact match keywords in the ad groups they came from?
Added 2 person backpacking tent and ultralight tent 2 person to 2 person tents, and waterproof hiking boots women to Women's boots, all exact match. Read back from Microsoft Advertising: all three are stored.
Created pausedBoth ad groups are active, so each keyword would start serving the moment it is turned on. Turn all three on?
Done, one at a time. Read back from Microsoft Advertising: all three keywords are now Active and can spend.
An exact keyword for a proven search gives you its own bid and its own row in reports, instead of leaving it to a broad keyword's luck. You should see each keyword's status reported back as Microsoft stored it.
Step 5: what is different on Microsoft
Two more checks are worth a minute on Microsoft Advertising: where the ads showed, and what a Google import left behind.
141 of the 318 wasted dollars came from search partner sites, nearly all in Generic tents. Across the account:
| Network | Spend | Conversions | Per conversion |
|---|---|---|---|
| Bing, AOL and Yahoo | 1,438 | 58 | 24.79 |
| Search partners | 402 | 6 | 67.00 |
Partners cost almost three times as much per conversion.
Done for its three ad groups, one at a time: 2 person tents, 4 person tents and Tent accessories. Read back from Microsoft Advertising: each now shows only on Bing, AOL and Yahoo. They are serving, so the change takes effect right away.
The search terms report can split every term by network, so partner traffic shows on rows of its own. Search partners are an ad group setting, on by default. Next, if your campaigns started as an import from Google, compare the negatives:
I read the negatives on both. Missing on Microsoft:
- Generic tents: pattern and jobs, phrase match on Google.
- Hiking boots: free, diy, pattern, jobs and rental, phrase match, and cheap, broad match on Google.
Microsoft negatives cannot be broad, so cheap would go on as phrase match, which blocks any search containing it. Add all eight?
Added 8 phrase match negatives, 2 to Generic tents and 6 to Hiking boots. Read back from Microsoft Advertising: all 8 are stored.
A negative you add on Google later does not appear on the Microsoft copy by itself. You should see the gap listed per campaign, and anything that cannot carry over as it is, like a broad negative, called out before it is added.
Step 6: the write-up
Found. 318 dollars in 30 days on 10 search terms that never converted, about 17% of spend. Search partners at 67 dollars per conversion against 24.79 on Bing, AOL and Yahoo. Three converting searches with no keyword of their own. Eight Google negatives missing on Microsoft.
Changed. 6 negatives from the waste review and 8 copied from Google. Three exact keywords added and turned on. Generic tents taken off search partners.
Look again on October 25. New wasted terms in Generic tents, whether the three exact keywords now carry their searches, and Generic tents' cost per conversion without partners.
Two weeks is enough for new terms to show up. Next time, start from this summary and ask what changed since.
Why this is safe to run
The reports change nothing. Negative keywords only restrict where ads show, so they cannot spend. Keywords added through Adplane cannot serve until they are turned on, and turning anything on goes through a single update tool, microsoft_update_object, which waits for your yes when the assistant is the one suggesting it. Every change is read back from Microsoft Advertising before it is reported as done. The full safety model is in the docs.
Run it on your account
Adplane connects to Claude from the connector directory, or to ChatGPT from the plugin directory. Then connect Microsoft Ads on your accounts page with the Microsoft account you use at ads.microsoft.com. The free trial is 14 days and does not require a card. Connect your account and run step 1.
Try it in your own chat
Sign in with Google, connect your ad accounts, and ask your first question.