The Google Ads search terms report: finding wasted spend

What the search terms report shows (and hides)

Your keywords are what you bid on; search terms are what people actually typed. With broad match and close variants, the gap between the two can be wide. The report lists each term with impressions, clicks, cost, conversions and conversion value, plus the keyword and match type that triggered it.

It doesn't show everything. Since 2020 Google only lists terms that enough people searched; the rest are summarized without the query text. Depending on the account, a meaningful share of spend can sit in terms you'll never see. Performance Max and AI-driven matching report queries differently from classic Search campaigns, and the detail available has been changing, so check what your account shows for each campaign type.

Setting thresholds for “wasted”

“Zero conversions” alone is not enough: a term with two clicks and no conversion tells you nothing. Use a threshold tied to your own numbers:

The clicks rule comes from simple expectation: at a 4% conversion rate you'd expect one conversion per 25 clicks, so 50 clicks with none is unlikely to be bad luck. It's a heuristic, not a statistical test, so use judgment for terms near the line.

Watch for conversion lag. If your sales cycle is long, recent clicks may still convert. Leave the last few days out of the window, or check conversions by conversion time.

Use n-grams to find the pattern, not just the term

Most wasted spend hides in the long tail: hundreds of terms that each spent a little. Individually none crosses your threshold; together they're a budget line. N-gram analysis fixes that by splitting every term into one-, two- and three-word phrases and adding up their cost and conversions across all terms.

If “jobs” appears in 40 different search terms that together spent $600 with no conversions, one phrase-match negative for jobs does more than 40 exact negatives. MegaMCP's ads_waste_sweep does this aggregation and only lists an n-gram as wasted when every term containing it together stayed at or under your conversion limit and it appears in at least two terms.

Typical junk themes: jobs and careers, free, DIY and how-to, courses and training, wrong locations, used or second-hand, and competitor names you've decided not to target. Typical traps: words that look odd but convert, such as “cheap” for a budget brand or a misspelling of your own name.

Choosing the negative match type and level

Negative keywords behave differently from positive ones. They do not match close variants, plurals or misspellings, so add the variants you need explicitly.

Then pick the level. Ad group negatives steer traffic between ad groups. Campaign negatives keep a campaign out of a search entirely. Shared negative keyword lists apply one list to many campaigns, which suits account-wide junk such as jobs or free. Microsoft Ads works similarly but supports exact and phrase negatives only.

A weekly wasted-spend workflow

1. Pull the last 30 days of search terms.

   Exclude the last few days if you have conversion lag. Sort by cost.

2. Apply your threshold.

   Spend over your target CPA, or clicks over 2 ÷ conversion rate, with zero conversions.

3. Run an n-gram pass.

   Find words and phrases that appear across many zero-conversion terms.

4. Decide, term by term or theme by theme.

   Keep anything that is on-topic but unlucky; negate what is clearly irrelevant.

5. Add negatives at the right match type and level.

   Exact for specific queries, phrase or broad for themes, shared lists for account-wide junk.

6. Check next week.

   Confirm the spend moved to better terms and nothing good got blocked.

This is the canonical job to give an AI with a human approval step: the analysis is mechanical, the decision is judgment. MegaMCP's Search Term Sweeper workflow runs weekly, flags terms over your spend threshold with zero conversions, and puts every proposed negative into one plan for review. Nothing is added until someone approves it, and the negatives can be removed again with undo.

Example prompts

Copy these into Claude, ChatGPT or Cursor with MegaMCP connected. Each lists the tools it typically calls; more are in the prompt library.

• Reads data

  Show search terms from the last 30 days with more than $60 spend and zero conversions, highest spend first.

  Tools it calls: google_ads_search_terms

• Reads data

  Run a wasted spend sweep with n-grams and group the candidates into themes.

  Tools it calls: ads_waste_sweep

• Proposes a change

  Add `jobs` and `free` as phrase negatives to every Search campaign where they wasted money. Show me the plan.

  Tools it calls: google_ads_add_negative_keywords

  Any change becomes a plan you review and approve. Nothing moves until you do.

• Proposes a change

  Create a shared negative list called Account junk with these terms and attach it to all Search campaigns.

  Tools it calls: google_ads_list_negative_listsgoogle_ads_negative_lists

  Any change becomes a plan you review and approve. Nothing moves until you do.

• Proposes a change

  Do the same wasted-spend check in Microsoft Ads for the last 30 days.

  Tools it calls: microsoft_ads_search_termsmicrosoft_ads_add_negative_keywords

  Any change becomes a plan you review and approve. Nothing moves until you do.

• Reads data

  Which converting search terms contain the word cheap? I want to make sure I don't block them.

  Tools it calls: google_ads_search_terms