A keyword is what you bid on. A search term is what someone actually typed. Google matches one to the other loosely, and the gap between them is where the money goes. Finding it takes a single filter: everything that spent more than your threshold in the last 30 days and converted zero times.
Which search terms are wasting my money?
Start with one rule. A term that spent more than you would happily pay for a customer, over a 30-day window, and produced no conversions. That is not a hunch, it is a filter you can run.
Here is what it looks like on a small account over 14 days.
| Search term | Spent | Conversions |
|---|---|---|
| free running shoes | $41 | 0 |
| shoe repair near me | $28 | 0 |
| nike outlet store | $19 | 0 |
Eighty-eight dollars, three queries, nothing to show for it. Nobody chose those as keywords — they are what loose match types brought in. In a mature account, more than 15 to 20 percent of spend sitting on non-converting queries is worth a serious look.
The free Google Ads waste audit runs this filter for you and returns the queries, the cost behind each one, and the negative keywords it proposes.
How do I find them?
1. Pull the search terms report for the last 30 days. In Google Ads it sits under Insights and reports, then Search terms. Fourteen days is enough on a high-volume account; thirty is safer when conversions are sparse.
2. Filter to spend above your threshold with zero conversions. The threshold should track your own cost per acquisition. Twenty dollars is a reasonable default, but it is generous at an $8 CPA and meaningless at $300. Set it to a number you would genuinely regret spending on a customer who never arrived.
3. Group by intent before you block anything. Free, DIY, job-seeker and competitor-name queries are usually easy calls. Everything else needs a second look, which is the next section.
The prompt
If your assistant has access to the account, the whole thing is one message. This is the check we run weekly.
Review my Google Ads search terms from the last 30 days. Find terms that spent more than $20 with zero conversions. Group them by intent and explain which are clearly irrelevant versus potentially valuable but unproven. Recommend the exact negative keywords and match types I should add, and whether each belongs at campaign or ad group level. Be careful not to block converting or high-intent variants. Do not apply anything. If there is nothing worth acting on, tell me that instead.
Change the twenty to your own number before you run it.
What you should not block
Not every expensive non-converting query is waste.
- Research queries on a considered purchase. Someone comparing options will cost you clicks for weeks and then convert on a branded search. Blocking them buys a cheaper CPA and a smaller business.
- Terms with too little data. Three clicks and no conversion is noise, not evidence.
- Near-misses on high-intent variants. A negative added at the wrong match type takes converting traffic with it. That is why the prompt asks for match types and levels instead of a bare list.
The judgement stays yours. Choosing the match type and the level is where most of that judgement gets spent, and the full 30-minute audit covers the other six places money leaks. In a local service business the list is predictable enough to pre-empt: the cleaning-company version names the four categories that waste money in that trade. In a repair shop the same list is mostly a parts catalogue, which the auto-repair version separates from the labour you actually sell. In a landscaping business the list is a garden centre, and the bigger question is which season and which crew the budget should be following. In a moving company it is a rental truck, and the report has to be read twice — once for where the searcher was, once for where they were going.
Make it a weekly check
Waste is not a one-off. Loose match types keep serving new queries, so the list refills.
For a scheduled review in ChatGPT or Claude, first verify that the scheduled run can access LoomaScale and return fresh account data. Follow the recurring-review guidance; otherwise, use a calendar reminder and run the saved prompt manually. Account setup: ChatGPT, Claude.
Then the waste gets caught on Sunday instead of next Wednesday. The weekly review folds this check in with five others and still fits in ten minutes.
And the same report has a profitable half: the queries that converted without being keywords at all. The keyword research walkthrough is the prompt that promotes them.
