Your keywords are what you'd like to show up for. The search terms report is what you actually paid for. Those two lists drift apart over time, and that gap is where a lot of budget slips away without anyone noticing.
I go through search terms on every account I manage, every week. Once you get used to it, it takes about twenty minutes.
Why it matters more now
Keyword matching is much looser than it used to be. Phrase and exact match now pick up close variations and searches with a similar meaning. Broad match goes further. AI Max for Search and Performance Max stretch it further again.
That's not all bad. It finds searches you'd never have thought to add. But it also lets through plenty you'd never want to pay for, unless someone's keeping an eye on it.
Getting set up
Open the search terms report for the last seven days. If the account is small, use the last 14 or 30 so there's enough to look at. Sort by cost, highest first, and make sure you can see conversions and cost per conversion.
If you're running Performance Max, look at its search terms too. Google added that report for PMax, and it's worth checking.
The twenty minutes
I start with the most expensive twenty or so searches. For each one I ask myself a simple question: if I were sitting next to the person typing this, would I be happy paying for their click? Anything that's obviously not a buyer gets marked.
Then I add negatives for the searches that will never be right. Job hunters, people looking for free or DIY versions, someone trying to find a competitor's customer service number, products the client doesn't sell. I put each negative where it belongs. If it applies to the whole account, it goes in a shared list. If it only applies to one campaign, it goes there. I also use the tightest match type that does the job, because a broad negative on a word like "free" can block more than you'd expect. I read every negative twice before saving it. Blocking a word that also shows up in good searches is the easiest mistake to make here.
After that I flip it around and look for winners: searches that brought in conversions at a good price but aren't keywords yet. Those are worth adding as keywords of their own, so you can control the bid and write ads that match them.
Then there are the in-between ones. Relevant, but expensive and not converting yet. I don't block those straight away. I make a note and look again next week. If they've had a fair amount of spend and still haven't converted, then I decide.
Finally, I write a line or two in a log. The date, what I excluded, what I added, anything odd. It feels like admin, but three months later it's the only way to understand why the account behaves the way it does.
Mistakes to avoid
Don't add negatives because of one bad click. One expensive click that didn't convert doesn't tell you much. Wait for a pattern.
Be careful with brand terms. It's easy to block your own brand name in Search while you're trying to keep it out of Performance Max.
Don't try to deal with every tiny search either. The point is to protect the budget, not to review every line.
And try not to skip weeks. A search that wastes a little this week can waste a lot more by the end of the month.
Why I bother
Most accounts don't fall apart because of one big mistake. They drift, slowly, a few bad searches at a time. Twenty minutes a week is the cheapest way I know to stop that and keep the money going to people who actually want to buy.