Google Ads broad match: 12 keywords alongside phrase match, zero spend in 10 days

Google's recommendation engine has one suggestion it makes more than any other: switch your keywords to broad match and pair them with smart bidding. The pitch is that broad match finds intent you'd never have written a keyword for, and the bidding algorithm sorts the good from the bad. It is presented as the modern way to run search.

Here is what happened in an account where broad match was enabled and available for the algorithm to use.

Twelve broad-match keywords were enabled across the highest-volume campaigns, solar, rebate, deal, deals, package, packages, bundle. Over ten days they recorded essentially nothing. Not a poor result. No result. The platform, given the choice between broad and phrase keywords in the same ad groups, served phrase.

Phrase already does what broad claims to

The argument for broad match is reach into queries you didn't anticipate. But that is what phrase match has been doing since 2021, when it absorbed broad match modifier. In the same ten days, 71% of phrase-match spend went to queries that didn't contain the keyword's words.

That's $30,736 of "discovery" traffic, found by phrase match, at a 45% premium. The reach broad match promises is already arriving through phrase. Adding broad on top doesn't expand the search space in any way this account could detect; it would only loosen the boundary further on a boundary that is already loose.

Exact does the job cheaper, at the same CPC

The counter-intuitive part of the first table is the CPC column, which wasn't shown. Exact match keywords in this account averaged $10.22 per click. Phrase averaged $11.43. Within 12%.

So exact match isn't cheaper per lead because it bids lower. It's cheaper because a much higher share of the clicks it buys come from people who typed the thing the advertiser meant. Conversion rate 5.8% against 2.7%. Same price per click, twice the leads per click.

Over the account's nine-year history the gap is narrower but consistent:

Exact produced 84% more conversions than phrase from 31% more money. There has not been a year in which that reversed.

What broad match is actually for

Broad match has a legitimate use, and it's narrower than the recommendation engine suggests: a new account or category with no keyword history, where you genuinely don't know what people search for. Run broad for a month, harvest the search terms, build exact and phrase keywords from what converted, and turn broad off. It's a discovery tool with a shelf life.

This account had nine years of search-term data and 507 campaigns' worth of keyword history. There was nothing left for broad match to discover that the paused campaigns didn't already contain.

The caveat

The ten-day window is short, and the twelve broad keywords were single generic words that phrase keywords in the same ad group would naturally outrank. A cleaner test would isolate broad in its own ad group with no phrase competition, run it for a month, and compare. That test would answer whether broad can work here. It wouldn't change the finding that, given the choice, the algorithm didn't reach for it.

And there are accounts where broad match with smart bidding does perform: usually large ones, with dense conversion data, where the algorithm has enough signal to make the loose matching pay. If yours is one, your data will say so. The recommendation card won't.

What to do

Look at your own search-term report by match type. If broad is a meaningful share of spend, check its CPL against exact on the same keywords. If broad is a small share, ask what it's contributing that phrase isn't.

Treat broad match as a diagnostic, not a setting. Turn it on when you need to learn something. Turn it off when you've learned it. And when the recommendation engine suggests moving your exact keywords to broad (which it will, repeatedly), look at your exact-match CPL first. In this account that number was $177. Nothing the recommendation offered was going to improve on it.