For a decade, "never use broad match" was correct advice. Broad match without Smart Bidding was a budget incinerator — it matched on vague relevance signals, Google had every incentive to spend your money regardless of intent, and anyone who tried it got burned once and told everyone they knew. That story hardened into dogma. Dogma is exactly the thing you should be suspicious of in an industry that changes every eighteen months.

Broad match today is a different product wearing the same name. Paired with Smart Bidding — specifically Target CPA or Target ROAS with a mature conversion signal — broad match is matching against your actual conversion data, not just keyword-to-query relevance. In the majority of head-to-head tests we run against phrase and exact on accounts with sufficient conversion volume, broad match plus Smart Bidding wins on volume and holds efficiency within a reasonable band. That is not a hot take anymore. It is what the data says when you actually run the test instead of repeating what you read in 2019.

What Actually Changed

Three things had to happen before broad match became testable again, and most accounts are only just catching up to all three.

Smart Bidding matured. Target CPA and Target ROAS now have years of machine learning refinement behind them, and they use signals phrase and exact match simply can't access — device, time of day, location, audience membership, and dozens of others layered into the auction-time bid. Broad match gives that algorithm more query-level surface area to find conversions on. Phrase and exact match restrict the algorithm's raw material.

Conversion data got cleaner. Enhanced conversions, server-side tagging, and consent-mode-aware modeling closed a lot of the measurement gaps that used to make Smart Bidding fly blind. A bidding algorithm making decisions on incomplete conversion data will make bad decisions no matter what match type feeds it. Fix measurement first — broad match on top of bad measurement is still a bad idea.

Negative keyword lists got smarter. Shared negative lists, applied at the campaign or account level, now do a lot of the filtering work that match-type restriction used to do manually. You're not relying on the algorithm alone to keep queries relevant — you're actively fencing off the categories you know don't convert.

When Broad Match Is Ready for Your Account

Not every account is ready for this test on day one.

The prerequisites matter more than the tactic.

1. **30+ conversions per campaign in the last 30 days.** Below that, Smart Bidding doesn't have enough signal to make broad match's flexibility pay off. You'll just buy noise.
2. **Enhanced conversions or offline conversion import configured.** Garbage measurement in, garbage bidding out — regardless of match type.
3. **A negative keyword list with actual history behind it**, not a fresh account with zero exclusions.
4. **Budget headroom to absorb a two-to-three week learning period** without panicking at day-four CPA swings.

If your account is under 30 conversions a month, skip this entire conversation and go fix volume first.

How to Run the Test Without Betting the Account

Don't flip your whole account to broad match on a Tuesday and hope. Structure it like the experiment it is.

Duplicate your best-performing exact match campaign into a new campaign using broad match keywords, same budget, same Target CPA or Target ROAS. Use Google Ads' built-in Campaign Experiments (formerly Drafts and Experiments) to run it as a true A/B split with statistical significance reporting, not a gut-feel comparison between two campaigns running in different weeks with different seasonality.

Let it run a minimum of two full weeks post-learning-period. Smart Bidding needs time to recalibrate against the new query surface, and judging it on day five is judging it before it's actually working.

Broad match isn't a shortcut. It's a bet that your measurement is good enough to let the algorithm do the targeting.

Where It Still Falls Apart

Broad match still fails in predictable places, and pretending otherwise is how you end up back in 2019's horror stories.

• **Low-volume accounts.** No signal, no smart matching — just expensive random walk.
• **Brand-sensitive categories** where one wrong query match creates a real reputational or legal problem, not just a wasted click.
• **Accounts without a real negative list.** Broad match delegates trust to the algorithm; you still have to hold up your end with exclusions.
• **New accounts under 90 days old.** Give the pixel and the bidding algorithm time to learn your business before handing them more rope.

The Actual Recommendation

Test it as a parallel experiment, not a full account migration. Keep phrase and exact running while broad match proves itself. If it wins on your numbers, expand it gradually — campaign by campaign, not overnight. If it doesn't win, you've lost two weeks of a controlled experiment, not your quarter's budget. That's the whole point of testing it this way instead of trusting a rule of thumb that was written for an auction that no longer exists.