Broad match deserves a second look, with the right safeguards in place
Broad match has a reputation it mostly earned in 2018. The match type has changed significantly, and dismissing it entirely is leaving incremental volume on the table. Here is how to give it a fair trial without blowing your budget on irrelevant queries.
Ask any performance marketer what they think of broad match and the answer has been consistent for years: do not trust it, review the search terms weekly, and switch everything to phrase unless you want Google to spend your budget on tangentially related queries for six weeks before you notice.
That reputation was earned. Broad match in the pre-Smart Bidding era was genuinely unreliable: it would trigger for queries that shared a word with your keyword and call it a match. The experience of watching it drain budget while your phrase-match campaigns sat on sensible CPAs has left a lot of practitioners with a permanent wariness.
But the match type has changed. Understanding what changed, and what safeguards make it viable now, is worth the effort.
What is different in 2026
Broad match now integrates with Smart Bidding in a way that changes its behaviour substantially. Rather than matching based purely on query tokens, it uses the bidding model's conversion signal to assess whether a given query is worth entering an auction for, even if the query looks loosely related. In practice this means that a well-fed Smart Bidding strategy acts as a filter on broad match reach.
A broad match keyword paired with Max Conversions or a tROAS target will, over time, weight towards the query patterns that convert within your target. It is not a perfect filter. But it is a materially better one than the raw lexical matching of five years ago.
Broad match works best when your campaign has strong conversion signal: at minimum thirty conversions per month, ideally seventy or more. Below that threshold, the bidding model does not have enough data to constrain broad match reach meaningfully, and you are back to the old problem.
The case for testing it
The argument for broad match is not that it is better than phrase. In most accounts with mature keyword lists it is not. The argument is that it finds queries that you did not know to write as exact or phrase keywords, particularly in product or category areas where user language is shifting.
Voice search, SGE-influenced reformulation and the general lengthening of queries mean there is real volume in query patterns that phrase match will not touch. If your competitor is using broad match with Smart Bidding and capturing that demand, your exact and phrase coverage has a structural gap.
Accounts using broad match with Smart Bidding saw an average of nineteen percent more conversions at the same CPA compared with phrase match alone, in matched control experiments.
Google Ads internal efficacy study, 2025
I would treat that figure with appropriate scepticism given the source, but the direction is credible, and the independent search-term analyses I have run in client accounts are broadly consistent with it.
A safe testing framework
The safeguard that makes broad match testable is negative keyword hygiene. Before adding any broad match keywords, your account needs a clean negative list. Not "we have negatives", but a systematically built list based on actual search term reports from the last ninety days, covering brand, competitor and irrelevant-intent terms.
Run the broad match test in a separate campaign or ad group with a dedicated budget, not by switching existing phrase keywords. This keeps the experiment clean and avoids contaminating your existing performance data.
| Setup element | Recommendation | Why it matters |
|---|---|---|
| Bidding strategy | tROAS or Max Conversions with target | Constrains reach via conversion signal |
| Minimum monthly conversions | 30+ (70+ preferred) | Below this, bidding model under-constrained |
| Negative keyword list | Systematic, 90-day STR review | Prevents brand, competitor and irrelevant spend |
| Campaign structure | Separate campaign, dedicated budget | Keeps test clean; easy to pause |
| Test duration | 4-6 weeks minimum | Smart Bidding needs time to learn the signal |
| Primary evaluation metric | Incremental conversions, not ROAS | ROAS can look fine while volume cannibalises phrase |
Reading the search terms report differently
When you run broad match, the search terms report becomes your primary diagnostic tool. You are not looking for proof that broad match is matching sensibly. You are looking for the queries it found that you did not have in your keyword list and that converted well. These are the finds.
Sort by conversions descending. Anything with three or more conversions and a CPA within your target is a candidate for promotion to exact or phrase match in your main campaign. This is the workflow that makes broad match a discovery mechanism rather than a liability.
When broad match is not worth testing
Three situations where the expected value is low: accounts with thin conversion data (under thirty conversions per month), accounts where brand terms are not cleanly negative-matched across all campaigns, and any account running manual or enhanced CPC bidding. In these cases the risk of wasted spend significantly outweighs the discovery upside.
The nuanced conclusion
Broad match is a tool, not a strategy. The practitioners still dismissing it wholesale are carrying a view formed in 2018. The practitioners treating it as a default because Google recommends it are confusing platform guidance with independent evidence. The right approach is a structured test with clean safeguards, honest evaluation against incremental conversions and a workflow for recycling the useful finds back into your core keyword structure.
That is more work than either extreme. It is also more likely to produce a result you can act on.