What is Broad match?

Broad match makes a keyword eligible for searches the platform judges related to it. Not searches containing it, not searches meaning the same thing — searches related, by the platform's own interpretation.

That interpretation is generous and it draws on more than the query. On Google Ads the signals reported to influence broad matching include the landing page content, the other keywords in the ad group, and the person's recent search activity. A broad keyword is therefore matched in a context, not in isolation, and the same keyword in two different ad groups can behave differently.

The result is the widest reach available in search advertising, and the setting with the largest gap between what an advertiser expects and what they get.

For years the standard advice was to avoid it. That advice was correct under the conditions it was formed in and is frequently wrong now, and the reason is bidding rather than matching.

Under a manual bid, a broad keyword is genuinely dangerous. One number applies to every search the keyword reaches. The advertiser cannot bid less for a poor match than for a good one, because they do not know which is which until after the click. The only available filter is a negative added once the money is spent, which is always retrospective.

Under a conversion-based automated strategy, the bidder evaluates each auction separately. A search it does not expect to convert gets a low bid or no bid. The filtering that the advertiser could not do in advance happens per auction, continuously, on signals the advertiser cannot see.

That is the whole of the change. Broad match did not become safer; the thing sitting next to it became able to decline.

Finding demand nobody named. People search in ways an advertiser cannot enumerate — different words, different framings, questions rather than nouns, problems rather than products. A broad keyword plus a bidder that can decline is a discovery mechanism, and the search-term report turns what it finds into exact-match keywords for later.

Covering a long tail. Most search volume in most categories sits in queries too infrequent to be worth adding individually. Exact and phrase coverage of a long tail is an endless clerical task; broad covers it structurally.

Feeding a model. Wider matching produces more conversions in one place, which is what an automated strategy needs to predict well. An account fragmented into many tightly matched groups often cannot accumulate enough events anywhere.

No automated bidding. Covered above, and it is the single largest factor.

No negatives. Broad match without a maintained negative list will find every adjacent query the platform considers related, including the free, the informational, the job-seeking and the competitive.

A shallow conversion. The bidder declines what it does not expect to produce the nominated event. If that event is easy to produce and unrelated to revenue, the filtering works perfectly toward the wrong end.

Thin conversion volume. With too few conversions, the bidder cannot predict, so the filtering is weak precisely where the reach is widest. This combination — broad match on an account with almost no conversion data — is the worst configuration available.

Loose ad groups. Because context influences matching, a keyword sitting among unrelated terms reaches further and less predictably than the same keyword in a tight group.

Pair it with a conversion-based strategy, always. If the account cannot support one, it cannot support broad match either.

Review search terms on a schedule and add negatives from what you find. This is the maintenance the setting requires and the work that makes it affordable.

Keep ad groups tight, so the context shaping the matching is the context you intended.

Harvest what works. Search terms that convert well are candidates for exact-match keywords with their own ads, which is how discovery becomes structure.

Check the aggregate, not only the report. Platforms do not surface every search term, so a share of spend goes to queries nobody can see. Watching the campaign's overall cost per conversion catches drift that the visible portion of the report does not explain.

Broad match is a tool for an account that has conversion data, a working conversion definition, an automated bid strategy and somebody maintaining negatives.

An account with all four will usually find it produces volume that tighter matching leaves on the table, at costs the strategy keeps in line.

An account missing any of the four should not be using it, and the advice to avoid it — which sounds outdated — remains exactly right for them. The disagreement in the industry about this setting is mostly two groups correctly describing their own circumstances.

The change is large enough to be worth staging rather than switching wholesale.

Start in one campaign, ideally one with healthy conversion volume and a target that already works. A campaign chosen because it is struggling is the worst place to test, because it has the least data for the bidder to filter with.

Bring the negative list first. Whatever negatives the account already has should be applied before the reach widens, not after the search-term report fills up.

Expect a noisy fortnight. The match change and, usually, the ad-group context change put the campaign into a period where delivery and cost are unrepresentative. Reading that fortnight as the verdict is the standard way this experiment fails.

Budget for discovery. Some spend on searches that do not convert is how the model learns which ones do not. That cost is real and it is front-loaded, which is another reason to stage rather than to convert the account at once.

Compare on the campaign, not on the keyword. Keyword-level figures under broad matching describe a shifting set of searches, so the campaign's cost per conversion and volume are the numbers that mean something.

Three changes are normal and all three alarm people who were not expecting them.

The keyword count stops describing the account. A handful of broad keywords can reach more than hundreds of exact ones, so the account looks smaller while covering more.

The search-term report grows and diversifies, including terms that seem unrelated until their conversion figures are read.

And impression volume rises faster than spend, because the bidder is entering many more auctions and declining most of them. A large increase in eligible impressions with modest spend growth is the filtering working, not a delivery problem.

Both are looser than exact match and they loosen in different directions. Phrase match requires the meaning of the keyword to be contained in the search: the search has to be about the thing the keyword names, with extra words allowed around it. Broad match requires only that the platform judge the search related, which can include searches sharing no words with the keyword at all. The practical difference is the size of the set each can reach and therefore how much the account depends on other controls. Phrase is bounded by a containment test, so the surprises in the search-term report tend to be adjacent rather than alien. Broad is bounded only by the platform's judgement of relatedness, so it needs a conversion-based bid strategy to decline the poor matches and a maintained negative list to refuse the specific ones — with a manual bid and no negatives it is the most expensive setting in search advertising. Neither is better in the abstract: the choice is about how much of the filtering the advertiser is doing in advance and how much the bidder is doing per auction.