Broad match with Smart Bidding works when three things hold: the campaign gets enough conversions for the bidder to learn from (as a rule of thumb, 30 or more a month), conversion tracking records the actions the business values, and, for Target ROAS, the values on those conversions are right. With that in place, broad match reaches queries your phrase and exact keywords never listed, and Smart Bidding bids them down or out when they do not convert. Without it, broad match matches widely and the bidder has nothing reliable to steer by, so spend goes to loosely related searches.

The safe way in is a custom experiment that adds broad match to a copy of the campaign, with negatives and brand exclusions in place, judged on CPA or ROAS over several weeks. Keep phrase and exact where the campaign is small, the tracking is weak, or the query space is narrow and already covered.

How broad match works with Smart Bidding

Broad match today reads more than the keyword. It uses the meaning of the query, the landing page, the other keywords in the ad group, and the searcher's recent searches to decide whether a query matches. A broad keyword 'crm software' can match 'sales pipeline tool for small teams' without sharing a word with it.

Smart Bidding then decides how much each of those matches is worth, from the predicted conversion rate and value in that auction. The pairing is the point: broad match finds queries, Smart Bidding prices them. On manual CPC or Maximize clicks, broad match has no conversion signal to price with, which is why Google recommends it only with Smart Bidding.

AI Max for Search campaigns, a campaign-level setting Google opened in beta in May 2025 and made generally available in April 2026, extends the same idea: search term matching (broad match plus keywordless matching), text customization and final URL expansion. From September 2026 Google is moving Search campaigns that used the campaign-level broad match setting into AI Max with search term matching on, so check which of your campaigns it now covers. Everything in this guide about prerequisites and guardrails applies to it with more force, because it widens reach further.

Prerequisites before you switch

  • Conversion volume: roughly 30 or more conversions a month in the campaign, or in the portfolio it bids with. Below that, the bidder learns slowly and broad match spends while it learns.
  • Tracking you trust: the primary conversion actions are the ones you want more of. Page views, scroll depth or newsletter sign-ups set as primary teach the bidder to buy the wrong traffic. The conversion tracking checklist covers this.
  • Enhanced conversions and Consent Mode v2 where you have European traffic, so the bidder sees conversions that cookies alone would miss.
  • Values, for Target ROAS: the values must reflect what a conversion is worth. A lead valued at a flat EUR 100 whether it becomes a EUR 500 deal or nothing teaches the bidder nothing about which queries bring the deals.
  • A target that reflects today: a Target CPA or Target ROAS close to the campaign's realized figure, so adding broad does not coincide with a squeeze or a loosening.

If one of these is missing, fix it first. Broad match on a campaign with 12 conversions a month and a thank-you page that fires twice will find a lot of new queries and cannot tell which ones are worth anything.

Guardrails: search terms, negatives and brand exclusions

Search term review

Broad match makes the search terms report the main control you have. For the first four weeks, review it weekly, sorted by cost. Look for patterns: job seekers ('crm jobs'), free intent ('free crm'), competitors you do not want to pay for, informational queries ('what is a crm'). Google hides some low-volume terms, so the report shows most of the spend, not all of it. The search terms guide goes through the review in detail.

Negative keywords

Add negatives at the level where the pattern applies. Themes that are never relevant, such as 'jobs', 'free' or 'course', belong in an account-level negative list or a shared negative keyword list. Phrase and exact negatives on the specific terms that wasted money go on the campaign. Negative keywords do not expand to close variants the way positive keywords do, so add plurals and common misspellings yourself.

Brand exclusions

A non-brand campaign on broad match will match your own brand queries, which are cheap and convert well, and report them as non-brand performance. Brand exclusions, set in the campaign settings from a brand list, stop that. They also keep a broad campaign off competitor brands you have decided not to bid on. Negative keywords do the same job for the exact spellings you list; brand lists also cover misspellings and variants Google associates with the brand.

Test broad match in an experiment

Switching a working phrase and exact campaign to broad is a change to a serving campaign that decides where the whole budget goes. Test it first.

  1. Under Campaigns > Experiments, create a custom experiment on the campaign.
  2. In the treatment arm, add broad match versions of the main keywords, or change the existing keywords to broad. Leave the bid strategy, target and budget alone, so the match type is the only difference.
  3. Apply the brand exclusions and account-level negatives to both arms before the start date.
  4. Split 50/50 on a search-based split.
  5. Run past the learning period and at least two conversion cycles, usually four to six weeks. See how long a Google Ads experiment should run.
  6. Compare CPA or ROAS, total conversions or value, and the share of spend on queries you would have excluded. Broad often wins on volume and loses a little on efficiency; decide in advance which trade you accept.

Watch one trap when reading it: the broad arm will show conversions on queries the control arm never entered, and some of them are brand or near-brand queries the exclusions missed. Check the treatment arm's search terms before crediting the gain to broad match. How to read Google Ads experiment results covers confidence intervals and when to call a test.

When to keep phrase and exact

  • Low volume: a campaign under about 30 conversions a month gives the bidder too little to filter broad queries with.
  • Weak or noisy tracking: conversions that fire on the wrong action, or values that do not track real revenue.
  • Narrow query space: a niche B2B product with a few dozen relevant searches, all already covered by exact match.
  • Brand campaigns: broad on a brand campaign reaches generic queries and blurs the brand numbers.
  • Tight regulatory or compliance limits on what queries the ads may appear on, where any unreviewed match is a risk.
  • A budget that cannot absorb the learning weeks, when broad match will spend on queries that do not yet convert.

Many accounts run both: exact and phrase for the known high-intent terms, with the tighter target, and a broad campaign or ad group with its own budget to find new queries. Terms that prove themselves in the broad campaign graduate to exact.

Where GoodLads fits

GoodLads reads each campaign's search terms against its keywords and conversions, and proposes negative keywords where terms spend without converting. Negatives are one of the few live-only levers, since a negative has to block a term to save anything, so the button is orange and asks for a second confirm. It does not switch campaigns to broad match or run a match-type experiment for you. The read-only MCP endpoint lets you point Claude or ChatGPT at your own search terms to work through a broad match review in conversation.

Keyword table showing spend split by match type, then each keyword with its match type, spend, clicks, CTR, conversions, value, ROAS and Quality Score.
Spend by match type and the keywords carrying it. Broad match takes 27% of this spend at the worst ROAS of the three, which is the gap two of the demo ideas are argued from. From the demo account

Questions people ask

Should I use broad match with Smart Bidding?

Yes, if the campaign has about 30 or more conversions a month, tracking you trust and, for Target ROAS, reliable values. Otherwise keep phrase and exact until those are in place.

Does broad match work with Manual CPC?

It runs, but without a conversion signal to price each query, so it tends to spend on loosely related searches. Google recommends broad match only with Smart Bidding.

How do I stop broad match from showing on my brand terms?

Apply a brand exclusion from a brand list in the campaign settings. Add your brand as a negative keyword as well if you want exact control over specific spellings.

How long should a broad match test run?

Past the learning period and at least two conversion cycles, which for most accounts is four to six weeks in a custom experiment split 50/50.