Negative keyword match types: broad, phrase, exact

Negative match types behave differently from their positive equivalents, and that asymmetry is the single most common reason people add a negative and then watch the same junk traffic keep coming in.

The fix takes two minutes once you understand the rule. The rule does not match your intuition from working with positive keywords, which is why everybody trips on it once.

The three types, in one paragraph each

Negative broad blocks any query that contains the word, in any order, regardless of what surrounds it. Adding the negative broad keyword free blocks "free repair," "free quote near me," "is plumber service free," and any other query that contains the literal token "free." It does not block "freezing pipes" because the matcher only blocks whole-word tokens, and it does not block "complimentary" or "no cost" because, unlike positive broad, negative broad does not expand to synonyms or close variants.

Negative phrase blocks any query that contains the exact sequence of words, in that order. Adding the negative phrase keyword "free repair" blocks "free repair near me" and "phoenix free repair," but does not block "free quote and repair," because the words have to appear in the same order without other words between them. It is more permissive than negative exact and more restrictive than negative broad.

Negative exact blocks only queries that match the exact phrase, with no extra words before, after, or between. Adding the negative exact keyword [plumber salary] blocks the literal query "plumber salary" and nothing else. Not "plumber salary 2026," not "average plumber salary," just "plumber salary." Negative exact is rarely the right tool unless you have one specific recurring query that you want to surgically remove.

The asymmetry that catches everyone

Here is the trap. Positive broad match in Google Ads has, for years, expanded to include synonyms and close variants. Bid on plumber as a broad-match positive and the matcher will also serve against "plumbing service," "pipe repair guy," and "drain unclogger." That synonym expansion is intentional; it is the whole point of broad match.

Negative match types do not do this. Negative broad blocks only the literal token. It does not expand to synonyms. So when you add "cheap" as a negative broad, you do not block "inexpensive" or "low cost" or "budget." You have to add each one separately.

This asymmetry is not documented in giant red letters anywhere obvious. Most people learn it by adding a negative, watching junk continue, double-checking that the negative was applied, and concluding that the system is broken. The system is not broken; the rule is just different.

Which one to reach for, by situation

Most rows you flag in a sweep fall into one of four patterns, and each pattern has a default match type.

Pattern one: a single unambiguous word. Examples: free, salary, jobs, wholesale, schools. The word almost never appears in legitimate buyer queries for the business. Use negative broad. One block catches every variant.

Pattern two: a word that has multiple meanings. Examples: parts, training, used. "Parts" might be irrelevant for a service business but valid for a parts retailer. "Training" might mean job training (negative) or service training (positive). Use negative phrase to scope the block to the actual junk pattern. Block "parts only" instead of parts.

Pattern three: a recurring exact query. Examples: "plumber salary 2026," "law firm hours per week." A specific phrase that keeps appearing and you have already decided about. Negative exact is fine here, but negative phrase usually catches the same query with less typing. Use negative exact only if a near-identical phrase has legitimate value and you want to keep that one.

Pattern four: a category mismatch with overlapping language. Examples: residential business getting wholesale searches, commercial business getting consumer searches. Use negative phrase with the most specific differentiator. Block "wholesale supply" instead of wholesale when "wholesale" might appear in valid queries elsewhere.

A worked example: when "cheap" is broad and when it is phrase

Imagine a residential plumbing campaign. The word "cheap" appears in three search terms over the last thirty days: "cheap plumber phoenix," "cheap drain cleaning," and "cheap pipe replacement parts." All three are bargain hunters, none converted, all three are negative-worthy.

The instinct is to add cheap as a negative broad and be done. The risk: somewhere in the next thirty days, a real searcher types "cheap reliable plumber" and is genuinely a buyer who happens to also be price-sensitive. With cheap blocked broad, that buyer is silently filtered.

The conservative play is to phrase-block the specific patterns: "cheap plumber", "cheap drain", "cheap pipe". Three blocks instead of one, but each one only catches the actual junk and leaves the word alive in any phrasing you have not seen yet.

The aggressive play is to broad-block cheap and accept the small risk of blocking some price-sensitive buyers. For most residential service businesses, price-sensitive searchers are a worse fit for the service anyway, so the broad block is fine.

The choice is not "right or wrong." It is "which mistake are you more willing to make." Over-block and you lose some good traffic. Under-block and you keep paying for some bad. Pick the side you can live with for that account, then apply it consistently.

What about close variants on negatives?

There is a partial exception to the no-synonym rule. As of late 2024, Google began applying close-variant matching to a limited set of negative phrases, specifically for plurals and minor misspellings. Adding "plumber salary" as a phrase negative will also block "plumbers salary" and "plumber salary" typed without the space sometimes.

Do not rely on this. The close-variant behavior is patchy, undocumented in any detail, and changes from time to time. The safe assumption is still that you have to enumerate plural and singular variants yourself for important blocks.

Bulk-applying match types in a CSV

If you are uploading negatives via Google Ads Editor, the match type for each row is a column in the CSV. The values are Broad, Phrase, and Exact (Negative is implied by the file context). Most spreadsheet workflows have one column for the term and one for the match type, set the match type for each row, and bulk-import.

There is one quirk worth noting: when you type "free repair" (with quotes) into the Google Ads UI, the quotes are treated as a phrase-match indicator and stripped from the stored term. When you upload via CSV, the quotes should not be in the term column; the match type column carries the indicator. If you accidentally include the quotes in the term column, the upload will treat them as literal characters in the term, which means the negative will only match queries that literally contain quotation marks (i.e., approximately none).

The first three are broad blocks: cheap, free, salary. Each one is a single, unambiguous word. The fourth is a phrase block: "supply wholesale". The fifth is a real buyer; leave it alone.

When match types fail to apply the way you expected

Two failure modes worth knowing.

The first is the close-variant gap mentioned above. You added cheap and a "cheap" query still showed up. Look closely; usually the actual query was cheaply or cheaper. The literal token cheap does not appear, so the broad block does nothing. Add the variants as separate negatives.

The second is the wrong level. You added a negative at the ad group level when the campaign has multiple ad groups, and the junk query keeps coming through one of the other ad groups. Move the block to the campaign level.

A third edge case, less common: a search-themes-driven Performance Max campaign is bidding on the same queries and the negative does not apply because PMax has its own negative-handling. The fix there is in the PMax negatives piece.

What this looks like in a list, after a few sweeps

After three or four monthly sweeps on a typical small account, the negative list shape settles into a recognizable pattern. About sixty percent of the entries are single-word broad negatives caught early. About thirty percent are phrase blocks that scope ambiguous words. About ten percent are exact-match negatives for specific recurring queries.

If your list looks heavily skewed toward exact-match negatives, you are probably under-blocking. Each exact-match block only catches one query, while the budget-eating waste usually comes from a pattern of similar queries that a phrase block would catch in one entry. Re-read the exact-match list and consider promoting some of them to phrase.

If your list is heavily broad-match, you are probably over-blocking. Audit the broad blocks for over-loaded words and demote some of them to phrase blocks where the word has legitimate uses.

The balanced shape of the list is itself a sign that the sweeping cadence is working.