Fix: Broad match running without continuous negative-keyword maintenance

Broad match doesn't self-clean. Even when it is correctly paired with Smart Bidding, the query universe expands faster than the negative-keyword list — so an account that hasn't added negatives recently is, by definition, leaking spend on queries the bid strategy has already learned to ignore but is still paying to serve [1][2].

Why this matters

This rule differs from the sibling check broad match without Smart Bidding — that finding asks "is the bidding signal capable of guiding broad match?", while this one asks "is the human operator still cleaning up after broad match's query drift?". A campaign can pass the bidding check and still fail this one.

Google's broad-match documentation positions broad as the default match type and recommends pairing it with Smart Bidding to manage the wider query surface [2]. What the documentation does not promise is that Smart Bidding will exclude every off-intent query on its own. Smart Bidding bids lower on low-converting queries, but lower is not zero — and at scale, "lower bid on garbage" still equals real money spent on impressions and clicks that will never convert.

The Search Terms Report is the canonical surface for catching this drift: Google's own example walks an advertiser through discovering that a broad match on flowers triggered an ad on wine glasses, and adds it as a negative [3]. The implied operating model is that the advertiser does this regularly — not once at launch. Practitioner consensus reinforces a recurring, intentional review tied to performance triggers and time windows, not a set-and-forget list [4].

Close-variant matching, which now applies to phrase and exact match too, compounds the problem [5]. Synonyms, reordered words, and "same intent" variations route adjacent-intent traffic to a broad-match keyword that may already have weak negatives — and every new variant Google decides to honor is a fresh chance for off-intent spend that your existing negative list doesn't cover.

The cost profile is asymmetric: each individual off-intent click is cheap, but the cumulative monthly leak on a mid-sized broad-match campaign typically runs 8-20% of spend before negatives are tightened (practitioner benchmark, not Google-published). Worse, the off-intent clicks that do sometimes convert at micro-rates pollute the Smart Bidding training data, dragging optimal-bid estimates toward the wrong queries.

How to verify the issue

1. Open Campaigns → select a Search campaign that uses broad match → Audiences, keywords, and content → Negative keywords. Note the count and the timestamp of the most recent addition.
2. Pull the change history: Tools → Change history, filter by Negative keywords, last 30 days. Fewer than 5 negative-keyword additions across all broad-match campaigns in the last 30 days = maintenance gap confirmed.
3. Check the shared library: Tools → Shared library → Negative keyword lists. Absent or unattached to active Search campaigns = no account-wide guardrail.
4. Pull Insights and reports → Search terms for the last 30 days. Filter to keywords with Broad match type. Sort by impressions descending. Read the top 100 — flag every off-intent query (informational, job-seeker, free-tier, competitor research, adjacent vertical).
5. Calculate the off-intent impression share: off_intent_impressions / total_broad_match_impressions. Above 10% on a campaign that has been live >30 days with no recent negatives = active leak.
6. Optional n-gram sweep: export the search-terms CSV, tokenize each query into 1-grams and 2-grams, aggregate cost and conversions by token. Tokens with high cost and near-zero conversions are negative-keyword candidates [4].

How to fix it

Total time: 90 minutes for the initial sweep, then a 30-minute weekly cadence.

1. Export 30-90 days of search terms for all broad-match campaigns. From Insights and reports → Search terms, download CSV. Filter to impressions ≥ 5 to focus on volume drivers — long-tail one-impression queries are usually noise, not patterns.

2. Run an n-gram sweep. Tokenize queries into 1-grams (single words) and 2-grams (bigrams). For each token, compute total cost, total clicks, total conversions, conv rate, CPA. Sort by cost descending among tokens with zero conversions [4]:

   token         cost   clicks  conv   notes
   "free"        $412     287     0    candidate negative (broad)
   "jobs"        $186     94      0    candidate negative (broad)
   "tutorial"    $98      71      0    candidate negative (broad)
   "salary"      $54      28      0    candidate negative (broad)
   "wholesale"   $312     78      1    review — conv exists, don't auto-negate
   

3. Build the negative list with correct match types. Single-token universal exclusions (free, jobs, careers, salary) as broad-match negatives. Multi-word patterns (free trial download, open source alternative) as phrase. High-volume specific terms as exact.

4. Apply at the right scope:

   • Shared list (MCC) — universal off-intent guardrails: jobs/careers/salary/free/torrent + competitor brand tokens (if you don't bid on them). Attach this list to every active Search campaign.
   • Campaign-level negatives — cluster-specific drift surfaced only in one campaign.
   • Ad-group negatives — cross-pollination between tightly themed ad-groups inside the same campaign.

5. Set a weekly cadence (practitioner benchmark, not Google-published): 30 minutes every Monday, pull the prior 7 days of search terms, add 5-15 new negatives, document them in a changelog. Tie additions to triggers (cost > X with zero conversions, or conv rate < Y of campaign average) rather than vibes [4].

6. Quarterly audit of the shared list itself. Negatives that block converting traffic are a hidden tax — review the list every 90 days and remove anything that has matched a known-good query in the last quarter. Cross-check with Search Query Mining for the deeper workflow.

7. Don't trust auto-applied recommendations blindly. Google's "Add negative keywords" recommendation surfaces candidates from your search terms — useful as a queue, but apply manually after reviewing each candidate against business context. A keyword that looks off-intent to an algorithm may be your most valuable query.

How to confirm the fix worked

Diagnostic checklist — re-measure 14 days after deploy

• Recent negatives added ≥ 5 in the trailing 30-day change history per active broad-match campaign.
• Shared negative list present at MCC level and attached to every Search campaign with broad-match keywords.
• Off-intent impression share in the post-deploy search-terms report < 10%.
• Wasted-cost share (cost on zero-conversion broad-match queries / total broad-match cost) trending down week-over-week.
• Weekly cadence documented — a written process, a calendar reminder, or an assigned owner. Without process, the leak returns within 60-90 days.
• Conv rate stable or rising; CPA flat or down within 14-21 days as Smart Bidding re-stabilizes on cleaner signal.

Re-run the audit — broad_match_without_negative_maintenance moves from failed → passed once the trailing-30-day negative-additions count crosses the threshold AND a shared list is attached. Both conditions must hold; either alone is insufficient.

Why this rule exists separately from the bidding-pairing check. Whitead's audit treats broad match as a two-part contract: signal quality (the bidding side, covered by broad match without Smart Bidding) and ops hygiene (the negatives side, this rule). The two failure modes co-occur often but are mechanically distinct — an account can pair Smart Bidding with broad match perfectly and still leak 15-20% of spend because nobody has touched the negatives panel since launch. Smart Bidding lowers bids on losing queries; it does not block them. Only negatives block. The blast radius is usually wider than operators expect because the wasted spend is invisible inside a single all-up CPA — it only surfaces when you cut the search-terms report by intent bucket and compare cost against zero-conversion buckets. Sequence this fix before any broader budget reallocation: a campaign that is 20% off-intent will look budget-constrained when it is actually negatives-constrained, and adding budget pours fuel on the leak.

Related rules + concepts

• Fix: broad match without Smart Bidding — sibling rule, focuses on bidding pairing rather than ops hygiene. Both can fail simultaneously.
• Negative keywords — underlying targeting concept, match-type semantics, scope hierarchy.
• Search Query Mining — recurring workflow that produces the negative-keyword input for this fix.
• Fix: negative keywords missing — existence tripwire for accounts with zero negatives; this rule is the next layer up (maintenance cadence).
• Smart Bidding — why "Smart Bidding will sort it out" is the wrong mental model for off-intent queries.
• Brand search cannibalization — related when broad match in non-brand campaigns matches brand queries.

Sources

1. Google Ads Help — About negative keywords. https://support.google.com/google-ads/answer/2453972 (accessed 2026-05-27)
2. Google Ads Help — About broad match. https://support.google.com/google-ads/answer/2497828 (accessed 2026-05-27)
3. Google Ads Help — About the search terms report. https://support.google.com/google-ads/answer/2472708 (accessed 2026-05-27)
4. Search Engine Land — The real strategy behind negative keywords in 2026. https://searchengineland.com/negative-keywords-strategy-476563 (accessed 2026-05-27)
5. Google Ads Help — About close variants. https://support.google.com/google-ads/answer/9342105 (accessed 2026-05-27)