Why does Google Ads show my ads for searches completely unrelated to my keywords?

Google's close-variant matching and Broad Match expansion interpret keyword intent semantically, so your ads are eligible for queries that share meaning, synonyms, or inferred topic — not just exact wording. This is intentional: Google's auction revenue grows when more queries match more advertisers, so 'related' is defined very generously. The practical consequence is that a B2B SaaS keyword like 'project management software' can trigger ads for 'excel project tracker template' or 'asana vs trello' — queries that rarely convert at your target CPA. The Search Query Bleed Audit measures how far each of your keywords has drifted from its original intent and quantifies the cost of that drift.

Exact match is no longer exact — how to deal with close variant matching

You cannot disable close variants in Exact match since Google removed that control in 2018, so the mitigation is additive rather than preventive: maintain aggressive negative keyword lists, review your Search Terms report weekly for drift, and use ad group-level negatives to prevent cross-contamination between tightly-scoped ad groups. Practitioners running 25,000+ negatives on exact-match-only campaigns still report irrelevant queries slipping through, which indicates this is now a managed-waste problem rather than a solvable one. The Search Query Bleed Audit identifies which of your Exact match keywords are triggering the most off-target variants and ranks them by the cost of that expansion.

How to stop broad match keywords from showing for irrelevant searches

Three levers: add aggressive campaign-level and shared list negative keywords for every irrelevant theme you see in the Search Terms report, tighten Smart Bidding's CPA or ROAS target so Google only chases high-confidence matches, and demote or pause any Broad keyword whose expansion-query CPA exceeds 2x its exact/phrase counterpart. Broad match will never be tight — the design assumes you feed negatives faster than it expands. The Search Query Bleed Audit calculates each Broad keyword's expansion rate (unique queries triggered per keyword) and Broad-vs-Exact CPA ratio, so you know which keywords are genuinely discovering value and which are pure leakage.

Is broad match just a research tool or can it actually convert?

It can convert — but only when three conditions hold: the campaign generates 30+ conversions per month so Smart Bidding has signal, the account has disciplined negative keyword hygiene, and the target CPA/ROAS is tight enough to suppress exploratory bids. Outside those conditions, Broad Match functions primarily as a research tool for discovering new converting queries, which you should then promote to Exact or Phrase match in a separate ad group. Running Broad as your primary match type without all three conditions is what drives the 20–40% waste rate practitioners report. The Search Query Bleed Audit tells you specifically whether Broad is discovering value or destroying it in your account.

Should I switch from broad match to phrase match to reduce waste?

If your Broad keywords have an AI-Expanded CPA more than 2x your Human-Controlled CPA, yes — switch the worst offenders to Phrase match and observe over 14 days. Phrase match still expands (it respects close variants too) but with a much tighter semantic boundary, typically reducing waste by 30–50% while preserving 70–90% of conversion volume. Don't do a blanket switch; Broad keywords that are actually converting well should stay as Broad so Smart Bidding retains discovery capacity. The Search Query Bleed Audit identifies which specific Broad keywords are the best switch candidates based on their expansion rate and CPA delta.

Search terms report doesn't show all queries — how much am I missing?

Google suppresses search terms with very low volume for user privacy, reporting them in aggregate under 'Other'. In most accounts, 'Other' accounts for 20–40% of search-term clicks; in high-volume accounts or sensitive verticals it can exceed 60%. This means you cannot audit a meaningful share of the queries your budget is funding. The Search Query Bleed Audit quantifies your unseen-query share, estimates the leakage within it using the conversion rate of your visible terms, and identifies which campaigns have the worst visibility gap.

How much of my broad match spend is actually converting?

In most accounts running Broad Match with Smart Bidding, 50–70% of Broad spend goes to queries with zero attributed conversions — but this sits hidden inside a blended campaign CPA that looks acceptable because the 30–50% that does convert covers for the waste. The waste is only visible when you split CPA by match type: Exact and Phrase typically land at or below target, while Broad lands 2–4x above. The Search Query Bleed Audit calculates Human-Controlled CPA (Exact + Phrase) versus AI-Expanded CPA (Broad + close variant) side by side for every campaign.

What is the difference between Human-Controlled CPA and AI-Expanded CPA?

Human-Controlled CPA is the cost per conversion from queries that closely match keywords you deliberately chose — Exact match terms and Phrase match terms where the core phrase is intact. AI-Expanded CPA is the cost per conversion from queries the algorithm added via Broad expansion, close variant matching, or semantic inference. In well-functioning accounts they are within 20% of each other; in leaky accounts AI-Expanded CPA runs 2–4x higher, meaning you are paying significantly more per conversion for traffic you did not choose. The Search Query Bleed Audit calculates both side by side and flags campaigns where the ratio exceeds 2x as candidates for match-type tightening.

How to use negative keywords to stop search query bleed

Work in themes, not individual terms — instead of adding 'free plumber london' as a negative, add 'free' as an exact match negative if your product has no free tier, and 'london' as a campaign-level negative if you don't serve London. Theme-level negatives compound: a well-scoped 50-negative list often outperforms a reactive 500-negative list. Avoid adding single-word negatives as Phrase match unless you've checked they won't block your converting long-tail queries (run a collision check first). The Search Query Bleed Audit groups your leaking queries into themes and produces theme-level negative recommendations rather than term-by-term, so you cut bleed at the source.

How to find all close variants triggering my exact match keywords

In the Search Terms report, set your match type filter to Exact and then compare each shown query against the keyword that triggered it — any divergence is a close variant. Queries that differ by pluralisation, word order, function word insertion ('for', 'to', 'of'), or semantic synonym are all close variants. Google does not flag them as such in the UI, so the audit has to be done by visual comparison or script. The Search Query Bleed Audit automates this — it flags every query where the triggered keyword differs from the actual search term and aggregates the close-variant cost per keyword.

How to stop search term bleed between ad groups

Add ad-group-level negative keywords that exclude the keywords owned by every other ad group in the campaign — for example, if Ad Group A targets 'running shoes' and Ad Group B targets 'hiking boots', add 'hiking' and 'boots' as negatives in A and 'running' and 'shoes' as negatives in B. Without this, Google routes queries to whichever ad group it judges most profitable, which typically concentrates spend in one ad group while starving the others. This matters most for accounts with tightly-themed ad groups where attribution purity affects Smart Bidding. The Search Query Bleed Audit maps cross-ad-group leakage and produces a per-ad-group negative list to restore routing discipline.

How do you manage overly broad search queries in Google Ads?

Run a weekly n-gram analysis on your Search Terms report — break every triggered query into 1-word, 2-word, and 3-word components and calculate cost and conversion rate per n-gram. Overly broad queries typically cluster into a small number of n-grams that drive most of the waste (often 10–15 n-grams account for 60–80% of bleed). Exclude those n-grams as theme-level negatives. The Search Query Bleed Audit pre-computes this n-gram analysis for your account and surfaces the highest-cost low-conversion n-grams ready to add to your shared negative lists.

ClickCatalyst

Search Query Bleed · Methodology

Back to Search Query Bleed Audit

Authority Layer · 8 metrics · 7 data sources

Inside the Search Query Bleed Engine

Every metric in your Search Query Bleed audit is derived from documented formulas. This page explains query waste rate calculation, match type waste attribution, keyword expansion rate, Human vs AI CPA comparison, and concentration ratio — with exact thresholds for every verdict.

Jump to metric

Query Waste Rate01 Match Type Waste Attribution02 Keyword Expansion Rate03 Human Controlled vs AI Expansion CPA04 Quality Score vs Expansion Rate05 Waste Concentration Ratio06 Negative Keyword Candidate Selection07 Query Length Analysis08

Data Sources

read-only OAuth · no campaign changes ever

All data is read via Google Ads API and GA4. Derived intelligence tables are pre-computed in our pipeline before your audit runs.

Google Ads API — Search Terms

google_ads_search_terms_historical

Search term text, cost, clicks, conversions for waste and query length analysis

Google Ads API — Matched Keywords

google_ads_search_terms_matched_keywords

Search term to keyword mapping with match type for expansion rate and control vs AI CPA

Google Ads API — Keywords Performance

google_ads_keywords_performance

Quality Score per keyword for QS vs expansion correlation

Google Ads API — Campaign Performance

google_ads_campaign_performance

Campaign-level spend and impression share for budget reallocation

Google Ads API — Hourly Pulse

google_ads_account_pulse_hourly

Hourly clicks, cost, conversions for hourly bleed pattern analysis

Derived Intelligence — Waste Management

waste_management_intelligence_daily

Pre-classified QUERY waste category with recovery potential per campaign

Derived Intelligence — Search Terms

search_terms_intelligence_daily

Negative candidate flags, recommendations, and match quality scores

Query Waste Rate

Percentage of search spend going to zero-conversion queries

Total query waste from waste_management_intelligence_daily (WHERE waste_category = 'QUERY') divided by total search spend from search_terms_historical. The intelligence layer pre-classifies waste at the campaign level with recovery_potential per category.

⚠️

Why this matters: Above 30% means the problem is too large for negative keywords alone — the account needs structural match type changes. Below 15% is healthy monitoring territory. The rate normalizes the waste figure so you can compare across accounts of different sizes.

💡

In plain English: We add up all the money spent on search terms that got zero conversions, then divide by total search spend. The result is the percentage of your search budget that produced nothing — no leads, no sales, no engagement.

< 15%Healthy — monitor with regular negative keyword reviews

15–30%Significant — negative keyword pruning needed

> 30%Structural — match type restructuring required

Match Type Waste Attribution

Which match type is responsible for the waste?

A proportional allocation query: campaign-level waste from the intelligence layer is distributed across match types based on each match type's cost share within that campaign. This avoids requiring waste classification at the individual search term level (which the intelligence layer doesn't store).

⚠️

Why this matters: Broad match dominating the waste chart is a structural problem. Exact match showing waste reveals Google's close-variant expansion. Phrase match waste indicates targeting specificity gaps. Each diagnosis requires a different fix path.

💡

In plain English: We can't directly measure waste per match type — so we use a smart workaround: we know each campaign's total waste, and we know how much each match type spent in that campaign. We distribute the waste proportionally. If broad match is 70% of a campaign's spend and the campaign wastes $1,000, broad match gets attributed $700 of that waste.

Broad > 60%Structural leak — migrate keywords to phrase match

Exact showing wasteClose variant creep — add variant negatives

Keyword Expansion Rate

How many unique queries does each keyword trigger?

From search_terms_matched_keywords: COUNT(DISTINCT search_term) per matched_keyword_text. Keywords triggering 5+ unique queries are included. The waste percentage per keyword is the cost of zero-conversion queries divided by total keyword cost.

⚠️

Why this matters: A keyword triggering 50+ unique queries with high waste means Google interprets it so broadly that most matched queries are irrelevant. The expansion rate identifies which specific keywords are the source of the bleed — not just that bleed exists, but WHERE it originates.

💡

In plain English: For each keyword in your account, we count how many different search queries it triggered. A keyword like 'shoes' might trigger 200 different searches — from 'shoe repair' to 'horseshoe'. The more unique queries, the broader Google interprets your keyword. High expansion + high waste = that keyword is too generic.

5–20 queriesNormal expansion — manageable with negatives

20–50 queriesHigh expansion — consider tightening match type

> 50 queriesExtreme — keyword too generic for this match type

Human Controlled vs AI Expansion CPA

The definitive performance comparison by match type

All search traffic split by search_term_match_type into two buckets: Human Control (EXACT, PHRASE) and AI Expansion (BROAD, AUTO, SMART_MATCH). CPA calculated per bucket. The gap quantifies the premium you pay for the volume that broad match provides.

⚠️

Why this matters: If AI expansion CPA is 2x human CPA, you're paying a 100% premium for every conversion from broad match. Whether that volume justifies the premium is a strategic decision — but most accounts don't know the premium exists until they see this comparison.

💡

In plain English: We separate all your search clicks into two groups: keywords you deliberately chose (exact and phrase match) vs keywords Google expanded into (broad and auto match). Then we compare cost-per-conversion. If Google's expansion costs 2x more per conversion, the extra volume comes at a steep price.

AI CPA ≤ HumanExpansion efficient — trust the algorithm

AI CPA 1–2x HumanVolume premium — evaluate if worth it

AI CPA > 2x HumanExpansion tax — tighten match types

Quality Score vs Expansion Rate

Low QS keywords expand more aggressively into junk

Cross-join between keyword expansion data (from matched_keywords) and Quality Score data (from keywords_performance). When a keyword has low QS AND high expansion, Google considers your ads less relevant to that keyword yet still matches it to many queries — compounding the irrelevance with each expansion.

⚠️

Why this matters: This reveals a structural paradox: the keywords Google rates as least relevant get expanded to the most queries. Each expansion degrades relevance further, driving up CPCs. Low QS + broad expansion = paying a premium for irrelevant traffic.

💡

In plain English: We combine two data points per keyword: Google's quality rating (Quality Score) and how many search queries the keyword triggers. The worst combination is a low quality score AND high expansion — it means Google thinks your ads are a poor match, yet keeps showing them to more and more different searches.

HIGH RISKQS < 5 + 10+ queries — worst combination

MODERATEQS 5–7 — room to improve relevance

OKQS ≥ 7 — expansion at acceptable quality

Waste Concentration Ratio

Can 20 negative keywords fix most of the problem?

Total zero-conversion query cost ranked descending, top 20 summed, divided by total waste. Above 60% means the waste is concentrated in a small number of high-cost terms — a short negative keyword list recovers the majority. Below 40% means waste is long-tail — structural match type changes needed.

⚠️

Why this matters: This changes the fix strategy entirely. Concentrated waste = 20 negatives + 10 minutes = 60% recovery. Long-tail waste = match type restructuring + keyword consolidation = weeks of work. Knowing which scenario you face determines the project scope.

💡

In plain English: We rank all wasteful search terms by cost, take the top 20, and calculate what percentage of total waste they represent. If the top 20 account for 60%+ of all waste, you can fix most of the problem with a short negative keyword list. If they account for less than 40%, the waste is spread across thousands of small terms and you need structural changes.

> 60%Concentrated — 20 negatives fix most of it

40–60%Mixed — negatives + match type review

< 40%Long tail — structural rebuild needed

Negative Keyword Candidate Selection

How the intelligence layer identifies terms to block

From search_terms_intelligence_daily WHERE is_negative_candidate = TRUE. The intelligence layer pre-flags terms using multiple signals: zero conversions, cost above account average, pattern matching against known non-commercial intent. Priority is determined by cost relative to the average wasteful query cost.

⚠️

Why this matters: Manual negative keyword review is subjective and time-consuming. The intelligence layer applies consistent criteria at scale — every term in the list has been verified against the same thresholds. Priority ordering ensures the operator works top-down for maximum impact per minute spent.

💡

In plain English: Our intelligence engine scans every search term and flags the ones that should be blocked. It uses three criteria: zero conversions, cost above average, and intent patterns. Each flagged term gets a priority: HIGH (costs 2x+ the average waste), MEDIUM (above average), or LOW (below average).

HIGHCosts 2x+ average waste — add today

MEDIUMAbove average — add this week

LOWBelow average — batch add monthly

Query Length Analysis

Longer queries = more intent = less waste

All search terms grouped by word count (using ARRAY_LENGTH(SPLIT(search_term, ' '))). CVR and waste percentage calculated per word-count group. Nearly universally, 1–2 word queries have the highest waste because intent is unknowable — 'shoes' matches everything from 'shoe repair' to 'horseshoe'.

⚠️

Why this matters: This analysis validates the match type problem from a completely different angle. If 1–2 word queries have 60% waste while 4+ word queries have 10%, the data proves that broader, shorter queries are structurally wasteful. It's not about individual bad terms — it's about query specificity.

💡

In plain English: We group all search terms by how many words they contain and compare conversion rates. One-word searches almost always waste the most because they're too vague. Four-word searches like 'buy red running shoes size 10' convert much better because the user knows exactly what they want.

1–2 wordsTypically highest waste — too generic

3 wordsModerate intent — acceptable for phrase match

4+ wordsHigh intent — usually converts well

Proprietary notice: The methodology, scoring models, waste classification logic, and recovery projections are proprietary to ClickCatalyst Digital and provided for informational purposes only. Results are subject to standard market volatility. Recovery projections are estimates, not guarantees. Contact us within 14 days if any metric appears inaccurate.

Also worth reading

How to Read Your Search Query Bleed Report →

Every section explained with exact action steps

See these metrics applied to your account

Run My Free Search Query Bleed Audit