> ## Content Index
> Fetch the complete content index at: https://cascader.io/blog/llms.txt
> Use this file to discover other available public pages before exploring further.

# How to Audit Negative Keywords in an Inherited Google Ads Account
- URL: https://cascader.io/blog/inherited-google-ads-negative-keyword-audit/
- Published: 2026-10-14T15:00:14.000Z
- Updated: 2026-10-18T23:27:50.000Z
- Description: Inherited a list nobody can explain? Trace each exclusion to its purpose, wording and reach—then decide what to keep, narrow, move, remove or investigate.
- Author: Rick Hecker

An inherited negative keyword can make you hesitate. It excludes a service the business appears to offer, but perhaps there's a reason you haven't found yet. That hesitation is reasonable. Someone may have learned something expensive before you arrived.

Your job is to give the exclusion a purpose you can defend today. What business decision should it express? Which searches do its words exclude? Where does it apply? A useful restriction can have excessive wording or reach. You may also find that it belongs exactly where it is. The number of rows deleted says little about the quality of the audit.

We'll use a constructed commercial HVAC account. Its service lead confirms that the provider accepts air-cooled chiller work and declines water-cooled chiller work. A broad negative, `chiller`, excludes both; its original reason is unknown. We'll trace that restriction, assess a narrower candidate, and leave a decision the next manager can continue.

## Get a business answer you can use

Bring accepted, declined, and uncertain enquiries to someone responsible for accepting the work. Service pages, campaign aims, lead dispositions, and handover notes help you prepare the question. A page tells you what prospects see; the responsible person confirms what the business will take on.

“Would you keep this negative?” gives the service lead an advertising problem. “Can the team take a commercial air-cooled chiller repair, and what would make it decline the job?” gives them a business question. Ask what information is needed when the equipment is unspecified too.

Keep three kinds of knowledge separate: **historical rationale**, if documented; **current confirmation**, with its source and date; and **inference**, a plausible explanation nobody has confirmed. Here, the equipment boundary is confirmed. The idea that someone originally added `chiller` to avoid water-cooled enquiries remains an inference. [Change history](https://support.google.com/google-ads/answer/2454137?hl=en&ref=cascader.io) can supply timing and authorship, but it doesn't guarantee the commercial explanation.

A missing note alone is a poor reason to remove a restriction. Equally, you needn't preserve it forever while reconstructing the past. An unknown current boundary is a different gap: if nobody can confirm which work is accepted, assign that question to a business owner before designing the change.

Serviceability establishes relevance, not an obligation to buy every relevant click. Keep that commercial question open as you follow the restriction into the account.

## Find every place the restriction is doing work

Map each source and every destination receiving it. Our relevant entries are:

| Negative                                 | Match type | Source                                 | Destinations                  |
| ---------------------------------------- | ---------- | -------------------------------------- | ----------------------------- |
| chiller                                  | Broad      | Shared list: Legacy service exclusions | Repair, Maintenance, Projects |
| chiller                                  | Broad      | Direct campaign entry                  | Maintenance                   |
| commercial air conditioning installation | Exact      | Shared list: Repair routing            | Repair, Projects              |

Maintenance has two sources blocking `chiller`. Editing the shared list alone leaves its direct entry in force. This is a good reason to do the apparently dull inventory work: it prevents a sensible decision from becoming an incomplete change.

The installation entry raises a different question. Projects seeks installation enquiries, yet receives that exact exclusion. Repair's reason needs confirmation. The list name can suggest a question; it cannot answer it. We'll return to that reach problem when planning the change.

Retain a dated baseline and the original downloads. Reconcile exports with native account records; Google's [download and edit guidance](https://support.google.com/google-ads/answer/7102298?hl=en&ref=cascader.io) doesn't mean one export contains every attachment. Capture account context, text, match type, source level, list name and owner where relevant, all receiving campaigns or ad groups, duplicates, and available application dates. Keep separate sources separate, even when their text is identical.

Inspect [account negatives](https://support.google.com/google-ads/answer/11396330?hl=en&ref=cascader.io), which automatically apply to eligible Search and Shopping inventory across relevant campaigns; [shared lists](https://support.google.com/google-ads/answer/2453983?hl=en&ref=cascader.io), whose edits affect attached campaigns; and [direct campaign and ad group entries](https://support.google.com/google-ads/answer/7102995?hl=en&ref=cascader.io), including overlaps with intended targeting. For [manager-owned lists](https://support.google.com/google-ads/answer/7519927?hl=en&ref=cascader.io), verify application from the client account. Finding one in the library isn't enough.

The three example rows don't establish that other blockers are absent. Finish this pass knowing where a restriction operates and what could survive an edit. Then test whether its wording fits the business boundary.

## Test everything the narrower candidate opens

Choose representative queries for offered work, declined work, and intent you cannot classify yet. Actual searches help where available; illustrative queries can expose a wording problem too. Keep the two identifiable in your notes.

For Search negatives, broad requires all terms in any order; phrase requires the ordered phrase and permits additional words; exact excludes that query without extra words. These are [negative match rules](https://support.google.com/google-ads/answer/2453972?hl=en&ref=cascader.io), so don't read them as positive keyword matching. Our negative keywords guide explains [how negatives match by spelling, not meaning](https://cascader.io/blog/negative-keywords-google-ads/#why-a-negative-misses-what-your-keyword-finds).

Compare broad `chiller` with a candidate phrase negative, `"water cooled chiller"`:

| Illustrative query          | Business judgment                                                | Broad chiller | Candidate phrase               |
| --------------------------- | ---------------------------------------------------------------- | ------------- | ------------------------------ |
| air cooled chiller repair   | Accepted equipment; other commercial qualifications still matter | Excluded      | Not excluded by this candidate |
| water cooled chiller repair | Declined equipment                                               | Excluded      | Excluded                       |
| chiller repair near me      | Equipment unspecified                                            | Excluded      | Not excluded by this candidate |

The first row shows that the wording also catches offered work; the second gives you a restriction worth preserving. Together they support considering narrowing. Simply deleting `chiller` without protecting declined work would leave half the business decision behind.

The candidate fits these three checks better. It doesn't establish comprehensive coverage of water-cooled searches. Review wording such as `water cooled chillers` and relevant synonyms. Google says singular/plural forms and synonyms need separate consideration; casing and misspellings “for which Google changes the search results” are handled automatically; when a misspelled version of the declined wording appears in your search terms report, you can add that form as its own negative. [Variant guidance](https://support.google.com/google-ads/answer/12437241?hl=en&ref=cascader.io). Build coverage around the boundary, rather than collecting spelling variants for their own sake.

Now look at the third row. It's tempting to put it in whichever column makes the decision easier. The query hasn't told you the equipment type. Leave it uncertain.

The candidate stops excluding both accepted **and unspecified** equipment in these tests. Naming accepted work as the goal doesn't separate those effects. Before proceeding, establish how appropriate ads or landing-page information will explain the boundary and how forms or call handling will capture equipment type. The service lead can confirm the work described; the PPC manager checks wording; the enquiry-handling owner establishes capture and classification.

If that qualification cannot be made workable, revise or defer the proposal. Later, qualified air-cooled enquiries at agreed acceptable cost would support continuing; declined or still-unclear enquiries at unacceptable cost would prompt reconsideration.

A budget-conscious manager could also reasonably prefer the broad restriction. A current, evidenced campaign or budget decision to deprioritize all chiller traffic could justify it despite the equipment being serviceable. That's a possible objection, not recovered account history. The offered-service conflict deserves attention; it doesn't settle the purchase decision.

These checks show what the wordings exclude, not observed lost leads, demand, or revenue. Ceasing to block a query doesn't guarantee serving, and serving doesn't establish a qualified enquiry. To assess the commercial choice, give the reports a question they can answer.

## Use reports for the questions they can answer

You may reasonably want performance data to settle this. First distinguish the claim you're testing. A confirmed refusal of water-cooled work needs current business confirmation, not a zero-conversion report. “Chiller searches produce poor leads” is a performance claim requiring dated traffic and spend, dependable tracking, qualification, and downstream outcomes. Campaign priorities need a current owner and reason.

The [search terms report](https://support.google.com/google-ads/answer/2472708?hl=en&ref=cascader.io) describes served searches and omits some low-activity rows for privacy. Its match-type column describes positive keyword matching. From that scope, we can infer that it cannot establish all demand a negative prevented. If air-cooled searches are absent after broad `chiller` was applied, you can't conclude that nobody wanted air-cooled work. You're examining a report produced while the exclusion was doing its job.

If an apparently blocked query appears, compare its date with the negative's application date and destination at the time. A query served before application doesn't show the current rule failing. [Google's date check](https://support.google.com/google-ads/answer/9701952?hl=en&ref=cascader.io).

Check what counts as a conversion and what happened afterwards. A form submission and an accepted repair job answer different questions. For a worked performance decision, [The $600 zero is not a verdict](https://cascader.io/blog/the-600-zero-is-not-a-verdict/) compares blocking a term, buying a bounded observation, and repairing tracking. Incomplete downstream records limit the conclusion. Allow outcomes time to arrive too: [conversion delay](https://support.google.com/google-ads/answer/9347141?hl=en&ref=cascader.io) can make recent CPA look higher and ROAS lower. The window needs to suit activity, lag, and the sales process.

These limits cut both ways. Missing rows don't prove an exclusion profitable; attractive illustrative queries don't prove recoverable value. A wording conflict and uncertain commercial value can occupy the same record. Credible evidence may support keeping the restriction. Inadequate evidence needs a specific investigation, rather than confidence borrowed from a tidy report.

## Choose an action and make the whole change

Use five working categories, not Google rules:

- **Keep:** current purpose, wording, and reach are justified.
- **Narrow:** preserve a valid restriction with less sweeping wording.
- **Move:** preserve it in fewer or different destinations.
- **Remove:** evidence supports ending the restriction at the relevant destination.
- **Investigate:** a material question remains; name its owner or answer source, decision-changing finding, and interim configuration.

Every action needs a reason. Keep preserves useful knowledge. There is no need to manufacture an edit to demonstrate effort.

For chiller, establish current intent, declined-wording coverage, and workable qualification. Agree how outcomes will be reviewed before implementation. If qualification, economics, or an evidenced campaign priority doesn't support proceeding, revise or defer. Record the broad restriction retained while investigating, who supplies the missing answer, and what would reopen the proposal. “Needs review” leaves the next reader nearly where you started.

When ready, write the intended sources and destinations. Add the justified narrower protection, save it, then reopen and verify the text, match type, and attachments. **Only after that protection is present** remove both broad sources: `chiller` from Legacy service exclusions and the direct Maintenance entry. This order prevents a gap in declined-work protection; addressing both sources prevents Maintenance retaining the block. Check other applying negatives from the inventory too.

For installation, ask the campaign-strategy owner: should Repair exclude the exact query `commercial air conditioning installation`, and what current purpose justifies that? Projects' conflict doesn't decide Repair's answer. A conflict with positive targeting can be intentional. The three answers lead to different treatments:

- **Justified:** add and save a direct exact Repair entry, then read it back before removing the exact entry from the shared source affecting Projects.
- **Rejected:** record the owner's answer, source, date, and reason. Remove the shared exact entry without creating a direct replacement by habit.
- **Unanswered:** keep the investigation open. Retain Repair's current restriction temporarily, assigned to the strategy owner. Revisit when that owner supplies a purpose decision or campaign intent changes, with an agreed date backstop. Hold shared-entry removal until Repair's treatment is decided. You can prepare a scoped Projects correction meanwhile. If using Projects-only list detachment, first inspect all contents and save and verify any justified Projects exclusions it would otherwise lose; Repair's existing attachment can remain temporarily. This precaution doesn't establish that Repair's restriction is right.

Whole-list detachment can remove useful rules alongside the troublesome one. Inspect and preserve those before detaching. And exclusion in Repair doesn't route a query to Projects or force it to serve. Projects still needs [suitable targeting](https://cascader.io/blog/when-a-search-term-deserves-its-own-keyword/), ads, a page, and the ability to accept the work.

After the agreed edits, reopen every changed source and affected destination. Inspect saved text, match type, level, and attachments. Confirm retained and replacement rules are present and intended old blockers are absent, including Maintenance's direct entry. Repeat the query reasoning against **all applicable saved negatives**, not just the edited row. Retain dated before-and-after evidence.

An edited sheet is a proposal. A successful save is an accepted edit. Read-back establishes whether the account matches the decision. Label the record **proposed**, **saved but unverified**, or **verified configuration** accordingly. Verified configuration still doesn't guarantee serving or commercial results; those need later observation.

## Leave a decision the next manager can continue

Keep one accessible record per decision, linking duplicates and related edits. Include entry and match type; sources and destinations; current purpose and confirmation source/date; historical rationale and labelled inference; action and reason; intended versus performed operations; verification evidence and status; open questions and owners; review criteria and owner.

The chiller note can be compact:

> **Reason:** Service lead confirms air-cooled acceptance and water-cooled refusal; attach confirmation and date. Original broad-negative rationale remains unknown. Possible water-cooled avoidance is inference, not history.
> 
> **Decision:** Proposed narrow. Illustrative checks show wording wider than the equipment boundary, without estimating lost demand. Sources: Legacy service exclusions → Repair, Maintenance, Projects; direct broad `chiller` → Maintenance.
> 
> **Pending:** PPC manager reviews variants, service lead confirms their meaning, enquiry-handling owner establishes equipment capture. A coverage gap or unworkable qualification changes readiness; retain broad pending investigation if unresolved. Campaign priority and qualified-lead economics also inform the choice.
> 
> **Configuration:** Replacement protection must be saved and checked before both broad removals. Implementation and verification pending; attach performed edits, destination checks, and dated evidence when available.
> 
> **Review:** Revisit readiness when answers arrive; reopen for changed service or campaign intent. PPC manager owns outcome review with service and enquiry owners supplying classification and downstream records.

Make the outcome review agreed earlier specific. Name the destinations, how accepted, declined, and unclear enquiries will be classified, and the qualified outcome and cost criteria that warrant another decision. Identify who supplies each answer or dataset. Choose enough activity and time for outcomes to mature, accounting for conversion lag and the sales process; set a date as a backstop.

Review sooner if tracking fails or declined enquiries reach an agreed concern threshold. If evidence remains thin, record what is inconclusive and the next observation needed. Concurrent budget, bidding, targeting, ad, page, or measurement changes limit how confidently you attribute an outcome to the negative edit.

In another account, the service terms will change. The useful habits travel: test offered, declined, and uncertain intent; judge purpose, wording, and reach separately; leave an answerable question where knowledge runs out.

The next manager may keep your restriction or reconsider it as the business changes. Give them enough context to do either thoughtfully. The negative keyword can stay brief; its explanation shouldn't have to be guessed.