Google Ads Quality Score audits for agencies: investigate low-score keywords before choosing what to change

Define the agency's diagnostic objective

An agency Quality Score review should begin with a decision the client actually needs to make. The team may be investigating an expensive keyword, checking whether a specialised campaign communicates its offer clearly, or deciding whether an obsolete service should remain eligible. Those questions need different evidence. A list of low scores can organise attention, but it cannot supply the business brief that makes a particular change sensible.

Write the objective in plain language before opening a correction flow. For example, investigate whether keywords promoting evening classes still match the available course schedule and landing page. This gives the account specialist, copywriter and client a common subject. It also prevents a numerical threshold from quietly replacing the real objective with making every diagnostic indicator look healthier.

Keep the client's current offer, service area, availability and intended outcome attached to the review. An agency employee may understand the account's naming conventions while another employee handles the website. Neither can infer from a keyword score whether the advertised course is still offered. Business facts supplied by the client remain important even when the provider read is technically complete.

The appropriate conclusion may be to retain a keyword while improving its explanation, stop promoting an unavailable offer, ask for more evidence, or leave a sparse observation unresolved. Each is a valid review outcome when supported. A successful agency process makes the reason and owner clear; it does not require a pause operation at the end of every diagnostic conversation.

Understand what Quality Score represents

Google's Quality Score guidance describes a keyword-level diagnostic from one to ten, informed by expected click-through rate, ad relevance and landing-page experience. Google explicitly distinguishes it from a business KPI and an input to the auction. Missing scores can reflect insufficient evidence rather than a numerical zero. Use these distinctions when explaining the audit to a client.

A low score is a reason to investigate the relationship between the query, the ad and the destination. It is not an invoice amount, a count of disappointed customers, a direct measurement of profitability or a complete explanation of actual click cost. The client should not hear that raising a displayed number by a particular amount guarantees an equally predictable reduction in spending.

Do not create an account performance target by averaging the displayed scores. That calculation can look precise while hiding important differences between keyword occurrences, traffic volumes and business purposes. A diagnostic review can list the returned low-score entries and discuss their actual context without turning an ordinal indicator into an agency success metric.

The agency should distinguish a provider description, an observation and an interpretation. The provider defines the diagnostic. The current read identifies a keyword with a reported score. The specialist explains why that occurrence deserves investigation in this client's circumstances. Keeping these three statements separate makes it easier to correct an interpretation without disputing the observed field.

Read the current audit scope

ppctor's current low-score reader requests positive keywords whose keyword, ad group and campaign are enabled, with a reported score at or below three. It returns the keyword text, score, campaign name, group name and an internal resource reference. The audit presents the names and diagnostic value for review; a client-facing explanation does not need the raw resource reference.

The account card is relevant when the account has an enabled Search campaign. It is an informational surface rather than a contribution to the overall audit score. Returned entries carry their campaign and group context and offer a pause option. That option describes an available operation, not a conclusion that the agency has already established the keyword's commercial value.

A campaign-level card reuses the account reader's returned entries and selects rows by campaign name. This is useful for a focused review, but it inherits the original collection's limits. It is not a new independent complete inventory query. If several campaigns share an identical name, a display-name filter should not be treated as proof that every returned occurrence belongs to one uniquely identified campaign.

Paused parents and paused keywords are outside this reader's displayed eligibility scope. A finding therefore describes currently enabled occurrences under enabled parents. Historical spending by a keyword that has since been paused belongs in a separately defined performance query. Neither an absent row nor an empty current finding establishes that the keyword never existed or never generated cost.

Preserve the result limit

The current low-score query orders the returned scores from lowest upward and limits the collection to 300 rows. A review with fewer returned entries is a useful set of observations. A collection that reaches the limit requires particular care: the displayed set does not establish that the account has exactly that many matching keyword occurrences.

Explain the unit before reporting a count. These are keyword occurrences in account structure, not necessarily distinct written phrases. The same phrase can appear in several groups and campaigns. Combining identical text into one unique phrase changes the question and discards parent context that may determine whether the agency should retain one occurrence while pausing another.

A complete-account count needs a report whose collection and coverage satisfy that request. Do not infer completeness from the size of the audit preview. Likewise, a campaign filtered from the capped account set can omit entries that would have appeared in a dedicated broader query. A local filter does not restore rows that the provider collection never returned.

Keep the visible list readable without disguising the limit. A client may need a few named examples in conversation and a fuller report for internal review. The summary should state the actual coverage when it matters. An exported preview still remains a preview if its source was capped; changing the file format does not make the underlying evidence complete.

Separate unavailable scores and failed reads

A successful low-score query with no matching entries is different from a failed provider read. The audit records failures for the low-score section and presents an unavailable check rather than silently treating the fallback empty array as a clean result. Preserve that distinction in a client summary, especially when the account also contains successful sections.

The query selects returned low scores; it does not provide a census of every keyword without a score. An empty successful result therefore supports the narrow conclusion that no matching entries were returned in that scope. It does not prove that every enabled keyword has a known score above three. A missing diagnostic is not transformed into a high score merely because it did not match the filter.

Do not interpret the card's compact healthy-state wording more broadly than its evidence. An agency report should distinguish no low-score matches from all keywords scored and inspected. This avoids making an unverified statement about missing diagnostics, other channels, paused structures or incomplete collection coverage.

If the client asks why one keyword has no score, first establish that the keyword and correct parent occurrence were identified. Then seek the relevant available diagnostic evidence. Do not invent a traffic explanation from an empty audit list. The provider's general description of missing-score conditions explains a possibility; it does not confirm the cause in this specific account.

Identify each keyword occurrence

Record the keyword text with its ad group and campaign. This is the minimum readable structure for a review that might later affect eligibility. A keyword called evening classes in one group is not interchangeable with the same text in another group. Its ad promise, destination, targeting and client purpose can differ even when the visible phrase is identical.

When match type matters, retrieve it from an appropriate supported field rather than assuming it from punctuation in a display label. The low-score reader itself does not include match type in its returned presentation fields. A later detailed report or edit target can establish that additional identity. The assistant should state what it has actually read, not silently complete the label from an expectation.

If names are duplicated, ask for the minimum context needed to disambiguate. A human-friendly proposal can use campaign and group names while the tool retains exact resource identity internally. Removing raw IDs from client output should never mean discarding internal ownership and identity checks.

Keep identities stable across an investigation. A retained dataset can help interpret a follow-up such as leave the weekend-class keyword alone, but a live proposal still needs the correct current target. Historical diagnostic rows, a renamed group and a newly created replacement keyword can all refer to different resources despite a familiar phrase.

Prioritise a client investigation

The score ordering helps an agency scan a diagnostic list, but it does not rank economic urgency. A keyword with a lower diagnostic score may have little current spending. Another occurrence may matter more because it promotes an important service and has substantial recent cost. The priority should follow the client's question and the observed evidence rather than the audit colour alone.

Define a performance window when the priority depends on spending or outcomes. Last seven completed days and last thirty completed days can tell different stories. A current low-score field does not supply either period's performance automatically. Keep the observation time of the diagnostic separate from the date bounds of the commercial report.

Combine evidence only at compatible identities and grains. A campaign revenue total repeated beside every keyword would falsely attribute the same outcome to each child. A proper keyword review needs supported keyword-level performance or a clearly labelled broader context. Parent totals can explain scale without pretending to measure the individual keyword's return.

Write a manageable investigation queue. For each selected occurrence, identify the business concern, missing evidence and responsible specialist. The first next step may be to verify the offer or inspect the client's supplied destination. A queue of every low-score row marked pause immediately is an action list without a business assessment.

Read commercial outcomes independently

Choose the outcome definition before interpreting a keyword's performance. Total primary conversions, purchase conversions, primary lead categories and a named conversion action answer different questions. Conversion value can include value attached to non-purchase actions when the selected metric includes them. Do not relabel a broad value metric as purchase-only revenue.

A primary call conversion can carry value, and an offline conversion may represent a later stage in the same customer journey. Those facts can matter for the client's keyword decision. A low diagnostic score does not make legitimate primary outcomes disappear from reporting. It also does not establish that several actions are independent customers whose counts can be treated as unique people.

Use strict and inclusive conditions accurately. Keywords with cost greater than a stated amount differ from those with cost at least that amount. A request combining substantial cost and zero reported conversions needs both conditions at the same supported row grain. Missing outcome evidence is not zero, and an undefined ratio cannot be substituted with a convenient numeric value.

Cross-source comparison requires the declared measurement scope. A named Ads conversion can be compared with a selected compatible GA4 event through the supported reporting framework, with availability checked for the requested plan. The low-score card alone does not select that event or establish equivalence. Keep the diagnostic review and measurement comparison as separate pieces of evidence.

Request the evidence behind a diagnosis

The current low-score list supplies the overall score, not the three component statuses. An agency should not tell the client that landing-page experience was below average merely because the overall score was low. That diagnosis needs the relevant component evidence from a supported provider report or a direct reviewed account surface.

When asking the assistant for a deeper explanation, specify the keyword occurrence and request the available diagnostic fields. The generic reporting framework can discover and validate supported fields for a query. Field compatibility and source availability still need checking. A field's existence in provider documentation is not enough to promise it in every current tool plan.

Keep the returned observations close to their interpretation. If ad relevance is the observed concern, discuss the connection between the selected phrase and actual ad copy. If the component remains unread, label the proposed investigation as a hypothesis. The agency can develop useful creative ideas without presenting an unobserved provider status as fact.

Do not burden the client with a technical field list when a plain explanation will do. The internal report can retain source identifiers and requested dimensions. The client conclusion should name the occurrence, explain the established concern and describe the next decision. Readability should simplify the evidence, not replace it with certainty.

Review the customer's search intent

A keyword phrase belongs to the client's offer, not to an abstract score target. Ask what the intended visitor wants and what the client can actually supply. For an evening-course advertiser, a phrase seeking a professional qualification may be unsuitable if the business provides informal workshops. The low score can prompt this review, but the offer mismatch must be established separately.

Use actual search-term evidence when the client requests account behaviour. A brainstormed list of possible queries is useful for discussion but is not a retrieved history. The agency should distinguish suggestions based on supplied product context from provider observations about searches that generated recorded activity.

Do not assume that a broad-sounding phrase is always bad or that a specialised phrase is always valuable. The campaign role, landing page and business model can change the decision. An occurrence promoting an introductory class may need a different message from one promoting an advanced programme, even when both relate to the same general subject.

A negative-keyword proposal is a distinct operation from pausing a positive keyword. Pausing changes that keyword occurrence's eligibility. Adding an exclusion affects matching within its selected campaign, group or list scope. The agency should choose the mechanism that addresses the established problem, rather than substituting whichever button appears nearest the diagnostic.

Assess the actual ad promise

Creative review should connect the customer's intent with the current offer. Read the relevant ads if the request concerns what is live in the account. If the client simply asks for alternative headlines from an already supplied brief, the assistant can draft suggestions without repeatedly retrieving account data. Creative work and live-state claims have different evidence needs.

Give the writer a specific constraint: the actual course duration, enrolment requirement, availability or service promise. A vague instruction to improve relevance invites generic copy that may sound polished while saying nothing new. A factual brief lets the agency judge whether the proposed message resolves the diagnosed mismatch.

Review the complete proposed ad when a supported creative edit is requested. Headlines, descriptions, destination and other supplied fields can interact. An operation that replaces creative should not be represented as a tiny invisible patch if its actual behaviour replaces a larger field set. The approval needs to describe the resulting asset clearly enough for the owner to judge it.

An accepted creative write does not establish policy approval, serving volume or a later diagnostic improvement. Keep the exact new content, provider acceptance and subsequent observations separate. The agency may have executed the requested change successfully while still waiting for enough evidence to assess the business result.

Coordinate landing-page responsibilities

The destination should support the ad's promise and help the intended visitor complete the relevant task. That is an agency review objective, not evidence that ppctor has automatically inspected every page. Account URL reads show configured destinations; they do not by themselves prove page content, loading behaviour, consent handling or successful checkout.

Ask the website owner to confirm the facts that require site access. For a course advertiser, those may include the available schedule, enrolment information and a working registration path. An account specialist can coordinate this evidence without presenting a configuration read as a live browser test.

Separate an Ads destination edit from a website change. Updating a final URL alters the provider configuration. Editing the destination's content belongs to the website workflow. An agency handoff should state which party can perform each task and what confirmation is required before the account message points visitors to the revised page.

Do not promise that a new landing page will produce a particular Quality Score or CPC. The client can approve a clearer destination based on its offer and customer journey. Later provider diagnostics and measured commercial results can inform the next review. This preserves a defensible sequence without inventing a numerical guarantee.

Consider group structure without promising a score

Group structure can help an agency keep related messaging and destinations understandable. If an ad group contains several different service promises, a review may find that separate groups would make ownership and creative work clearer. That operational reason needs explaining independently from a claim that reorganising names will automatically repair a diagnostic number.

A new group and new keyword occurrence are separate resources from the existing ones. The agency should preserve the distinction when discussing historical results and approval. Creating a replacement occurrence does not transfer the original keyword's history into a new object merely because the text matches.

Use the supported local dependency plan for requested creation work. A new keyword needs a valid group, and a new group needs a valid campaign. Existing verified parents can be reused. A client request to create a group should not silently create another campaign, and a request to review a diagnostic should not initiate either operation.

Check structural changes against the supported review safeguards. A valid sequence can share one clear confirmation with each action separately described. If the complete setup cannot be prepared safely, ask for narrower proposals before writing. The proposal should reveal each creation and status decision, rather than hiding several effects inside a technical paragraph called optimisation. The client must understand the intended state of every affected resource.

Follow an invented agency review

The following example is invented. Cedar Shore Media manages Rowan Vale Workshops, a business offering short evening classes. An account review returns a low-score occurrence for advanced woodworking in the Introductory Classes group of the Local Courses campaign. The phrase and parent names are illustrative; no client result, price reduction or conversion improvement is implied.

The specialist first checks the offer with the client. Rowan Vale confirms that its advanced programme has been discontinued while introductory classes remain available. This is the deciding business fact. The diagnostic did not reveal the discontinued programme; it helped the agency notice an occurrence whose purpose needed review.

The agency then asks for the selected occurrence's recent performance using a defined completed period. The response is kept separate from the diagnostic score and labelled with the chosen outcome metric. Even if the current period contains little activity, the client can still decide not to advertise an unavailable programme. That decision does not require fabricated wasted-spend arithmetic.

The specialist prepares a pause for the advanced woodworking keyword in the named group and campaign. The introductory-course keyword in another group remains untouched. The confirmation describes that exact occurrence and action, so the owner can see that the proposal is narrow. The proposal does not delete the group, remove the campaign or add campaign-wide exclusions.

After the connected owner confirms, the tool executes the supported edit and checks the resulting status through the conversational verification flow. The receipt establishes the requested keyword pause. It does not announce that every low-score keyword is fixed or that the account's profitability improved.

The agency's handoff records the discontinued offer and the exact retained alternatives. A later request to restore the programme would need a new current review and approval. This example shows why business relevance, parent identity, action scope and technical outcome belong together even when the initial diagnostic appears simple.

Choose a defensible next action

Retaining the keyword can be appropriate when it promotes a valid offer and the available evidence supports continued observation. The agency should record what it is waiting to learn and when it will review again. Retention is a deliberate decision, not an admission that every low-score occurrence must remain indefinitely.

Pausing is appropriate when the owner wants that occurrence to stop being eligible. It is a concrete status operation. Removing an occurrence, replacing its text, revising creative and adding an exclusion have different effects. The specialist should not describe them interchangeably as cleaning up low scores.

If the issue is uncertain, request the missing evidence rather than a speculative write. The required evidence might be an offer confirmation, the correct destination, a component status or the chosen conversion definition. A precise question is more useful than a broad request for the client to explain the whole account.

Confirm current support for the selected operation. The audit offers per-keyword pausing, while the conversational edit framework has its own declared actions, constraints and checks. A provider capability does not guarantee that every convenience tool exposes it. A clear proposal states the actual operation the current tool can prepare.

Understand the separate bulk action

The existing audit also has a pause-all-low-score action. Its execution query is broader than the display query: it selects enabled positive keywords with scores at or below three without repeating the display reader's enabled campaign and group filters or 300-row limit. The agency must not assume that this button operates only on the displayed preview rows.

This scope difference is important when a client has enabled keywords under paused parents. The current display may omit those occurrences, while the separate bulk execution query can select them. A review that intends to affect only named occurrences should use an exact supported proposal rather than relying on an assumed identity between display and bulk collections.

The bulk action permits partial failure. It rereads matching enabled keywords afterward to identify what remains. One operation being refused does not necessarily prevent other selected keywords from being paused. A bulk result therefore needs its actual outcome summary, not a claim that all changes were rolled back when any one failed.

Per-keyword audit pausing has an ownership check and returns the mutation outcome. It does not itself perform the same dedicated status readback as the conversational verification flow. These are distinct paths. Explain the evidence supplied by the path actually used, and do not attach an executed-and-verified claim solely because an API response was accepted.

Review an exact conversational proposal

A conversation-based edit should name the keyword, its ad group and its campaign, followed by the exact action. Keep each logical action in its own readable step. The client should not have to decode an internal mutation object to discover whether the proposal pauses a keyword or an entire parent group.

Use one confirmation for a supported sequence with clearly separated actions. If the complete request needs smaller proposals for safe review, explain that requirement before writing. Do not quietly apply a subset while omitting the rest. Similarly, do not silently add creative changes because they might improve the diagnostic under discussion. Approval remains attached to the specific listed effects.

The pending message says that no changes have been made yet and that the proposal expires in fifteen minutes. The connected owner's confirmation is the boundary before live execution. A request to explain a finding or suggest better wording should remain a conversation until a supported edit has been deliberately proposed and confirmed.

Review the whole resulting state when a request includes several effects. Pausing a keyword and enabling its group are independent operations; one does not imply the other. Each intended status should be visible. The tool's local dependency ordering can organise prerequisites, but it cannot supply business permission for an unrequested change.

Distinguish execution and later results

After confirmation, a visible change-in-progress state tells the owner that the click has been received. The durable execution flow then performs the checks and requested provider work. The progress message is an acknowledgement of ongoing execution, not proof that the provider has accepted every operation.

Current validation checks whether the prepared request remains applicable before sending it. Result verification checks the observed resulting settings afterward. A proposal can become stale between preparation and confirmation. The account may have changed, a resource may have been removed or the requested status may already be present.

A verified keyword pause is a technical milestone. It does not establish a new Quality Score, a policy outcome, increased conversions or reduced cost in a later reporting period. The agency can communicate the successful setting change prominently while leaving the business assessment for a subsequent evidence-based review.

For partial or unverified results, retain the distinction between sent, accepted and read back. Do not claim nothing happened merely because complete verification was unavailable. The owner needs the actual confirmed scope and unresolved portion so that a follow-up does not accidentally repeat a change that already took effect.

Reconcile an uncertain change

A delayed Slack receipt, failed message update and uncertain provider outcome are different problems. The durable execution record and current resource state help determine which occurred. A notification failure is not permission to repeat the mutation, especially when the requested operation might include creation or replacement alongside a status change.

Read the affected current setting when reconciliation is required. If the keyword is already paused, that observation matters even when an earlier message did not arrive. If it remains enabled, the team still needs to understand the recorded attempt and relevant failure before preparing another approval-based request.

Do not promise automatic rollback across a multi-step proposal. The execution may include several accepted operations and an unresolved later step. A recovery plan should preserve known outcomes and identify the remaining work. Restoring a setting is a new deliberate edit with its own current target and approval.

Agency operators should coordinate while reconciliation is underway. Two people interpreting a missing message as a failed change can submit competing follow-ups. Record who is checking provider state and who will communicate the conclusion. This practical ownership prevents uncertainty from becoming several overlapping live requests.

Continue the investigation in Slack

Start with the business question: review low-score keywords in the Local Courses campaign and explain which need an offer check. A useful reply names the observed occurrences, preserves the collection boundary and separates established facts from interpretation. It should not convert the diagnostic question into a bulk pause request.

Follow-ups can use the retained subject while preserving exact identity. Ask for recent cost and primary conversions for one named occurrence, then request draft headlines using the confirmed offer. The performance request requires provider evidence. The headline brainstorm can use the existing brief without another account query merely to produce creative suggestions.

When a chart is requested, choose a meaningful performance series rather than inventing an average Quality Score success line. The native chart flow supports readable compatible metrics and limits each chart series to twenty points. Longer periods can use a declared coarser grain unless daily detail was explicitly requested, in which case necessary splits should be explained.

Keep conversational replies readable and put the user mention before feedback so feedback remains last. Internal request records belong in their appropriate logs; executed edits have their execution record. The client should see the diagnostic, proposed action and outcome in clear language rather than raw resource IDs or provider request bodies.

Reuse relevant evidence efficiently

A current audit dataset can support explanation of its returned rows without rereading unrelated account settings. The campaign card already reuses the low-score collection. Reuse should preserve provenance and limits: the same retained preview does not become complete merely because the client asks a second question.

Request additional fields only when they answer a new requirement. Match type, diagnostic components and period performance are absent from the compact low-score reader. A supported focused query can obtain the necessary additional evidence. There is no reason to scan every creative asset, all product inventory or unrelated Analytics events for a keyword-status question.

Local planning can order dependencies and validate known action shapes without provider calls. Live-state checks remain necessary for exact current edits. Efficiency means avoiding redundant reads while retaining the checks that protect scope and correctness; it does not mean substituting stale observations for current mutation prerequisites.

Do not promise a fixed call count or instant response for every review. Collection breadth, missing evidence, provider conditions and requested outputs affect the work. Explain delays only when they help the owner understand progress. The client's report should remain about the agency decision, not infrastructure details that do not change that decision.

Report a useful client conclusion

Lead with the decision and its evidence. For the invented discontinued-course case, the conclusion is that the selected keyword was paused because the offer was unavailable. Mention the low-score diagnostic as the review trigger, not as proof of the business mismatch or a mathematical explanation of wasted spending.

State the reviewed scope and any collection limit that affects interpretation. A campaign review based on a capped account collection is different from a complete dedicated count. Avoid a confident account-wide claim when the actual work concerned a small selected set.

Keep technical and commercial milestones on separate lines when necessary. The setting change can be executed and verified today. A later performance comparison needs defined periods, consistent outcomes and attention to other changes. The agency should not attribute every subsequent movement to the pause simply because it was the last visible operation.

The report can also explain a decision to retain an occurrence or seek missing evidence. Clients benefit from knowing what was reviewed and why no edit was made. A clear unresolved question is more valuable than a polished improvement claim unsupported by the available account facts.

Record ownership and review conditions

Leave the next specialist the named occurrence, business reason, evidence window, selected outcome and actual edit status. The handoff should also identify the client owner who confirmed the action and any website or creative work that remains separate. This prevents a setting receipt from being treated as completion of a larger campaign project.

Define the next review condition sensibly. A discontinued offer may stay paused until the client confirms its return. A retained keyword under investigation may need another completed observation window or a missing diagnostic read. Avoid a universal schedule detached from the reason the occurrence was reviewed.

Preserve disagreements explicitly. If the client chooses to retain an occurrence despite the agency's recommendation, record that decision and the agreed observation plan. The diagnostic should not become a covert route to apply an unapproved pause later. Current conversational edits still require the supported proposal and connected-owner confirmation.

Use the broader agency audit workflow for account-wide context and the Quality Score audit page for this check's focused explanation. Related negative-keyword review remains a separate targeting decision. These connections help the team navigate the work without claiming that one diagnostic covers every account concern.

Answer practical agency questions

Should every low-score keyword be paused?

No. Establish the offer, occurrence, performance evidence and owner decision first.

Does this list show component statuses?

No. The compact reader returns the overall score and parent names.

Are missing scores equivalent to zero?

No. Unknown diagnostics and numerical scores remain distinct.

Is the displayed account collection unlimited?

No. The current query returns at most 300 entries.

Does the bulk action have exactly the display scope?

No. Its current query has different parent filters and collection breadth.

Does a verified pause establish improved profitability?

No. It establishes the requested setting change, not a later commercial result.

Can the assistant draft ad ideas without reading the account again?

Yes. A supplied factual brief can support creative suggestions without a new provider read.