The quick answer
Performance Max search term analysis should answer three questions: Which queries are creating value? Which are consuming meaningful spend without value? And what do those patterns say about your search themes, creative, feed, and landing pages?
HireOtto can pull a dedicated PMax search terms report into your AI assistant. The current report is campaign-level and includes the search term, match type, impressions, clicks, CTR, average CPC, cost, conversions, cost per conversion, and conversion value. You can filter it to one campaign, sort it, export it to CSV, or segment it by day.
The important caveat: this is not a complete log of every query. Google withholds some low-volume queries for privacy. Use the report for evidence, not false precision, and use Google’s Search terms insights alongside it when you need category-level context.
A reliable workflow is simple:
Pull enough data to see patterns.
Let AI classify terms by intent and relevance.
Separate proven waste from low-data uncertainty.
Review search themes, assets, feed, and landing pages before blocking traffic.
Approve any negative keywords yourself.
What the PMax search terms report tells you
The report shows the actual search terms associated with Performance Max activity and their campaign-level performance. That is useful for four jobs:
Waste control: find irrelevant or low-value queries consuming meaningful budget.
Demand discovery: spot relevant language, use cases, and objections that convert.
Message alignment: compare query intent with your headlines and landing page promise.
Campaign steering: decide whether search themes, exclusions, feed data, or creative need attention.
Do not read a PMax search term row as if it were a Search keyword row. Performance Max has no standard ad-group and keyword structure. Google aggregates PMax search term performance at the campaign level, so the dedicated HireOtto report does not attribute a query to a specific asset group, keyword, or ad.
Search terms report vs. Search terms insights
Google exposes two related views, and they solve different problems:
Search terms report: individual queries and their performance. Use this for concrete term review and negative-keyword candidates.
Search terms insights: Google-generated categories and subcategories that include aggregated context for queries that may not appear individually because of privacy thresholds. Use this for market themes and demand patterns.
HireOtto’s current dedicated PMax search terms report returns individual campaign-level terms and metrics. It does not currently return Google’s insight categories or subcategories. If you need that broader view, check Search terms insights in Google Ads and compare the categories with the term-level HireOtto analysis.
The five-pass AI workflow
1. Start with the right window and output
For a quick review, pull the last 30 days and sort by cost. For lower-volume accounts or high-consideration sales cycles, use 60–90 days if seasonality and recent changes make that window sensible. Avoid mixing a major pricing, targeting, feed, or conversion-tracking change into one undifferentiated period.
Prompt:
Pull the Performance Max search terms report for the last 30 days.
Filter to campaign [CAMPAIGN NAME OR ID]. Sort by cost descending.
Return the top 50 rows in chat and include the full CSV export.
Do not recommend changes yet.HireOtto shows 50 rows inline by default. A summary-and-CSV request gives you the readable sample plus a fuller export; the CSV export defaults to up to 5,000 rows, and its signed link expires after 30 minutes by default.
2. Classify intent before judging performance
Raw query lists are messy. Have the AI assign each term to a small, business-specific taxonomy such as:
Core commercial: directly describes the product or service.
Problem-aware: expresses the pain you solve but not the solution.
Comparison or alternative: evaluates vendors or approaches.
Brand: your brand, competitor brands, or unrelated brands.
Informational: research-led intent that may or may not assist a later conversion.
Irrelevant: no credible path to the offer.
Prompt:
Classify every PMax search term by intent and business relevance for [WHAT WE SELL] to [TARGET CUSTOMER].
Add columns for intent class, relevance (high/medium/low), evidence, and confidence.
Do not label a term as waste from wording alone; use cost and conversion evidence too.3. Build three decision queues
Ask the AI to create three queues rather than one giant list of “good” and “bad” terms.
Queue A: investigate or exclude
Prioritize terms that are clearly irrelevant and have meaningful cost, or relevant-looking terms that repeatedly spend beyond your acceptable economics without primary conversions. A useful threshold can reference your target CPA – for example, zero-conversion terms above one target CPA deserve review – but it is not an automatic exclusion rule.
Queue B: expand and reinforce
Look for terms with strong conversion volume, efficient CPA, or high conversion value that reveal language missing from your assets, landing pages, product feed, or search themes. The next action may be better copy or a dedicated page, not merely adding another search theme.
Queue C: observe
Put low-volume, ambiguous, recently emerging, or strategically relevant terms with thin evidence here. Monitoring is a decision. It is usually safer than blocking uncertain demand.
Prompt:
Create three queues: investigate/exclude, expand/reinforce, and observe.
Rank each term by expected business impact.
For every recommendation, cite cost, conversions, conversion value, CPA, relevance, and data sufficiency.
Flag anything that requires checking conversion quality or the landing page before action.4. Diagnose the source of the pattern
A poor term does not tell you which control caused it. Before changing the campaign, inspect the likely source:
Search themes: Are optional themes too broad, outdated, or missing a high-value concept?
Final URL expansion and landing pages: Is Google finding pages that imply a different offer or intent?
Merchant Center feed: Are titles, descriptions, product types, or attributes attracting the wrong demand?
Assets: Do headlines and descriptions overpromise, omit a qualifier, or fail to match the best queries?
Conversion setup: Are weak actions counted as primary conversions, making poor traffic look successful?
Search themes are signals, not keywords. Google says they are optional and additive to the queries Performance Max predicts from your assets, feeds, and landing pages. You can add up to 50 search themes per asset group, but more themes are not automatically better.
Prompt:
For each high-impact query pattern, propose the most likely control to inspect: search themes, landing pages/final URL expansion, feed data, assets, exclusions, or conversion setup.
Explain what evidence would confirm or reject the diagnosis.
Do not make changes.5. Turn recommendations into reviewed actions
Use negatives for clear exclusions, not as a reflex for every weak row. Google supports account - or campaign-level negative keywords for PMax Search and Shopping inventory. Those exclusions do not control Display or video inventory; Google directs advertisers to content-suitability exclusions for those channels.
HireOtto can add campaign-level negative keywords with broad, phrase, or exact match types. Because this is a live account change, first ask for the proposed terms and match types, review them, and approve the final set.
Review prompt:
Draft a proposed campaign-level negative keyword set from the approved investigate/exclude queue.
Show keyword text, match type, evidence, expected benefit, and overblocking risk.
Include close variants, singular/plural forms, and synonyms only where needed because negative keywords do not behave like positive keywords.
Do not apply anything.Execution prompt after approval:
Add only the approved campaign-level negative keywords below to PMax campaign [CAMPAIGN ID].
Use the match type shown for each term. Return the number added and any errors.
Do not change search themes, assets, URLs, budget, bidding, or conversion settings.A worked example
Suppose a B2B payroll platform sees these patterns in one PMax campaign:
“payroll software for 20 employees” converts efficiently and shows repeat demand.
“free payroll spreadsheet template” has high spend, no primary conversions, and is outside the product’s acquisition strategy.
“how to calculate payroll tax” has low spend and assists some later conversions, but the direct report alone cannot prove the assist.
The right response is not “exclude every informational term.” Instead:
Reinforce the first theme in creative and consider a use-case landing page.
Review “free payroll spreadsheet template” as an exact or phrase negative only after confirming it does not contribute qualified leads elsewhere.
Keep the tax-calculation query in the observe queue, compare a longer window, and review Search terms insights and assisted-conversion evidence.
This is where AI is most useful: sorting and explaining the evidence quickly while leaving the irreversible judgment with the operator.
How to compare two periods without fooling yourself
A single snapshot can hide a trend. Pull two complete, comparable windows – such as the previous 30 days and the 30 days before that – then ask the AI to compare normalized metrics.
Prompt:
Compare these two PMax search term exports.
Group identical terms, then summarize new terms, disappearing terms, cost changes, conversion changes, CPA changes, and conversion-value changes.
Separate genuine movement from low-volume noise.
Note any campaign, tracking, budget, bid, feed, or landing-page change that could break the comparison.HireOtto can also segment the dedicated report by day. Use that when timing matters, but expect a much larger dataset and export it rather than flooding the conversation.
Common mistakes
Blocking terms from wording alone
A strange-looking query can still convert. Check business relevance, conversion quality, and enough data before excluding it.
Using CTR as the verdict
Google advises prioritizing conversion data when evaluating PMax. Because PMax spans multiple channels and formats, click- and impression-based rates can be misleading when isolated from value.
Assuming the report contains every query
Some low-volume queries are omitted for privacy. Search terms insights may retain their impact inside aggregated categories even when the individual term is hidden.
Treating campaign-level data as asset-group attribution
The current HireOtto report is campaign-level. It cannot prove which asset group, ad, or landing page caused a term’s performance.
Adding broad negatives too quickly
Broad negative keywords can remove more demand than intended. Start with the narrowest justified match type, review variants, and monitor after the change.
Changing several controls at once
If you change themes, assets, feed titles, URLs, and negatives together, you lose the ability to learn what worked. Sequence changes and keep a change log.
Where HireOtto fits
HireOtto lets you run this workflow inside your AI assistant instead of manually exporting and reorganizing the report. It can retrieve campaign-level PMax search terms with impressions, clicks, CTR, average CPC, cost, conversions, CPA, and conversion value; return an inline summary or CSV; and help classify the visible queries into exclude, expand, and observe queues.
Once you approve the recommendations, HireOtto can add selected terms as campaign-level PMax negatives. Keep the boundary clear: let the AI retrieve the data, calculate the metrics, and organize the evidence – but let the practitioner judge intent and approve every exclusion. HireOtto’s current report does not provide asset-group attribution or Google’s aggregated Search terms insight categories, so use the Google Ads interface when that additional context is needed.
Frequently asked questions
Can I see every Performance Max search term?
No. Google omits some individual queries for privacy and low-volume reasons. Use the term-level report for visible queries and Search terms insights for aggregated categories and subcategories.
Can I see which asset group generated a search term?
Not in HireOtto’s current dedicated PMax search terms report. It aggregates performance at campaign level. Google’s Search terms insights can show asset-group context for categories, but that is a different report.
Does HireOtto return landing pages or ad formats with each term?
Not in the current dedicated report. Google’s interface now offers a “Search terms and landing pages for Performance Max” view with landing-page and ad-format detail, but HireOtto’s report currently returns the term, campaign, match type, and performance metrics only.
Can AI add PMax negative keywords for me?
HireOtto can add campaign-level negative keywords after you approve the terms and match types. For PMax, Google says campaign- and account-level negatives apply to Search and Shopping inventory, not Display or video inventory.
Should winning search terms become search themes?
Sometimes. Add a theme when it represents strategically relevant demand that your existing signals may not express. If the campaign is already reaching and converting that demand, the better next step may be stronger creative, feed data, or a landing page – not another signal.
The operating principle
Do not use AI to generate a longer negative list. Use it to produce a better decision trail: what happened, why it may have happened, what evidence supports action, what could go wrong, and what should be monitored next.
That is the useful combination: Google Ads supplies the performance data, HireOtto moves it into your AI assistant, and the practitioner keeps control of the account.
FAQs
What is a good Google Ads impression share?
A good level is one that matches the campaign’s role and economics. Brand campaigns may justify aggressive visibility. Exploratory non-brand campaigns may not. Compare against your own history, profitability, and strategic importance rather than a universal percentage.
Should I increase budget when lost impression share from budget is high?
Only if the campaign is producing acceptable incremental value. High lost share from budget identifies available visibility; it does not prove that buying it will be profitable.
Review relevance, ad quality, landing-page experience, assets, query intent, and bidding. Raise bids only when the expected incremental value supports the added cost.
What is the difference between impression share and top impression rate?
Impression share compares received impressions with estimated eligible impressions. Top impression rate measures the portion of received impressions that appeared among top ads.
You can pull the campaigns across an account in one request, but Google reports impression share separately by campaign type rather than as one meaningful account-wide aggregate. Analyze campaign rows and group them by comparable roles.
Does HireOtto return exact-match impression share?
Not in the current dedicated impression-share report. It returns click share, overall/top/absolute-top impression share, and the related budget- and rank-loss metrics. Exact-match impression share would require a separate supported report path.
About Me
I’m Suyash – badminton junkie, ex‑GroupM ad‑ops grunt, first marketer at a B2B SaaS startup, and creator of Hiretto: Google Ads MCP Server.
My mission: less clicking, more thinking.
Let’s build leverage together.