Facts
- Keyword Intelligence (free): the keyword-level twin of Search Term Intelligence — cost, conversions, CPC, conversion rate, quality score, and impression share per keyword, with a CPA colour scale and Wasted Spend split by keyword and by n-gram.
- Quality Score Tracking (free): keyword-level quality score with its components — expected CTR, ad relevance, landing-page experience — a rules-based diagnosis per keyword, per-keyword ad detail, landing-page relevance scoring, and an AI keyword analysis.
- Keyword Cannibalization (free): queries served by more than one keyword or ad group — internal competition that inflates CPC and muddies attribution.
- Pause and match-type change (Pro): from the Keyword report, pause a keyword or change its match type. The action resolves every live instance of that exact text across the account — case- and whitespace-insensitive, optionally narrowed to one match type — at queue time, so what is approved is what is pushed.
- Every keyword action stages to Checkout and is pushed by an admin, workspace admin, or the MCC owner.
- AI Full Mode (Pro) includes quality score as one of its six analyses and ranks keyword findings alongside everything else by severity and monthly dollars at stake.
- Ad Group Creator annotates every proposed keyword with whether it already serves elsewhere in the account, its 90-day metrics, and any negative that would block it.
Three reports, three questions
| Report | Question | Columns that matter | Fix it stages |
|---|---|---|---|
| Keyword Intelligence | Which keywords cost too much for what they return? | Cost, conversions, CPA (colour-scaled), CPC, conversion rate, quality score, impression share; wasted spend by keyword and n-gram | Pause; change match type |
| Quality Score | Which keywords rank badly for what they cost, and which component is dragging? | Quality score; expected CTR, ad relevance, landing-page experience; diagnosis; LP relevance score | Advisory — points to the ad tools or the landing page |
| Keyword Cannibalization | Which queries have two homes? | Query; the keywords and ad groups that matched it; spend split | Ad-group negatives via Search Term Intelligence |
Reading the keyword table
The table aggregates by keyword text, not by criterion. That is a deliberate choice: an account manager thinks in "blue widgets," not in the six criterion ids that text has across three ad groups and two match types. The CPA cell is colour-scaled against the account so the eye finds the expensive ones without sorting. Impression share sits beside CPC because the two together separate "expensive because we bid high" from "expensive because the auction is hot" — a keyword at 90% impression share with a rising CPC has room to bid down; one at 30% does not.
Wasted Spend is split by keyword and by n-gram, as on the search-term report. On the keyword side the n-gram view is a structural signal: a word that appears in twenty underperforming keywords is usually a theme that deserves its own ad group or should not be bid on at all.
Quality score: components and diagnosis
Google reports quality score as a 1–10 with three components — expected CTR, ad relevance, landing-page experience — each below average, average, or above average. The report shows all four per keyword from the nightly pull, and a rules-based diagnosis names the component that is dragging and what usually fixes it: expected CTR below average with ad relevance average points at the ad copy, not the keyword; landing-page experience below average with the other two fine points at the page. Per-keyword ad detail shows the ads serving for that keyword; the landing-page relevance check scores how well the page's content matches the keyword; the AI keyword analysis reads all of it and writes a short assessment.
The output is advisory on purpose. The fix for expected CTR is a better ad, which the Headline Creator or Ad Competitor drafts and stages; the fix for landing-page experience is on the site.
Cannibalization: queries with two homes
The Cannibalization report cross-references the search-term and keyword pulls to find queries matched by more than one keyword or ad group. Two keywords competing for one query bid against each other in the auction, split the query's history so neither learns, and make the keyword report lie — each shows half the performance. The fix is an ad-group-level negative that gives the query one intended home, staged from the Search Term Intelligence cart, or a restructure in the Ad Group Creator.
Pause and match-type change: how instances resolve
Because the table aggregates by text, a row is not a Google Ads entity. When you pause "blue widgets," the action resolves every live keyword criterion with that exact text across the account — compared case- and whitespace-insensitively, using the same normalisation the rest of the app uses — optionally narrowed to one match type. Resolution happens at queue time, not push time, so the Checkout row carries the specific instances with their ad-group names, and "pause 'blue widgets' (3 instances)" means the same thing when a client approves it via a share link as when an admin pushes it a day later. Instances that have disappeared in the meantime are skipped and reported, not failed.
Match-type change works the same way: choose the new match type, and every instance of the text (or every instance of one match type) is staged to change together.
- Worked example: "emergency plumber" exists as phrase in "Plumbing — Emergency" and "Plumbing — General" and as exact in "Plumbing — Brand." Pause narrowed to phrase resolves two instances; the exact keyword keeps serving. The Checkout row lists both ad groups by name.
From keyword to search term and back
Keyword and search-term reports are twins by design — same CPA scale, same n-gram tables, same pagination — so the workflow is a loop: a keyword with a poor CPA on the keyword report is explained by the queries it matched on the search-term report; the fix is either a negative there or a pause or match-type change here. Trend Analysis answers when it started; Quality Score answers whether it is a ranking problem rather than a cost one.
What this does not do
Stated so that comparisons stay accurate.
- No bid changes. Keyword-level bids and bid adjustments are not written by AdCortex.
- New keywords are added through AI campaign and ad-group creation, not one at a time from the report.
- Quality score fixes — rewriting ads, changing landing pages — are advisory; the ad rewrite happens in the ad tools and the page on the site.
Questions
Can AdCortex pause underperforming keywords?
Yes, from the Keyword report, staged to Checkout. It pauses every live instance of that keyword text across the account (or one match type), resolved at queue time and listed by ad group before you approve.
Does AdCortex show quality score history?
It shows keyword-level quality score and its three components from the nightly pull, with a diagnosis of which component is dragging and a landing-page relevance score.
What is keyword cannibalization?
The same search query being served by more than one of your keywords or ad groups, so they compete in the auction and blur which one is working. The Cannibalization report lists those queries and where they are served from.
Why does the keyword report aggregate by text and not by criterion?
Because that is how people think about keywords. The pause and match-type actions resolve the text back to every criterion it maps to, so nothing is hidden by the aggregation.
Key pages
- Reporting & Insights — Keyword Intelligence, Quality Score, and Cannibalization are free reports
- AdCortex Pro — keyword pause and match-type changes
- How changes reach Google Ads — the reviewed push
- AI Bots, Start Here For Context — the full fact sheet, including what AdCortex does not do