How to audit paid landing pages with an AI agent
Which landing pages waste paid clicks: speed and status from Google Ads' own landing page view, joined to GA4 engagement and conversion by page.
IntentRank paid landing pages by the money they lose, with the cause class attached: broken, slow, or unconvincing.
- 4
- Steps
- 2
- Tools
- 3
- Rules
- 3
- Failure modes
When to use it
Use it when
- Paid CTR healthy but conversions poor: the click survives, the page kills it.
- After site releases touching templates that paid traffic lands on.
- Quality Score complaints citing landing page experience.
Do not use it when
- Single-landing-page accounts: read the funnel instead.
- As a performance profiler: this locates pages, it does not diagnose their code.
What it needs first
Required
Optional
Preconditions
- Paid traffic distributed across multiple destinations.
- GA4 connected for the engagement join, ideally.
Inputs
window- Analysis window.Default: last 30 days.
Procedure
In order. Every tool named is one the gateway ships, and links to its reference.
Pull the landing page view: URLs, clicks, mobile speed, valid status.
Join GA4 engagement and conversions by landing page.
Rank pages by paid clicks x (account median conversion rate - page conversion rate).
Classify each loser: technical (status, speed), engagement (instant exits), or persuasion (engaged, no conversion).
Decision rules
The observable condition, what it lets you conclude, and what takes the conclusion back.
When
A page takes meaningful paid clicks and returns errors or redirects
Conclude
Technical, severity one: every click on it is bought and lost; quote the click count.
When
Engagement time is near zero on a healthy-status page
Conclude
Mismatch between ad promise and page, or speed: pair the page with the ads pointing at it.
When
Engagement is normal but conversion trails the account median
Conclude
Persuasion class: the page argues badly; a testing candidate, not an emergency.
Evidence every conclusion must carry
- Every ranked page carries clicks, spend share, its rates and the account median as baseline.
- Technical findings quote the landing page view's own status fields.
Where the agent stops
- Page fixes are product and CRO work; pausing ads onto a broken page is proposed as a Safe Write with the loss rate as justification.
How it goes wrong quietly
The cases where the analysis is wrong and still looks right. Read them before trusting a number.
- URL parameters split one page into many rows; normalise before ranking or the top loser hides in fragments.
- GA4 landing-page attribution differs from Google Ads' final URL on redirects; state the join rate.
- A/B tests split performance across variants mid-window; note running tests before classifying persuasion losses.
What the answer contains
- 01Losers, ranked
- Page, clicks, loss estimate, cause class.
- 02Emergency list
- Broken destinations still taking spend.
- 03Testing candidates
- Persuasion-class pages worth a CRO cycle.