The short answer
Diagnose a landing page as a chain, not a single rate:
Delivery: Did the right ad reach the right audience?
Click: Did the ad earn relevant traffic at a sensible cost?
Arrival: Did paid visitors reach a measurable landing page?
Engagement: Did the page hold attention and support the intended task?
Outcome: Did the visit produce the key event or business result that matters?
Google Ads is strongest at the first two layers. GA4 adds evidence about arrival, engagement, and on-site outcomes. When you preserve that boundary, AI can help you classify a weak result as a traffic-quality problem, page-experience problem, message mismatch, conversion-path problem, tracking issue, or simply insufficient evidence.
The goal is not to find the page with the lowest conversion rate and declare it broken. The goal is to identify the most plausible constraint, show the evidence for and against it, and define the next human test.
Why Google Ads and GA4 answer different questions
The Google Ads landing pages report starts with an ad click and the URL associated with it. Depending on campaign type and report, it can show delivery metrics, expanded landing-page URLs, and mobile-friendliness information. Google Ads also holds the campaign, ad group, ad, keyword, search term, device, cost, and advertising-conversion context.
GA4 defines a landing page as the first pageview in a session. Its landing-page reporting can pair that page with session-scoped traffic dimensions and metrics such as sessions, engaged sessions, engagement rate, average engagement time, key events, and revenue.
Those are related views of a journey, not interchangeable datasets.
Question | Best starting evidence | Important limitation |
|---|---|---|
Which paid traffic did we buy? | Google Ads impressions, clicks, CTR, cost, CPC, campaign, ad group, search term, and device | A click does not prove a measurable session or useful visit |
Which page started the visit? | GA4 landing page plus query string | It is session-scoped and depends on Analytics collection |
Did visitors engage? | GA4 engaged sessions, engagement rate, and average engagement time | Engagement does not prove persuasion or business value |
Did the visit produce the intended action? | GA4 key events and Google Ads conversions | Definitions, scope, counting, attribution, and processing may differ |
Is the page technically fast and usable? | Browser testing and PageSpeed Insights | Platform reports can suggest a problem but cannot reproduce every user experience |
Expect clicks and sessions to differ. A visitor can click more than once before one session, leave before Analytics loads, reject consent, encounter a redirect that loses parameters, or return in a separate session. Time zones, date boundaries, filters, modeling, and processing can create additional differences.
That means a low session-to-click ratio is a signal to investigate – not a verdict that the page failed.
Lock the comparison before diagnosing anything
Write down the comparison contract before pulling data. At minimum, specify:
the exact Google Ads account and GA4 property
one complete date range and one equal comparison period
the campaign or campaigns in scope
the intended business outcome and its exact conversion or key-event name
the landing-page URL rule, including whether query parameters matter
the traffic scope, such as Google Ads sessions only
the device split you will preserve
the minimum evidence required before assigning a diagnosis
Use a mature period. For an initial review, the last 28 complete days ending yesterday is often a practical default. Extend the window for low-volume accounts, long conversion cycles, or infrequent high-value outcomes. Keep the previous equal period as context, but do not let a percentage change from tiny counts drive the queue.
Do not combine pages merely because their paths look similar. Query parameters may be disposable tracking noise, or they may select a materially different offer, location, product, or experiment. Normalize only after documenting the rule.
A reliable landing-page diagnosis workflow
1. Confirm the measurement handoff
Verify that the intended Google Ads account and GA4 property are connected, auto-tagging is enabled where appropriate, and the campaigns in scope actually send traffic to the pages you plan to evaluate.
Then check the outcome definition on both sides. A Google Ads lead conversion and a GA4 form_submit event may represent different points in the funnel. If the definitions differ, record that before comparing rates.
Do not proceed as if configuration proves collection. A platform can show the correct setting while a redirect, consent state, tag condition, form behavior, or browser error prevents the signal from arriving.
2. Pull Google Ads traffic and cost evidence
Start with campaign, ad-group, device, and other useful delivery cuts for the selected period. Keep impressions, clicks, CTR, cost, average CPC, and the relevant conversion metrics visible.
For exact URL-level Google Ads evidence, use the landing-page and expanded-landing-page reporting resources. In HireOtto, that exact custom Google Ads query may require an eligible Agency account. Do not claim a core report contains final URLs when it does not.
If exact URL-level Ads data is unavailable, keep the analysis honest: use the available campaign, ad-group, ad, and device evidence, pair it with the current destination mapping you can verify, and label page-level cost allocation as unavailable rather than estimating it.
3. Check GA4 metadata and compatibility
Before requesting the report, confirm the property-supported API names and whether the proposed dimensions and metrics are compatible.
A useful starting request is:
dimension: landing page plus query string
a Google Ads session dimension or another paid-traffic scope supported by the property
sessions and engaged sessions
engagement rate and average engagement time per session
the relevant key-event count or rate
revenue when the business outcome supports it
GA4 Core reports reject incompatible combinations. Realtime reporting also uses a different field set. Let the property metadata and compatibility result decide the final query rather than forcing a memorized template.
4. Pull the GA4 landing-page evidence
Run the compatible landing-page report for the same complete date range. Filter to the intended Google Ads traffic where possible, exclude blank landing pages from the ranked analysis, and keep them in a separate data-quality check.
Retrieve enough rows to cover the account, and make truncation visible. Record the requested rows, matching rows, collected rows, filters, and sort order. A neat top-50 table is not a complete audit if the long tail contains meaningful spend or outcomes.
Preserve device where volume permits. A blended result can hide a page that works on desktop and breaks on mobile.
5. Normalize URLs without erasing meaning
Create a reviewable URL key. A safe first pass can:
standardize host casing
remove fragments
handle trailing slashes consistently
strip known click and analytics parameters such as gclid and UTMs
preserve parameters that change the offer, page content, location, product, or experiment
Keep both the original URL and normalized key. When two URLs are merged, document why. Never let AI silently collapse distinct pages because their strings look similar.
Redirects deserve their own check. The advertiser-specified URL, expanded final URL, and GA4 landing page may differ legitimately. They may also expose a redirect chain, cross-domain handoff, or parameter-loss problem.
6. Join only the evidence that can be joined
If exact Google Ads URL reporting is available, join its normalized expanded URL to the normalized GA4 landing-page key. Retain source-specific columns so nobody mistakes clicks for sessions or Google Ads conversions for GA4 key events.
If exact Ads URLs are not available, diagnose at the narrowest shared level supported by the evidence – often campaign, ad group, or device – and treat the GA4 page ranking as a related view, not a mathematically exact cost allocation.
For each reliable page row, calculate or retain:
Google Ads clicks, cost, average CPC, and conversions when available at that grain
GA4 sessions, engaged sessions, engagement rate, and average engagement time
GA4 key events, session key-event rate, and revenue when compatible
click-to-session ratio as a diagnostic signal, not a quality score
change versus the equal comparison period
a volume or confidence label
Do not invent a single composite “landing page score.” It hides the reason a page needs attention and implies precision the data does not support.
7. Classify the evidence pattern
Use the combination of signals to select the next check.
Evidence pattern | Most plausible explanations | Next human check |
|---|---|---|
Strong CTR and efficient clicks, but weak GA4 engagement across devices | Message mismatch, slow or unstable arrival, weak first screen, or unwanted traffic | Review search terms and ad promise; test the page and load path |
Desktop is healthy, mobile is weak | Mobile layout, speed, tap targets, form friction, or device-specific tracking | Reproduce the journey on real mobile devices and run PageSpeed Insights |
GA4 sessions are materially lower than Ads clicks | Consent, Analytics loading, redirect, tagging, filters, repeat clicks, or fast exits | Inspect the discrepancy by day and device; validate with browser tools |
Engagement is healthy, but key-event rate is weak | Offer, CTA, form, eligibility, checkout friction, or missing event collection | Complete the real conversion path and verify each step |
GA4 outcomes look healthy, but Google Ads conversions are weak | Import, action status, goal scope, attribution, counting, or reporting-window difference | Reconcile the exact event and conversion action before changing the page |
One URL variant underperforms while peers are stable | Broken parameter, experiment, redirect, or variant-specific content | Compare the rendered variants and their routing rules |
Volume is low or the period is immature | Insufficient evidence or conversion lag | Extend the period or wait; do not manufacture a diagnosis |
Classification is not proof. Each row should state whether the explanation is confirmed, plausible, ruled out, or unverified.
8. Turn the diagnosis into a review queue
Prioritize issues by expected business impact, evidence strength, and reversibility – not by the most dramatic percentage.
A useful queue separates four kinds of work:
Measurement: missing sessions, duplicate events, broken parameters, or mismatched outcome definitions
Traffic: irrelevant queries, weak ad-to-page intent alignment, or expensive low-quality segments
Experience: mobile usability, speed, first-screen clarity, navigation, or form friction
Experiment: a specific copy, layout, offer, or CTA hypothesis with one primary success metric
Each item should name the page or segment, summarize the evidence, show what remains uncertain, assign an owner, and define the review gate. Measurement problems should usually be resolved before creative tests; otherwise the experiment may optimize against unreliable feedback.
The exact prompt
Copyable prompt
Connect to my Google Ads and GA4 data. First list the accounts and GA4 properties I can access, and wait for me to select the intended account and property. Do not make changes.
Diagnose landing-page performance for the last 28 complete days ending yesterday, with the previous 28 complete days as context.
Confirm the selected Google Ads account, GA4 property, property time zone, Google Ads link, and auto-tagging status. Ask me for the exact business outcome and conversion or key-event name if I have not supplied them.
Pull Google Ads traffic and cost evidence by campaign, ad group, and device. If exact URL-level reporting is available, use landing-page or expanded-landing-page rows. If it is not available, state the limitation and do not estimate page-level cost.
In GA4, inspect the reporting metadata and check compatibility for landing page plus query string, an appropriate Google Ads session dimension, sessions, engaged sessions, engagement rate, average engagement time, the relevant key-event metric, and revenue where applicable. Use only compatible fields.
Run the GA4 landing-page report for the same period and traffic scope. Keep requested rows, matching rows, collected rows, filters, and truncation status visible. Report blank landing pages separately.
Preserve original URLs. Propose a normalization rule, strip only known tracking parameters, and do not merge parameters that may change content or intent.
Join the platforms only at a grain supported by both datasets. Keep clicks and sessions, Ads conversions and GA4 key events, and all source-specific metrics in separate columns.
Segment by device when volume is sufficient. Flag low-volume rows rather than over-interpreting rates.
Classify each issue as traffic, arrival or measurement, page experience, conversion path, Ads-to-GA4 reconciliation, or insufficient evidence. Mark each explanation confirmed, plausible, ruled out, or unverified.
Produce a prioritized review queue with the evidence, business impact, next human check, owner, and review gate. Do not recommend a page change when the strongest evidence points to traffic or tracking.How HireOtto supports the workflow
HireOtto can bring the Google Ads and GA4 evidence into the AI client where you already work, then turn it into a review-ready diagnosis.
On the Google Ads side, HireOtto can retrieve core performance reports at campaign, ad-group, keyword, search-term, ad, geography, device, and conversion levels. Eligible Agency accounts can use custom GAQL when the workflow needs Google Ads’ exact landing-page or expanded-landing-page resources.
On the GA4 side, HireOtto can discover properties, inspect configuration, search the property’s metadata, check Core-report compatibility, run filtered landing-page reports, and export rows. The GA4 connection is read-only.
HireOtto can help normalize the evidence, identify patterns, and build the review queue. It cannot reproduce the page in every browser, test every consent state, prove a tag fired, submit the form as a real prospect, edit the landing page, or modify GA4 configuration. Use a human review for URL-merging decisions and business-outcome definitions, and browser-test the important mobile and conversion journeys before changing spend or page design.
For the supported workflows, use the Google Ads reporting guide, GA4 custom reporting guide, and GA4 tools reference. To connect a new AI client, follow Connect a HireOtto server.
What a useful diagnosis should produce
The final output should be an evidence packet, not a leaderboard of “good” and “bad” pages.
Output field | What it should contain |
|---|---|
Scope | Account, property, dates, traffic scope, outcome, device, and URL-normalization rule |
Source evidence | Separate Google Ads and GA4 metrics with their original grain |
Diagnosis | Confirmed or plausible constraint, plus evidence against the conclusion |
Confidence | High, medium, low, or insufficient evidence, with a volume note |
Next check | One specific browser, traffic, tracking, or page review |
Action | Owner, priority, expected impact, and approval or validation gate |
That packet lets a paid media operator fix traffic, an analyst inspect measurement, a designer review the page, and a developer test one specific journey. It also makes the audit repeatable after the change.
Frequently asked questions
Should Google Ads clicks equal GA4 sessions?
No. Clicks count ad interactions; sessions measure Analytics activity. Repeat clicks, consent, redirects, page exits before collection, filters, time zones, and processing can all create differences. Investigate a material or sudden change, but do not use equality as the health criterion.
Is a low engagement rate proof that the landing page is bad?
No. It can reflect a poor page experience, but it can also reflect irrelevant traffic, a mismatch between the ad promise and page, measurement gaps, or the nature of a fast single-purpose task. Compare segments and validate the journey before choosing a page fix.
Should I use landing page or page path in GA4?
Use the landing-page dimension when the question is about the first page of a session. Use page-path or page-location dimensions when analyzing views and behavior across pages regardless of where the session started. They answer different questions.
How much data is enough?
There is no universal minimum. Require enough volume to make the intended decision at the chosen grain. A page with eight sessions should not outrank a high-spend page because one conversion changed its rate dramatically. Label thin evidence and extend the period when necessary.
Can AI decide which landing-page change to ship?
AI can rank evidence, suggest hypotheses, and prepare a test plan. A practitioner still needs to confirm the business outcome, inspect the actual page and traffic, choose the hypothesis, approve the implementation, and judge the result against a predeclared metric.
About Me
I’m Suyash – badminton junkie, ex‑GroupM ad‑ops grunt, first marketer at a B2B SaaS startup, and creator of Hiretto: MCP servers for performance marketers.
My mission: less clicking, more thinking.
Let’s build leverage together.