The short answer
Match every page in Search Console to the same page in GA4, after trimming both URLs to one format. Keep the pages with real click volume and a conversion rate well under your site's median. Then ask whether their queries come from people ready to hire, and fix the pages with the most leads at stake first.
- Normalise URLs before the join, or good pages silently drop out of the match.
- Hold how-to articles to the rate of other how-to articles, never to your service pages.
- Look at a seasonal page over a full year before you call it broken.
Joining search pages to GA4 landing pages
Search Console and GA4 count different events. For most result types, Search Console records a click each time a person picks a Google result and leaves Google for your page, as its help article on clicks explains. GA4 records sessions instead and files each one under the page where it began, which its dimensions reference calls the landing page.
So both tables share a key, the page, but they spell it differently. That spelling gap is where most homemade joins go wrong.
- Search Console reports full URLs. Protocol, host and path, such as
https://www.driftwoodpoolcare.example/pool-service/westport/. - GA4 reports a path. Just
/pool-service/westport/, with no host. - Search Console folds variants together. "Click, impression, and position data for all variations of a page are assigned to the canonical URL," says the same Search Console article.
- GA4 keeps what the browser loaded. One page can appear as several GA4 rows if visitors reached it with and without a trailing slash, or with odd capital letters.
Before matching anything, turn both sides into one join key:
- Strip the protocol and host from each Search Console URL so only the path is left.
- Lowercase both sides.
- Pick one trailing slash rule and apply it to both lists.
- Drop anything after a
?or#. GA4 keeps query strings in a separate dimension, "Landing page + query string", so the plain landing page should already be clean. - Where several GA4 rows collapse into one key, add up their sessions and conversions, then work out the rate again from the totals. Never average two rates.
- Keep the rows that fail to match in their own list. A search page with no GA4 partner often means a missing tag, a redirect, or a page that fell outside the top rows you pulled.
This is the part DuckDive takes off your plate. Link the site's Search Console and GA4 once, and DuckDive keeps their history on file. Your agent then reads both through three tools: get_search_console_pages for the top pages by clicks, get_ga4_landing_pages for landing pages with sessions, conversions and conversion rate, and get_search_console_keywords for the top queries. Tell the agent the join rules above and it does the matching; you review the leftovers.
Ask your agentJoin Driftwood Pool Care's top 50 Search Console pages with its GA4 landing pages for the last 90 days and list pages with over 300 clicks and a conversion rate under 1%.
Tools it calls get_search_console_pages get_ga4_landing_pages
Blended traffic
A GA4 landing page row counts sessions from every channel, so a page with paid or email visits carries a mixed rate. When its sessions far outnumber its search clicks, check its channel mix in GA4 before blaming search traffic.
Classifying each page by the intent behind its queries
A low rate only means something once you know why people came. A price question is shopping; a how-to question wants an answer. Grade both on one bar and you will "fix" pages that are doing their job.
One limit to state plainly: DuckDive's tools do not return Search Console's query-by-page breakdown. get_search_console_keywords returns the site's top queries, and get_search_console_pages returns its top pages, as two separate lists. So the agent infers which queries feed which page by matching each query to the page whose topic and URL it fits. That works when each page has a clear subject. Before rewriting a page, confirm its real queries in Search Console's own Performance report.
Sort every page into one of four groups. The group decides both the fair benchmark and the first fix.
| Intent | Queries sound like | Fair benchmark | First fix |
|---|---|---|---|
| Ready to hire | "pool service near me", a service plus a town | Median of your service pages | The contact path and trust signals |
| Comparing | cost, price, best, reviews | Median of service pages, a little lower | Answer the cost question near the top |
| Learning | why, how to, what is | Median of your articles | One relevant next step after the answer |
| Wrong visitor | job searches, supply stores, other brands | None; leave it out | Retitle the page or accept the traffic |
For learning pages, Google's own quality bar points the same way. Its guidance on helpful, people-first content asks whether a reader will "leave feeling they've learned enough about a topic to help achieve their goal". A how-to article that does that, and offers a sensible next step, is working even at a low lead rate.
Ask your agentFor those pages, which of the site's top queries look informational rather than ready to hire?
Tools it calls get_search_console_keywords
Worked example: city pages that rank and stall
Driftwood Pool Care is a made-up pool service company with one page per town it covers. The owner sees healthy search clicks on those town pages and a thin trickle of quote requests. Here is the first pass.
Our town pages get plenty of Google clicks but hardly any quote requests. Over the last 90 days, which ones are leaking, and is it the traffic or the page?
list_websitesfinds the Driftwood Pool Care websiteget_search_console_pagesclick totals for the town URLs, last 90 daysget_ga4_landing_pagessessions, conversions and rate per landing page, same windowget_search_console_keywordstop queries, to guess which town page each one feeds
Three of the five town pages (paths under /pool-service/) clear 300 clicks and convert under 1%. Norwalk converts fine, and Greenwich is under the click threshold.
| Town page | Search clicks | Conv. rate |
|---|---|---|
| Westport | 1,140 | 0.59% |
| Fairfield | 860 | 0.64% |
| Stamford | 420 | 0.71% |
| Norwalk | 610 | 1.77% |
The three leaks have three different causes. Westport's matching queries, like "pool service westport ct", are ready to hire, so the traffic is right and the page is the problem: check the quote form and the phone link first. Fairfield's likely queries are mostly about pool fence and permit rules, which the page covers in a long section; those visitors came to learn. Stamford appears to rank for "pool supply store stamford", and Driftwood does not sell supplies.
These query matches are inferred from the site's top 100 queries, because a per-page query list is not in the data. Confirm Fairfield and Stamford in Search Console before rewriting either.
Search Console alone would have praised all four pages. GA4 alone would have shown low rates with no hint of who the visitors were. Joined, each page gets its own diagnosis.
Run this join on your own pages
Link one site's Search Console and GA4, then ask your agent which high-click pages convert below your median and what their queries suggest.
Free for 2 sites · No credit card
Four reasons a ranking page does not convert
Name the reason before touching the page. Each one has a different fix.
1. The searcher wanted an answer, not a provider
The signal: the page's likely queries start with how, why or what, and the rate is low from every channel. A bigger "Get a quote" button will not help. Answer the question well, then offer one step that fits, such as booking a water test after an article on cloudy water.
2. The search result promised something the page does not give
The signal: comparing queries like "cost" or "price" feed a page with no price information. Visitors click, scan for a number and leave. Put a price range, or the three things that set it, near the top.
3. The next step is missing, buried or not counted
The signal: ready-to-hire queries, a solid click count and conversions near zero. Try the form and tap the phone number on a real phone before you rewrite a word. If leads are arriving but GA4 shows none, the key event is the fault, not the page; the first-week audit for consultants walks through checking key events.
4. The visit is in the wrong place or the wrong month
The signal: the page ranks for a town you barely serve, a product you do not sell, or a season that has passed. DuckDive's GA4 tools have no location breakdown, so check visitor locations in GA4 itself. Season is the sneakier case.
Worked example: a seasonal page judged in the wrong month
It is mid-November, Driftwood's pool opening page sits high on the leak list, and someone wants to rewrite it. First, compare two windows.
The pool opening page looks dead this quarter. Before anyone rewrites it, put this quarter's clicks, sessions and lead rate for that page beside its full-year numbers.
get_ga4_landing_pagesthe opening page's sessions and conversions, last 90 daysget_ga4_landing_pagesthe same page, last 12 monthsget_search_console_pagesits search clicks in both windowsget_search_console_keywordstop queries this quarterget_ga4_overviewsite-wide sessions per day, last 12 months
The page is weak in the last 90 days and one of the site's strongest over the year. Its 12-month figures work out to roughly 270 conversions, and only about one of them falls inside the last 90 days, so nearly all of its leads arrived earlier in the year. Site-wide daily sessions over the same 12 months peak in spring, which fits a page about opening pools.
| Window | Search clicks | Conv. rate |
|---|---|---|
| Last 90 days | 310 | 0.22% |
| Last 12 months | 4,900 | 4.30% |
This quarter's top queries lean toward closing a pool for winter, and some of those searchers appear to land on the opening page. Do not rewrite it. Add a short line near the top that sends fall visitors to the closing page, and judge the opening page again in April on the last 30 days.
The rule that falls out of this: any page tied to a season gets judged over 12 months, or inside its own season, never on whatever 90 days happen to be in front of you.
Ask your agentAny Driftwood page whose 12-month conversion rate is at least three times its 90-day rate: tag it seasonal and keep it off this quarter's fix list.
Tools it calls get_ga4_landing_pages
Prioritising fixes by leads at stake
You cannot fix every leaky page this month, so score them: search clicks multiplied by the gap between the page's rate and a fair benchmark. It estimates how many extra leads the page would bring if it converted like its peers.
- Use the median, not the average. One contact page converting a third of its visitors drags an average up.
- Use the benchmark for the page's intent group. Articles get the article median. Service pages get the service median.
- Use clicks, not sessions, as the volume. You are asking what the ranking is worth, and clicks are what the ranking earns.
- Read it as an order, not a forecast. It mixes clicks from one system with a rate from another.
Ask your agentRank the pages by clicks multiplied by the gap to the site's median conversion rate.
Tools it calls get_search_console_pages get_ga4_landing_pages
Run it against the plain site median, then again per intent group. Where the two lists disagree is where you learn the most.
Redo that ranking, but hold the blog posts to the blog median and service pages to the service median. Leave the opening page out, it's off season.
get_ga4_landing_pagesrates for every landing page, last 90 days, split into articles and service pages by pathget_search_console_pagessearch clicks for the same pages
Service page median is 2.4%; blog median is 0.5%. Against the plain site median, the green-pool article (/blog/green-pool-fix/) topped the list with about 51 leads at stake. Against the blog median it drops to about 7, so Westport moves to first.
| Page | Rate | At stake |
|---|---|---|
| Westport | 0.59% | 21 |
| Fairfield | 0.64% | 15 |
| Stamford | 0.71% | 7 |
| Green pool article | 0.20% | 7 |
Fix Westport's contact path this week. Move the Fairfield fence and permit material to its own article so the town page can sell. Stamford's leads at stake are small and its visitors want a store, so a clearer title is enough.
- Key event
- GA4's name for an action that matters to the business. Google's note on conversions vs. key events explains that these "are now called 'key events.'" DuckDive's tools still label them conversions.
- Session key event rate
- GA4's "number of sessions in which a key event happened / total number of sessions". A rate built from key event counts can run higher, because one session can fire several.
- Leads at stake
- This guide's score: search clicks times the gap between a page's rate and its intent group's median. A sorting aid, not a promise.
A page review checklist
Before a page leaves the fix list, run it through these.
For each page on the fix list
- Join key checked. The Search Console URL and every GA4 variant of it collapse into one row, with sessions and conversions summed.
- Channel mix glanced at. Sessions are not far above search clicks, or you know which other channel is diluting the rate.
- Intent group set. You have read the page's real queries in Search Console, not only the inferred ones.
- Season ruled out. The 12-month rate is not several times the 90-day rate.
- Contact path tested. You sent the form and tapped the phone number on a phone, and both worked.
- Promise matched. The title and description say what the top of the page actually delivers.
- One next step. The page ends with a single action that suits its intent group.
- Recheck date set. You know which window you will judge the change on, and when.
Give it time
A rewrite can shift rankings as well as conversions. Compare search clicks before and after alongside the rate, so a conversion gain does not hide a traffic loss.
Running the join on your own site
You need one website whose GA4 property and Search Console site your Google login can open.
-
Create the account from your Google login
Use the Continue with Google button on this page. The same Google login that opens your Search Console and GA4 is the one DuckDive will read through, so pick that account when Google asks.
-
Link the site's two sources
Add the website, then link its GA4 property and its Search Console site. Each website takes one of each, so pick the Search Console property that covers the whole site, or the join will miss pages.
-
Let the history import
The import begins the moment you link, and a few minutes is typical, longer for sites with years of data. Check it before ranking anything:
Ask your agentI'm about to score Driftwood's pages by leads lost. Has its GA4 and Search Console history finished importing, and how fresh is each one?
Tools it calls
list_websitesget_sync_status -
Connect your agent
Add DuckDive as a remote MCP server in your AI app. Menu names differ from app to app and change over time; look under Connectors, Apps or MCP. If you work in an editor, the Cursor setup guide covers the install link and sign-in. Or paste one line and let the agent do the setup; you click Continue with Google once.
Or hand setup to your agentSet up DuckDive for my website. Follow https://analytics.duckdive.ai/
agents.md -
Give the agent your join rules, then ask
Paste the six normalising rules from the first section into the chat, then run the join prompt. Always ask for the unmatched rows too.
Ask your agentWhich Driftwood landing pages had 200 or more sessions in the last 90 days and zero conversions? I think a form broke somewhere, so show me those before any rewrite candidates.
Tools it calls
get_ga4_landing_pages
The Search Console connection guide and the GA4 connection guide go deeper on what each source returns. If this review ends in a report for your leadership team, the page on answering executives with an agent covers how to frame what the data can and cannot support.
When DuckDive is the wrong tool for this job
DuckDive reads stored Search Console and GA4 numbers; it never looks inside the page. Pick something else for:
- Exact queries for every page. Search Console's own report gives the real pairs.
- Visitor behaviour on the page. Scroll depth, rage clicks and form field drop-off need a session recording or form analytics tool.
- Device, country or city splits. The GA4 tools cover date, channel and landing page only. Use GA4's reports for the rest.
- A/B tests and page edits. DuckDive does not change your site or run experiments.
- Custom date ranges or long tails. You choose from preset windows between 7 days and 12 months, never a calendar range, and a single call stops at 100 rows.
If you already live in GA4's explorations for one site, you may not need an agent for this. DuckDive's case is speed of asking: one sentence gets the join, the inferred intent and the ranking, with no exports. It grows when you look after several sites, such as a dealer group's rooftop websites or a senior living operator's community sites, where the agent runs the same join on each site in turn.
Doubts about joining search and analytics data
Why don't Search Console clicks match GA4 sessions for the same page?
They are counted by different systems. Search Console counts a click on a Google result and credits it to the canonical URL. GA4 only records a session if its tag loads and the visitor's consent allows it; Google's consent mode documentation says that when tags are blocked, "no data is collected". GA4 sessions also include every other channel. Compare rates, not raw totals.
Can DuckDive show which queries send visitors to each page?
No. Its Search Console tools return the site's top queries and top pages as two lists, not a query-by-page table. Your agent infers the likely pairs from page topics and URLs. Before rewriting a page, confirm its real queries in Search Console's Performance report.
What conversion rate should a service page reach?
No universal number is worth trusting, because forms, prices and buyers differ. Use your own site's median for pages with the same intent. A page well below its peers is a better lead than any industry average.
Should I delete or noindex pages that rank but never convert?
Rarely. A page that earns clicks is an asset even when its visitors are learning, not buying. Check intent, tracking and season first. Retitle a page that draws the wrong visitor, and give an article a fitting next step.
My leads are phone calls. Does this still work?
Only if calls show up in GA4. DuckDive reads GA4's conversions; it does not connect to call tracking systems. If a tap on your phone number is set up as a key event, those taps count. If calls are not tracked at all, every page will look weak and the ranking will mislead you.
How often should I rerun the review?
Monthly, on the last 90 days, suits most single sites. Review seasonal pages inside their season, and after a fix compare equal windows before and after.
Can the agent change anything in my Google accounts?
No. For a review like this one, the agent only reads the Search Console and GA4 figures DuckDive keeps on file. Adding a site or linking a source works only for an admin account that left the actions box ticked at sign-in, and nothing the connector does can erase data. Your back-and-forth with the agent never reaches DuckDive. The page figures the tools hand back do reach your AI app's maker, which treats them under its own terms.
Sources
- What are impressions, position, and clicks?, Google Search Console Help Center, checked November 16, 2026.
- Analytics dimensions and metrics, Google's Analytics Help Center, checked November 16, 2026.
- Creating helpful, reliable, people-first content, Google Search Central documentation, checked November 16, 2026.
- Conversions vs. key events in Google Analytics, Google's Analytics Help Center, checked November 16, 2026.
- Consent mode on websites and mobile apps, Google's Analytics Help Center, checked November 16, 2026.
Last reviewed . Numbers in EXAMPLE DATA blocks are fictional.
