# How to tell if a topical map is working after you publish

You published the first pages of your map and now you are refreshing Search Console. Here is what normally happens in the first days, weeks and months, why positions jump around, the six numbers that show a map working, and when to decide it is not.

By Kevin Jeppesen, AI growth engineer. Founder, FrameworkSEO & Ascendily. Updated 2026-11-13. Source: https://frameworkseo.com/learn/topical-maps/measure-a-topical-map

**Measuring a topical map:** Measuring a topical map means checking, cluster by cluster and month by month, whether its pages get indexed, shown, chosen and paid for, against the plan the map set out.

## Key takeaways

- Google says crawling takes a few days to a few weeks and site-wide improvements can take several months to register, so judge a new cluster on indexing and impressions first and on clicks last.
- An average position that gets worse in the first weeks is usually a sign the map is working: Search Console averages over every search a page shows for, and new searches arrive low before they climb.
- No topical authority checker can read how Google rates your site on a subject; the score that matters is the share of your own map's searches you show for, which you can count in Search Console.
- AI answers break the old reading: one store we worked with went to 5.4 to 6.5 million impressions a month from May to October 2025 while its click rate fell from 5.85% to about 2%, so measure citations and sales, not only clicks.
- Clusters you have not published yet are your control group: if live clusters rise beyond last year's season and the unpublished ones do not, the map is doing the work.

## What happens in the days, weeks and months after you publish

After you publish the pages of a topical map, Google crawls them within days to weeks, shows them for many searches at low positions in the first weeks, moves them up and down for one to three months, and settles them over several months. Each phase has its own number to watch.

A topical map is the plan of every page a site needs on its subject, grouped into clusters and linked in a set order. A cluster is one group of pages on one part of the subject. Measuring the map means measuring those clusters, because Google judges a group of related pages differently from one page on its own.

Two words come up all through this lesson. Indexed means Google has stored a page and can show it. Impressions are the times Google showed one of your pages in its results, whether or not anyone clicked.

![A timeline of four phases after publishing a topical map, from crawling to settled rankings, with the number to check in each phase.](https://frameworkseo.com/learn/measure-a-topical-map/after-you-publish.webp)

*Each phase has its own measure. Judging clicks in the second phase is judging too early.*

### The four phases, and where each time range comes from

| Phase | Typical timing | What happens | What to check |
| --- | --- | --- | --- |
| 1. Crawl and index | Days to weeks | Google finds the pages and decides whether to store them | Pages indexed, out of pages live |
| 2. First showing | The first 2 to 4 weeks | Pages appear for many searches, mostly low on the page | Impressions and the number of different searches |
| 3. Re-ranking | About 1 to 3 months | Positions jump up and down as Google learns how searchers respond | Impressions per cluster and the share of the map's searches you show for |
| 4. Settling | Several months | Clusters hold their places, and clicks and sales follow | Clicks to the pages that sell, brand searches, AI citations, sales |

The first and last ranges come from three Google statements. The middle two are my rule of thumb, read from one practitioner's dated cases below, not a Google figure.

- **Crawling:** Google's help page on recrawling says ["Crawling can take anywhere from a few days to a few weeks."](https://developers.google.com/search/docs/crawling-indexing/ask-google-to-recrawl) (last updated December 2025).
- **Judging a change:** Google's [SEO starter guide](https://developers.google.com/search/docs/fundamentals/seo-starter-guide) says some changes take hours and others several months, and that "you likely want to wait a few weeks to assess whether your work had beneficial effects".
- **Site-wide trust:** Google's [core updates page](https://developers.google.com/search/docs/appearance/core-updates) says it "could take several months for our systems to learn and confirm" that a whole site has improved, and that after a few months with no effect you may be waiting for the next core update (a broad change to how Google ranks, made several times a year).
- **A dated case:** Koray Tuğberk Gübür, whose [2021 case study](https://www.oncrawl.com/technical-seo/importance-topical-authority-semantic-seo/) put the term and his method in front of the SEO industry, reports 10,000 searches, 300 clicks a day and 40,000 impressions in the first 20 days on one education site in 2021, which he labels initial ranking, and 33,000 searches with 11,000 clicks a day by day 65, which he labels re-ranking ([Holistic SEO, first published November 2021](https://www.holisticseo.digital/theoretical-seo/ranking)). On the same site, his first searches in the top three came after two months for one group of pages and on day 15 for a second group launched later ([Oncrawl, January 2022](https://www.oncrawl.com/technical-seo/creating-semantic-content-networks-with-query-document-templates-case-study/)).

Koray's numbers are his reported numbers, from his own screenshots, with no control site. Read them as one practitioner's record of the shape, not a promise of the size. A new site with no history moves slower than an established site adding a cluster.

> **A line you will see quoted that Google no longer says:** You may see Google quoted as saying SEO takes "four months to a year". That sentence is not on Google's current "Do you need an SEO?" page (checked 13 November 2026), so do not plan around it. Use the three statements above instead.

The order you publish in decides which clusters reach each phase first. That is the subject of [publishing order](https://frameworkseo.com/learn/topical-maps/publishing-order). This lesson starts the day the pages go live.

## Why rankings jump, drop and settle in the first months

Rankings move after you publish because Google first places a new page with little evidence about it, and then moves it as it learns how searchers respond. Koray Tuğberk Gübür calls the first placement the initial ranking, the moves re-ranking, and the site-wide result the ranking state.

### Initial ranking and re-ranking, in Koray's words

Koray defines the two in his [January 2022 Oncrawl case study](https://www.oncrawl.com/technical-seo/creating-semantic-content-networks-with-query-document-templates-case-study/): "The initial ranking represents the rank of a document on the SERP immediately after being indexed." Re-ranking is "changing a document's ranking on the SERP based on the feedback of the users". He adds a line worth keeping on a sticky note: "If a Search Engine indexes your web page, it doesn't mean that the search engine understood the web page."

These are Koray's terms and his model, not names Google uses. Google's own record points the same way. In a 2016 internal presentation shown at the US antitrust trial, Google's Eric Lehman wrote ["We do not understand documents. We fake it."](https://www.justice.gov/d9/2023-11/417516.pdf), explaining that Google looks at how people respond, and that this fails "maybe because a document is new, recently-changed, or rarely-shown". The court's [2024 opinion](https://storage.courtlistener.com/recap/gov.uscourts.dcd.223205/gov.uscourts.dcd.223205.1033.0.pdf) describes Navboost, a system that "pairs queries and documents through memorizing user click data", trained on 13 months of data since 2017. A page published last week has none of that record.

Google also patented a method that scores a page for a search by the share of clicks that lead to longer views ([US 8,661,029, granted 2014](https://patents.google.com/patent/US8661029B1/en)). A patent shows what Google designed, not what runs today. Together with the trial record, it explains why a new page wobbles before it settles: the evidence about it is still arriving.

### Why the average position gets worse while a map is working

An average position that gets worse in the first weeks is usually a sign the map is working. Search Console averages position over every search your pages show for. During re-ranking, a cluster starts showing for many new searches, and new searches usually enter low on the page. Those low positions drag the average to a worse number even while every search you already had improves.

![Made-up example: four searches all improve, yet the average position worsens from 5.5 to 21.6 because four new searches arrive at low positions.](https://frameworkseo.com/learn/measure-a-topical-map/average-position-falls.webp)

*A worse average with more searches is usually growth, not decline. Read the number of searches next to the average.*

Here is the arithmetic, with made-up numbers and one impression each. In month one, a page shows for four searches at positions 3, 5, 6 and 8, so the average is 5.5. In month two, those four have moved up to 2, 4, 5 and 6, and four new searches appear at 28, 35, 41 and 52. Every old search improved. The average is now 21.6.

Koray recorded the same pattern on his 2021 education site: "Whenever more web pages are indexed for more queries, the average position decreases, after collecting the necessary amount of historical data and relevance consolidation from the content network, the average position rises again" ([Holistic SEO](https://www.holisticseo.digital/theoretical-seo/ranking)). So never read average position alone. Read it next to the number of different searches the pages show for. More searches with a worse average is phase two. Fewer searches with a worse average is a real drop.

### Ranking state: why the same page can rank higher on one site than another

Koray's third term covers the whole site. In a [March 2024 case write-up](https://medium.com/@ktgubur/achieving-positive-ranking-state-120-growth-in-4-months-bcf51b590979) he defines it: "Positive Ranking State refers rankability of a web source for a specific timeline according to assigned quality and authority score of ranking algorithms." In plain words, Google holds a running opinion of your site on a subject, and the same page can rank higher or lower depending on it. In 2023 he went as far as ["Topical Authority is a positive ranking state."](https://www.holisticseo.digital/seo-research-study/topical-map)

He also says why he named it: to help SEOs read a site's current state "so as not to abandon the optimization journey and project or change the course of the efforts while about to get proper results". That is the practical point for you. Weeks four to twelve are the worst time to rewrite a page that is about to climb. More of his vocabulary, in plain words, is in [Koray Tuğberk Gübür's topical map framework](https://frameworkseo.com/learn/semantic-seo/koray-topical-map-framework).

> **Do not rewrite a page in its first month:** A page in re-ranking is collecting the evidence Google uses to place it. A rewrite in that window resets part of what Google has learned and hides whether the first version worked. Fix pages that are not indexed or that answer the wrong question. Leave pages that are indexed and gaining searches alone for at least four weeks.

## Why a topical authority checker cannot measure your map

No tool can read how Google rates your site on a subject. A topical authority checker shows its own estimate, built from its own keyword database and its own ranking samples, so it measures how you do on the vendor's list of keywords, not on your map.

Koray Tuğberk Gübür, whose case studies put the term in front of the SEO industry, says the same, in [February 2024](https://medium.com/@ktgubur/3-suggestions-about-topical-authority-1cd73963a9cb): "Refrain from believing that anyone can measure TA with a tool. You can only predict it by neutralizing PageRank and simulating organic potential in a comparison way." He adds that money spent on the "measurement of TA" is wasted, and that the useful comparison is how a site is prioritized against others for one topic or one form of query.

Google's internal documentation, leaked in 2024, has a field called siteFocusScore, described as a "Number denoting how much a site is focused on one topic", and one called siteRadius, "how far page_embeddings deviate from the site_embedding" ([Content Warehouse documentation](https://hexdocs.pm/google_api_content_warehouse/0.4.0/GoogleApi.ContentWarehouse.V1.Model.QualityAuthorityTopicEmbeddingsVersionedItem.html)). Google confirmed the documents are real and cautioned against "inaccurate assumptions" ([The Verge, May 2024](https://www.theverge.com/2024/5/29/24167407/google-search-algorithm-documents-leak-confirmation)). Microsoft holds a patent on "multiple topical authority scores" per site for different topics ([US 11,874,882, granted 2024](https://patents.google.com/patent/US11874882B2/en)). Whether and how much either is used is not public. If they are, they live inside the engines, and no outside tool can read them.

| Who looks | What they can see | What it tells you about your map |
| --- | --- | --- |
| A checker tool | Its own sample of keywords and where you rank for them | How you do on the vendor's list, which is not your plan |
| You, in Search Console | Every search your pages were shown for, with impressions, clicks and position | How your planned pages do on your planned searches |
| Google | Its own data, including fields such as siteFocusScore whose use and weight are not public | Whatever Google decides, which nobody outside Google can see |

Even inside Google, one number does not do it. A 2018 internal presentation shown at trial lists relevance, page quality, popularity, freshness, authority, topical diversity and more as aspects of search quality, and ends: ["Capturing everything in a metric is tough!"](https://www.justice.gov/atr/media/1316761/dl?inline)

So drop the search for one score. Your map is your own list of searches, written down before you published. The share of that list you show for is the coverage score tied to your plan, and Search Console gives it to you free. The next section shows how. Why the idea of authority on a subject holds up at all, and what Google has said about it, is in [topical authority](https://frameworkseo.com/learn/topical-maps/topical-authority).

## How to measure a topical map in Search Console, step by step

Measure six things for each cluster in Google Search Console, in the order they move: pages indexed, impressions, searches per page, the share of the map's searches you show for, internal link reach, and clicks to the pages that sell. Early measures tell you the map is being read; late ones tell you it pays.

![Six measures of a topical map in the order they move, from pages indexed in the first weeks to clicks on the pages that sell months later.](https://frameworkseo.com/learn/measure-a-topical-map/six-measures.webp)

*The measures move in this order. A late measure that is flat while an early one rises means wait, not fail.*

You need Search Console access for the site, your map in a spreadsheet and your analytics. All three are free. Before you start, write down for every page in the map its cluster, its address and the one search it is built for. That list is what you measure against. Then, before a cluster's first batch goes live, take its numbers once as a baseline, so month one has something to be compared with.

### 1. Pages indexed, out of the pages you published

1. Make a sitemap that lists only the map's live pages, and submit it under Indexing, then Sitemaps.
2. Open Indexing, then Pages. Use the filter at the top to show only that sitemap.
3. Write down the indexed count and the not indexed count, and open the reasons under "Why pages aren't indexed".
4. For the newest pages, paste each address into URL Inspection to see its state today.

Two reasons cause most worry and mean different things. Google's [Page indexing help](https://support.google.com/webmasters/answer/7440203) says "Crawled - currently not indexed" is "a common message for many sites and doesn't necessarily indicate an issue", while "Discovered - currently not indexed" means Google found the page and rescheduled the crawl. A few weeks in, most planned pages should be indexed. A cluster where most pages still sit in "Crawled - currently not indexed" after a month is worth a closer look: Google has seen those pages and not kept them yet, so check what each one adds that your indexed pages do not.

### 2. Impressions per cluster

1. Open Performance, then Search results. Set the date to the last 3 months.
2. Click Add filter, choose Page, then Custom (regex), which lets one pattern match several addresses at once.
3. Type a pattern that matches the cluster's addresses. If they share a folder, use the folder, such as `/grinders/`. If not, list the endings with a bar between them, such as `burr-grinder|hand-grinder|grind-size-chart`.
4. Choose Date, then Compare, and compare the last 3 months with the 3 before. Switch the chart to weekly.
5. Write down impressions, clicks and average position for the cluster.

Google's [filtering help](https://support.google.com/webmasters/answer/17011165) confirms the filter uses RE2 regular expressions, is not case sensitive, and that weekly or monthly views neutralize "the day-of-the-week effect". Impressions are the first number to move, because a page has to be shown before it can climb to where people click it.

### 3. Searches per page

1. In Performance, open the Pages tab and click one page of the cluster.
2. Open the Queries tab and set rows per page to the maximum.
3. Count the searches with at least one impression. Write the count next to the page in your sheet.

A page shown for 60 different searches reaches more of the subject than one shown for 3. Rising counts across a cluster are the plainest sign that Google connects your pages to more of the subject. Two limits apply, both from Google's [data help](https://support.google.com/webmasters/answer/17010575): the table shows at most 1,000 rows, and rare searches are left out of the table for privacy, though they count in the chart total unless a query filter is on. Data usually arrives 2 to 3 days late, so measure on the same day each month.

### 4. The share of the map's searches you show for

This is the measure the checker tools cannot give you, because it starts from your plan. It answers one question: of the searches your map was built for, how many does Google show you for at all, and how many on page one?

1. Take the target search of every live page in the cluster from your map.
2. In Performance, add a Query filter, Custom (regex), and list the searches exactly, such as `^(burr grinder|best hand grinder|coffee grind size chart)$`.
3. Open the Queries tab. Count the target searches that appear, and how many have an average position of 10 or better.
4. Divide each count by the number of target searches. For a long list, export the Queries tab and match it against the map in your spreadsheet instead.

![Made-up cluster of ten target searches: seven show in Google, so 70% shown, and four sit on page one, so 40% in the top ten.](https://frameworkseo.com/learn/measure-a-topical-map/share-of-map-searches.webp)

*Shown for is the coverage half. In the top ten is the competing half. Track both every month.*

In the made-up cluster above, 7 of 10 target searches show, so the share shown is 70%, and 4 of 10 sit in the top ten, so the top ten share is 40%. Expect the first number to climb in phases two and three and the second in phase four. One caution: a target search that very few people type can be hidden from the table for privacy, so "not shown" can also mean "too rare to list". A target search with real demand that never shows after three months is a page to look at, with [the checks further down](#not-working).

### 5. Internal link reach

1. Open Links. Under Internal links, open Top linked pages.
2. Find each page of the cluster and write down how many of your pages link to it.
3. Click a page to see which pages link to it, and check that the cluster's main page is among them.

A new page that no live page links to has only its own record to go on. Google's [Links report help](https://support.google.com/webmasters/answer/9049606) warns that the report "shows a sample of internal and external links" and that tables stop at 1,000 rows, so use it to find pages with zero or one link, not to count exactly. A page no other page links to may not appear in the list at all, so a page you cannot find there is a warning too. How the links in a map should run is in [internal links inside a topical map](https://frameworkseo.com/learn/topical-maps/internal-links-in-a-topical-map).

### 6. Clicks to the pages that sell

1. In Performance, filter Page with Custom (regex) to the pages that sell: services, products, categories, booking or pricing pages.
2. Write down their clicks, and compare with the same months a year earlier.
3. In your analytics, count the orders, leads or bookings from organic search whose first page was one of the map's pages.

The pages that sell are what Koray calls the core section. The guides around them, the outer section, exist to bring in impressions and searches that lift the core over time. So a map works when the guides rise first and the core follows. The split is explained in [core section and outer section](https://frameworkseo.com/learn/semantic-seo/core-and-outer-sections).

> **Add one number for reputation:** Filter Query with Custom (regex) to your brand name and its common misspellings, and write down the impressions each month. Google patented a site quality score built from how many searches refer to a site by name, compared with how many searches end in a click on it ([US 9,031,929, granted 2015](https://patents.google.com/patent/US9031929B1/en)). Brand searches in Search Console are only a stand-in for that idea, not the score itself. Rising brand searches over a year mean more people seek you out by name; write the number down either way.

## How AI answers change what impressions and clicks mean

AI answers mean impressions can rise while clicks fall even when rankings hold. Search Console counts a link inside an AI Overview as an impression at one shared position and adds it to your ordinary web results. A separate report now shows those AI impressions, but not the clicks, so the AI answer can still take the click without your numbers saying so. Measure citations and sales as well as clicks.

### How Search Console counts AI Overviews and AI Mode

- **Position:** "An AI Overview occupies a single position in search results, and all links in the AI Overview are assigned that same position" ([Search Console Help, how data is counted](https://support.google.com/webmasters/answer/7042828)).
- **Clicks:** clicking a link to your page inside an AI Overview or AI Mode counts as a click. A follow-up question in AI Mode counts as a new search.
- **Where it shows:** sites in AI features are "reported on in the Performance report, within the "Web" search type" ([Google, AI features and your website, December 2025](https://developers.google.com/search/docs/appearance/ai-features)).
- **The AI view:** since 31 August 2026 every site has a Generative AI performance report that shows impressions in AI Overviews and AI Mode on their own, by page, country, device and date ([Search Console Help](https://support.google.com/webmasters/answer/16984139); [Google Search Central, 3 June 2026](https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports)). It reports impressions only, not clicks.

Searchers click less when an AI summary is on the page. Pew Research Center tracked 900 US adults in March 2025 and found they clicked a normal result in 8% of visits to pages with an AI summary, against 15% without one, and clicked a link inside the summary in 1% ([Pew, July 2025](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/)). Two other measurements point the same way, each with its own weakness. SparkToro, using clickstream data from Datos (a Semrush company), counted 58.5% of US Google searches ending with no click in 2024, before AI Overviews spread ([SparkToro, 2024](https://sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/)). Authoritas estimated a 47.5% drop in click rate on desktop and 37.7% on mobile when an AI Overview appeared on UK news searches, in a report Google called flawed ([Press Gazette, 2025](https://pressgazette.co.uk/media-audience-and-business-data/google-ai-overviews-publishers-report-clickthroughs-authoritas-report/)). Bing's own guidelines now say your content can appear as "impressions, citations, or grounding references" without a click ([Bing Webmaster Guidelines](https://www.bing.com/webmasters/help/webmaster-guidelines-30fba23a)).

### What it looked like on a real site

One store we worked with, an online nicotine pouch store selling mostly to the US, peaked at 169,000 Google clicks in September 2024, at a click rate (the share of impressions that become clicks) of 5.85%. From May to October 2025, Google showed it more often than ever, 5.4 to 6.5 million times a month, yet clicks settled between 106,000 and 118,000 and the click rate sat around 2%.

![Chart of one store's monthly Google impressions rising to 6.5 million while its click rate fell from 5.9% to about 2%, September 2024 to October 2025.](https://frameworkseo.com/learn/measure-a-topical-map/impressions-up-clicks-down.webp)

*Real client data from Google's monthly Search Console emails, rounded as Google printed them. By our March 2025 reading, rankings mostly held while AI Overviews spread across its searches.*

When we looked at the pages in March 2025, our report to the client said the missing clicks came from AI Overviews, "Not a decrease in ranks". That was our judgment at the time, and the store's guide addresses had also moved in late 2024, so the cause cannot be fully separated. On 13 November 2026, all six of the store's biggest question searches we checked opened with an AI Overview, and the store was quoted in three of them.

If you only watch clicks, this looks like a site in decline. If you watch impressions too, it looks like a site Google still trusts, answering questions people no longer need to click for. Neither reading is complete without the next two numbers.

### Measure citations and sales, not only clicks

- **Citations in Bing.** Bing Webmaster Tools reports Total Citations, Average Cited Pages and the grounding queries "the AI used when retrieving content" for its AI answers ([Bing, February 2026](https://blogs.bing.com/webmaster/2026/2/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview/)). Since June 2026 it adds Citation Share, the percentage of citations for a grounding query that go to your site, and a Topics view that groups those queries into themes ([Bing, June 2026](https://blogs.bing.com/search/2026/6/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare/)). Bing describes Citation Share as an observational metric rather than a ranking system. It is a free engine report of citations grouped by topic, which is close to how a map groups its pages.
- **A fixed set of questions, checked by hand.** Write down 10 to 20 questions your map answers and check once a month whether ChatGPT, Perplexity and Google's AI answers cite you. Answers differ by person and by day, so treat one check as a snapshot and look at the trend over months.
- **Sales and leads from the map's pages.** This is the number that decides whether the map paid. Google says people who click from results with AI Overviews are "more likely to spend more time on the site", which, if it holds for you, shows up as sales per click, not clicks.

How to write and structure pages so AI answers quote them is a separate job, covered in [getting a topical map cited in AI answers](https://frameworkseo.com/learn/semantic-seo/ai-answers-and-topical-maps).

## How to tell the map caused the change, not the season

To tell whether the map caused a rise, compare clusters you have published with clusters you have not published yet over the same months, and compare each month with the same month a year earlier. A rise in the live clusters only, beyond last year's season, points to the map.

Most traffic changes have causes that have nothing to do with your pages. Google's [guide to debugging traffic drops](https://developers.google.com/search/docs/monitor-debug/debugging-search-traffic-drops) names algorithm updates, technical issues, security and spam issues, seasonality and changing interests, and site moves, and offers a year-over-year comparison of the last 3 months as one way to see what changed. A garden site that grows in April did not grow because of its map.

### Your publishing order already gave you a control group

Companies that run SEO split tests compare changed pages with a forecast built from pages they left alone. You can do a rough version for free, because a map goes live in batches. The clusters not published yet sit on the same site, in the same season, through the same Google updates. They are your untreated group.

| Made-up site, impressions, last 3 months vs the 3 before | Live clusters | Not yet published | What it means |
| --- | --- | --- | --- |
| Case A | +85% | +6% | The map is doing the work |
| Case B | +40% | +38% | Season or a Google update; the map has not shown its effect yet |
| Case C | -20% | -22% | A site-wide drop; check the causes in Google's debugging guide before touching the map |

For the untreated group, measure the existing pages on those subjects, or the searches the unpublished pages are built for. Write the publish date of every batch in your sheet so you can line the curves up later. This is not a controlled experiment, and the clusters differ in demand. It is still far better than reading one site-wide line and guessing.

## When to conclude a topical map is not working, and what to check first

Conclude a cluster is not working only when its pages have been live and indexed for about three months, its impressions are flat, and the share of its target searches you show for has not grown. Then check eight things in order, cheapest first, before you rewrite anything.

### The verdict by time

- **After 4 weeks:** judge only indexing. Most pages should be indexed. If they are not, that is the problem, whatever else the numbers say.
- **After 3 months:** judge impressions, searches per page and the share of target searches. Flat on all three for an indexed cluster is a real signal.
- **After 6 to 12 months:** judge clicks to the pages that sell and sales. Rising impressions with flat sales is a question about the pages that sell, or about AI answers, not about the map.

Those time boxes are my rule of thumb, set inside Google's own ranges above. A new site sits at the slow end. An established site adding one cluster often sits at the fast end.

![Eight checks in order for a topical map cluster that is not working, from reporting errors and indexing to reputation and market changes.](https://frameworkseo.com/learn/measure-a-topical-map/not-working-checks.webp)

*Check in this order. The first four take minutes, so rule them out before you touch the pages.*

### The eight checks, in order

1. **Is it the report?** Check the property, the date range and any filter left on. Google's debugging guide lists reporting glitches among the causes. A regex that misses half the cluster looks like a failure.
2. **Is it indexed?** Go back to measure 1. A page Google has not indexed cannot rank, and a single page that stalls is covered step by step in [publishing order](https://frameworkseo.com/learn/topical-maps/publishing-order).
3. **Do live pages link to it?** Go back to measure 5. Pages with zero or one internal link have only their own record to go on, so fix them before you judge the cluster.
4. **Do two of your pages share one search?** Click a target search in the Queries tab and open the Pages tab. Two of your pages splitting the impressions is [keyword cannibalization](https://frameworkseo.com/learn/topical-maps/keyword-cannibalization).
5. **Does the page match what ranks?** Search the target and look at the type of page on page one. A guide written for a search where Google shows product grids will not climb, however good it is.
6. **Do the pages build on each other?** When we review maps that buyers built themselves, the same three mistakes come up: pages that do not build on each other, off-topic pages that could have been part of another post, and searches that do not deserve their own page. The test for the last one is in [when a search deserves its own page](https://frameworkseo.com/learn/topical-maps/page-or-section), and what to do with pages that do not belong is in [pruning and merging](https://frameworkseo.com/learn/topical-maps/pruning-and-merging).
7. **Is reputation the missing half?** If the pages are indexed, linked and right for the search, and still sit on page two or three, other sites may not vouch for you enough on the subject yet. That half is covered in [links, mentions and reputation](https://frameworkseo.com/learn/topical-maps/links-and-mentions). The store above is a caution against jumping here first: its clicks grew about 29 times in 15 months while the number of sites linking to it went from about 80 to about 200, and hardly moved during the fastest climb.
8. **Did the market change?** Check demand for your main searches against last year, and ask what happened to the business. That same store sold mostly to the US, and in late August 2025 the Nordic postal service stopped taking parcels to the US. Nothing on a topical map protects you from that.

> **Wait for a core update before you give up on a whole site:** If the checks find nothing and the whole site is flat, Google's core updates page says it can take several months for its systems to confirm a site has improved, and that you may be waiting for the next core update. Keep publishing in order while you wait. Stopping is the one move that guarantees nothing changes.

## The monthly review sheet to keep, with a made-up example

Keep one row per cluster per month with the six measures, brand searches, AI citations and sales, plus one line on what changed that month. Fill it on the same day each month, and read the rows of one cluster top to bottom: the measures should start moving in the order of the timeline.

[Download the monthly map review sheet (CSV)](https://frameworkseo.com/learn/measure-a-topical-map/monthly-map-review.csv) (19 columns, one row per cluster per month, with where each number comes from, and a made-up three-month example of one live cluster and one unpublished control cluster to overwrite. Opens in Google Sheets or Excel.)

Here is how three months of one made-up cluster read, a store's 8-page cluster on coffee grinders.

| Month | Indexed | Impressions | Searches shown for | Avg. position | Target searches shown | Top 10 | Clicks to pages that sell |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Month 1 | 5 of 8 | 1,900 | 64 | 18.2 | 3 of 8 | 1 of 8 | 21 |
| Month 2 | 8 of 8 | 6,400 | 310 | 27.5 | 6 of 8 | 1 of 8 | 24 |
| Month 3 | 8 of 8 | 11,800 | 520 | 19.8 | 7 of 8 | 3 of 8 | 49 |

Month two is the one that scares owners. The average position got worse, from 18.2 to 27.5, while impressions tripled and the searches shown for went from 64 to 310. That is re-ranking: new searches arriving low. In month three the average recovers, three target searches reach page one, and clicks to the pages that sell double. The numbers are made up to show the reading, not to promise a pace.

The columns, and why each one is there:

- **Pages planned, live and indexed.** The gap between live and indexed is your first warning.
- **Impressions, clicks and average position for the cluster.** From the Page regex filter.
- **Searches shown for.** Read it beside average position, never apart from it.
- **Target searches shown and in the top ten.** The share of the map's searches, your coverage score.
- **Pages with fewer than two internal links.** From the Links report. The goal is zero.
- **Clicks to the pages that sell, and organic orders or leads.** The late measures that decide whether the map paid.
- **Brand search impressions and Bing citations.** The reputation and AI side, which clicks miss.
- **What changed this month.** Batches published, pages rewritten, Google updates announced. Without this line you cannot explain the curve in six months.

Once a quarter, the monthly rows feed a bigger decision: grow, fix or prune. That review, with its own sheet, is in [expanding a topical map](https://frameworkseo.com/learn/topical-maps/expanding-a-topical-map).

We built our own system for planning maps, which is how a map of about 159 pages, our median (maps run from about 50 pages to over 20,000), takes us about two to two and a half hours instead of a week or two by hand. Measuring it afterwards is the part every owner can and should do. If you would rather start from a map planned for your business, so the list you measure against already exists, that is the [growth plan](https://frameworkseo.com/).

## Sources

- [Importance of topical authority: a semantic SEO case study](https://www.oncrawl.com/technical-seo/importance-topical-authority-semantic-seo/), Oncrawl (Koray Tuğberk Gübür), 2021
- [Creating semantic content networks with query and document templates: a case study](https://www.oncrawl.com/technical-seo/creating-semantic-content-networks-with-query-document-templates-case-study/), Oncrawl (Koray Tuğberk Gübür), 2022
- [Importance of initial-ranking, re-ranking, and historical data for SEO](https://www.holisticseo.digital/theoretical-seo/ranking), Holistic SEO (Koray Tuğberk Gübür), 2021
- [How to expand a topical map for higher topical authority?](https://www.holisticseo.digital/seo-research-study/topical-map), Holistic SEO (Koray Tuğberk Gübür), 2023
- [Achieving positive ranking state: 120% growth in 4 months](https://medium.com/@ktgubur/achieving-positive-ranking-state-120-growth-in-4-months-bcf51b590979), Medium (Koray Tuğberk Gübür), 2024
- [3 suggestions about topical authority](https://medium.com/@ktgubur/3-suggestions-about-topical-authority-1cd73963a9cb), Medium (Koray Tuğberk Gübür), 2024
- [UPX0203: Q4 Search All Hands (trial exhibit, US v. Google LLC)](https://www.justice.gov/d9/2023-11/417516.pdf), US Department of Justice, 2016
- [UPX0204: Ranking for Research (trial exhibit, US v. Google LLC)](https://www.justice.gov/atr/media/1316761/dl?inline), US Department of Justice, 2018
- [United States v. Google LLC, memorandum opinion (Document 1033)](https://storage.courtlistener.com/recap/gov.uscourts.dcd.223205/gov.uscourts.dcd.223205.1033.0.pdf), US District Court for the District of Columbia, 2024
- [Modifying search result ranking based on implicit user feedback (US8661029B1)](https://patents.google.com/patent/US8661029B1/en), Google Patents, 2014
- [Site quality score (US9031929B1)](https://patents.google.com/patent/US9031929B1/en), Google Patents, 2015
- [Extracting key phrase candidates from documents and producing topical authority ranking (US11874882B2)](https://patents.google.com/patent/US11874882B2/en), Google Patents (Microsoft), 2024
- [QualityAuthorityTopicEmbeddingsVersionedItem (siteFocusScore, siteRadius)](https://hexdocs.pm/google_api_content_warehouse/0.4.0/GoogleApi.ContentWarehouse.V1.Model.QualityAuthorityTopicEmbeddingsVersionedItem.html), google_api_content_warehouse 0.4.0 (hexdocs), 2024
- [Google confirms the leaked Search documents are real](https://www.theverge.com/2024/5/29/24167407/google-search-algorithm-documents-leak-confirmation), The Verge, 2024
- [Google users are less likely to click on links when an AI summary appears in the results](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/), Pew Research Center, 2025
- [Introducing AI Performance in Bing Webmaster Tools (public preview)](https://blogs.bing.com/webmaster/2026/2/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview/), Microsoft Bing, 2026
- [New AI visibility insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare](https://blogs.bing.com/search/2026/6/New-AI-Visibility-Insights-in-Bing-Webmaster-Tools-Intents-Topics-Citation-Share-Compare/), Microsoft Bing, 2026
- [2024 zero-click search study](https://sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/), SparkToro (data from Datos), 2024
- [Google AI Overviews: publishers report clickthroughs (Authoritas report)](https://pressgazette.co.uk/media-audience-and-business-data/google-ai-overviews-publishers-report-clickthroughs-authoritas-report/), Press Gazette, 2025
- [Bing Webmaster Guidelines](https://www.bing.com/webmasters/help/webmaster-guidelines-30fba23a), Microsoft Bing, 2026
- [Ask Google to recrawl your URLs](https://developers.google.com/search/docs/crawling-indexing/ask-google-to-recrawl), Google Search Central, 2025
- [SEO starter guide](https://developers.google.com/search/docs/fundamentals/seo-starter-guide), Google Search Central, 2025
- [Google Search's core updates](https://developers.google.com/search/docs/appearance/core-updates), Google Search Central, 2025
- [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features), Google Search Central, 2025
- [Debugging drops in Google Search traffic](https://developers.google.com/search/docs/monitor-debug/debugging-search-traffic-drops), Google Search Central, 2025
- [Generative AI performance report (Search), rolled out to all sites 31 August 2026](https://support.google.com/webmasters/answer/16984139), Search Console Help, 2026
- [Introducing Search Generative AI performance reports in Search Console (3 June 2026)](https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports), Google Search Central, 2026
- [How are impressions, position, and clicks counted?](https://support.google.com/webmasters/answer/7042828), Search Console Help, 2026
- [Advanced filtering and comparison](https://support.google.com/webmasters/answer/17011165), Search Console Help, 2026
- [Troubleshooting data discrepancies](https://support.google.com/webmasters/answer/17010575), Search Console Help, 2026
- [Page indexing report](https://support.google.com/webmasters/answer/7440203), Search Console Help, 2026
- [Links report](https://support.google.com/webmasters/answer/9049606), Search Console Help, 2026