A Q&A on topic mapping, internal links, thin duplication, and knowing when a cluster is done

In 1986, Metroid dropped players onto planet Zebes with no map and no instructions. Every corridor led somewhere, every locked door eventually opened, and by the end you knew that planet better than your own street. That feeling is exactly what a topic cluster is supposed to create. Build enough connected rooms around one subject, and Google, human readers, and the new AI answer engines all walk away with the same conclusion: this site owns the place.

1. What is a topic cluster, in plain English?

A pillar page covers a subject broadly. Around it sit cluster pages, each going deep on one facet. Everything links to the pillar, the pillar links back to everything, and siblings link to each other when a reader would naturally need the next one. HubSpot popularized the model around 2016. Its 2017 guide by Mimi An described a pillar that “acts as the main hub of content for an overarching topic” while related pages “link back to the pillar page and to each other” (quoted by Practical Ecommerce). HubSpot later reported that “the more interlinking they did, the better the placement in search engine results pages.”

1. The pillar is the overworld map. Broad, skimmable, points everywhere.
2. The cluster pages are the dungeons. Each one has a single job and finishes it.
3. The links are the doors. Without them you have thirty rooms and no castle.

2. Why does AI search reward clusters even more than old-school Google did?

Because AI engines don’t run one search. They run a dozen. Google calls it query fan-out. At Google I/O 2025, Head of Search Elizabeth Reid explained that AI Mode “calls on our custom version of Gemini to break the question into different subtopics, and it issues a multitude of queries simultaneously on your behalf.” Google VP Robby Stein’s example: “things to do in Nashville with a group” quietly spawns searches for restaurants, bars, and kid-friendly spots. Semrush estimates a complex prompt can fire 8 to 20+ subqueries.

1. Rank for the subtopics, get cited for the topic. Ahrefs found that in July 2025, 76% of AI Overview citations ranked in the top 10 for the same query. By early 2026, across 863,000 keywords and 4 million cited URLs, that fell to 38%. About 31% of citations now come from pages ranked 11 to 100, and another 31% from pages outside the top 100 entirely.
2. Questions trigger the AI. Pew Research found AI summaries appeared on only 8% of one- or two-word searches, but on 53% of searches with 10+ words and 60% of searches starting with who, what, when, or why.
3. Google appears to measure focus. The 2024 Google API documentation leak listed an attribute called siteFocusScore, described as a “number denoting how much a site is focused on one topic,” plus siteRadius, which measures how far each page drifts from the site’s overall theme. Weights are unknown, so treat it as a strong hint.

3. How much traffic is actually on the line?

Plenty. Pew Research tracked 68,879 real Google searches from 900 U.S. adults in March 2025. When an AI summary appeared, people clicked a regular result in 8% of visits, versus 15% without one. They clicked a link inside the summary in just 1% of visits, and 26% of sessions simply ended, versus 16% on normal results pages.

1. Position one is leaking. Ahrefs measured a 34.5% lower click-through rate for the top result when an AI Overview appeared (March 2025 data, 300,000 keywords). Its December 2025 rerun put the gap at 58%.
2. Citation is the lifeboat. Seer Interactive tracked 53 brands, 5.47 million queries, and 2.43 billion impressions. On AI Overview results in 2025, cited brands averaged 2.07% organic CTR versus 0.94% for uncited brands, and 15.74% versus 11.19% on paid. Study lead Tracy McDonald was candid: “We cannot definitively prove that citation causes higher CTRs.” Either way, cited beats uncited.
3. The audience is enormous. AI Overviews reach more than 2 billion monthly users, per Alphabet’s Q4 2025 earnings call. OpenAI reported 900 million weekly ChatGPT users in February 2026.

4. How do I map the topic before I write a single page?

Think like a kid with Pokémon Red and Blue. The Pokédex had 151 slots, and you always knew exactly which ones were still blank. Build a Pokédex for your subject before you write anything.

1. State the pillar in one sentence. If you can’t, the topic is too big. Split it.
2. List the facets: definitions, how-tos, tools, costs, comparisons, mistakes, troubleshooting, alternatives, glossary.
3. Mine real questions from People Also Ask, Reddit, support tickets, and sales calls. Then type your core question into ChatGPT, Gemini, and Perplexity and write down every subtopic their answers touch. That is a free preview of the fan-out.
4. One intent per page. If two keywords return mostly the same top ten results, they belong on the same page.

Facet	Pages	Example titles
Pillar	1	Retro Console Repair: The Complete Guide
Console fixes	8	NES blinking red light (the 72-pin connector) · Genesis no video · Game Boy screen lines · N64 stick drift · PS1 laser alignment
Cartridge & disc care	5	Cleaning contacts · Replacing save batteries · Restoring cracked labels
Tools & parts	6	3.8mm vs 4.5mm security bits · First soldering iron · Multimeter basics · Capacitor kits
Techniques	5	Recapping a board · Retrobriting yellowed shells · Desoldering without lifting pads
Buying & value	3	Spotting reproduction carts · Fair prices for untested lots
Safety & glossary	2	Working safely around power supplies · Repair terms A to Z

Table 1. A sample 30-page map for a hobby site about fixing old consoles. Every spoke answers one question the pillar raises.

5. How should the internal links actually be wired?

The best data comes from Zyppy, which studied 23 million internal links across 1,800 sites and about 520,000 URLs, matched against Google Search Console clicks. 53% of URLs had three or fewer internal links pointing at them. URLs with 0 to 4 internal links averaged about 2 clicks, while URLs with 40 to 44 saw roughly four times that. Past about 45, traffic started falling. Pages with at least one exact-match anchor had around 5x the traffic. Founder Cyrus Shepard’s advice: “we should be increasing not necessarily the number of internal links, but increasing our anchor text variations.” Zyppy flags all of this as correlation.

1. Pillar links down to every spoke. Every spoke links up with a descriptive anchor.
2. Link sideways where the reader goes next, like Metroid’s doors. “Recapping a board” should point to “Capacitor kits.”
3. Vary anchors naturally. “NES blinking light fix,” “72-pin connector swap,” and “why your NES won’t boot” all describe the same page.
4. Keep everything within about three clicks. Super Mario Bros. hid a warp zone in World 1-2 that skipped you to Worlds 2, 3, or 4. Your pillar is that warp zone.
5. Respect the kill screen. Pac-Man breaks at level 256. Links have a ceiling too, and forty well-chosen ones beat four hundred sitewide footer links.

6. What counts as thin duplication, and how do I dodge it?

In 1982, Atari rushed E.T. the Extra-Terrestrial out in roughly five weeks. Unsold cartridges ended up in a New Mexico landfill in Alamogordo, and diggers pulled them back out in 2014. Thirty near-identical pages spun up to fill a cluster are your E.T. cartridges. In March 2024 Google launched its scaled content abuse policy, aimed at “producing content at scale to boost search ranking,” whether “automation, humans or a combination are involved.” Google later said users saw 45% less low-quality, unoriginal content, beating its own 40% goal. The broader backdrop: Ahrefs found 96.55% of roughly 14 billion pages get zero Google traffic.

1. The swap test. If you can swap the city, product, or console name between two pages and nothing else changes, merge them.
2. The SERP test. Two pages chasing queries with the same top results will cannibalize each other. Combine them and 301 the loser.
3. The Missingno test. Pokémon’s famous glitch duplicated whatever sat in your sixth item slot. Great for Rare Candy, awful for a site. Every page needs something the others lack: a number, a photo, a step, a tool, a mistake you made.

7. How do I know when a cluster is finished?

Castlevania: Symphony of the Night (1997) showed a map-completion percentage, and players famously discovered it tops out at 200.6% once you find the inverted castle. That counter is the lesson. Give your cluster its own completion meter.

Signal	What “done” looks like	How to check
Coverage	Every fan-out subquestion has a home URL	Run 15 to 20 prompts in ChatGPT, Gemini, Perplexity; map each subtopic to a page
Link health	Zero orphans; pillar and spokes link both ways	Crawl with Screaming Frog or Sitebulb
No cannibals	One URL ranks per query	Search Console: queries showing 2+ of your URLs
Citations	You’re named in AI answers for core questions	Track in a brand-mention tool or a simple monthly log
Diminishing returns	New pages earn impressions but no new queries	Search Console query count per new URL
Freshness	Stats, author bios, visible “last updated” dates	HubSpot’s State of AEO 2026 tied these to more AI citations

Table 2. The cluster completion meter. Hit all six and you’ve earned your 200.6%.

Finished is a save point. Return every quarter, because the fan-out changes as the questions do. Then pick the next subject and start mapping a new castle.

Sources

1. Pew Research Center, “Google users are less likely to click on links when an AI summary appears” (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/
2. Ahrefs, “AI Overviews Reduce Clicks by 34.5%” and “Update: AI Overviews Reduce Clicks by 58%”. https://ahrefs.com/blog/ai-overviews-reduce-clicks-update
3. Ahrefs, AI Overview citations vs. rankings (1.9M citations). https://ahrefs.com/blog/search-rankings-ai-citations
4. Search Engine Journal, “Google AI Overview Citations From Top-Ranking Pages Drop Sharply” (March 2026). https://www.searchenginejournal.com/google-ai-overview-citations-from-top-ranking-pages-drop-sharply/568637/
5. Seer Interactive, “AIO Impact on Google CTR: 2026 Update”. https://www.seerinteractive.com/insights/aio-impact-on-google-ctr-2026-update
6. Zyppy, “23 Million Internal Links: SEO Case Study”. https://zyppy.com/seo/seo-study/
7. Wix SEO Hub, internal linking webinar with Cyrus Shepard. https://www.wix.com/seo/learn/resource/internal-linking-webinar
8. Search Engine Journal, Robby Stein on query fan-out (July 2025). https://www.searchenginejournal.com/query-fan-out-technique-in-ai-mode-new-details-from-google/552532/
9. Semrush, “What is query fan-out?” (Reid I/O 2025 quote). https://www.semrush.com/blog/query-fan-out/
10. Google, “New updates to address spam and low-quality results” (March 2024). https://blog.google/products-and-platforms/products/search/google-search-update-march-2024/
11. Search Engine Land, “Unpacking Google’s massive search documentation leak”. https://searchengineland.com/unpacking-googles-massive-search-documentation-leak-442716
12. Ahrefs, “96.55% of Content Gets No Traffic From Google”. https://ahrefs.com/blog/search-traffic-study/
13. HubSpot, “Topic clusters: The next evolution of SEO” (State of AEO 2026 findings). https://blog.hubspot.com/marketing/topic-clusters-seo
14. Practical Ecommerce, “SEO: Topic Clusters Improve Rankings, Traffic” (Mimi An quote). https://www.practicalecommerce.com/?p=165019
15. SQ Magazine, AI search engine statistics (AI Overviews 2B users; ChatGPT 900M weekly). https://sqmagazine.co.uk/ai-search-engine-statistics/