Reverse Engineering Competitor Topic Clusters with Search Operators and Google Sheets

The modern SERP is a multidimensional array of entities, intents, and semantic relationships, and your competitors have already decoded it for you. But instead of paying for a subscription to a data aggregator that spits out a sanitized backlink report, you can manually reverse engineer their entire topic architecture using nothing more than Google search operators, a free Google Sheets instance, and the kind of logical inference that separates actual SEO engineers from keyword bingo players. This isn’t about cheating; it’s about reading the public signals your competitors leave in their wake. Every internal link, every subtopic mention, every piece of anchor text is a breadcrumb pointing to the way they have organized their knowledge silos, and you can rebuild that map with a few clever queries and some spreadsheet gymnastics.

Start by identifying a competitor whose topical authority you want to dissect. Do not pick a domain that ranks for a thousand unrelated queries; pick one that clearly owns a niche, where every piece of content orbits a central commercial or informational theme. Once you have that domain, your first investigative move is to use `site:competitor.com` paired with a modifier that exposes internal directory structure. Most savvy publishers use subfolders to categorize content, and a quick `site:competitor.com/blog`, `site:competitor.com/guides`, or `site:competitor.com/product-category` will often reveal the skeleton of their information architecture. But you can also go deeper with `inurl:keyword` to find pages that Google has indexed with that term in the URL, which frequently coincides with their primary topic segmentation. The key is not to stop at the homepage but to enumerate as many deep pages as possible. You can do this manually by clicking through the site directory, but the free version of Screaming Frog allows a 500-URL crawl, which is often enough for a mid-tier domain. Export that crawl to a CSV, then pull the URL, page title, and internal link counts into Google Sheets.

Now the real reverse engineering begins. You are not looking for keywords; you are looking for clusters — groups of pages that topically cohere through dense internal linking and shared semantics. In your spreadsheet, create a column for the “parent topic” by examining the URL structure and the breadcrumbs that Screaming Frog captures. Then use the `=IMPORTXML` function in Google Sheets to pull the H1 and meta description for each URL directly from the live site. This gives you a raw corpus. Next, apply a simple TF-IDF analysis to the aggregated text from these titles and descriptions. You can do this with a free Google Apps Script or a manual formula using `SPLIT` and `COUNTIF`, but the goal is to identify the frequency and exclusivity of terms across the cluster. A term that appears in the titles of five URLs and in the anchor text pointing to those URLs is almost certainly a structural topic core. A term that appears once is probably a satellite or a long-tail variant. This distribution tells you how your competitor has chosen to prioritize their topical coverage.

The second layer of analysis involves the SERP itself. For each of the core topics you’ve identified, run a manual Google search with `keyword site:competitor.com` to see which of their pages actually rank. Pay attention to the People Also Ask boxes and the bottom-of-the-page related searches. These aren’t just user behavior artifacts; they are Google’s own approximation of a semantic graph. Record those related entities in a separate sheet, because they represent the boundaries of the cluster your competitor has staked out. Then cross-reference that list with your crawl data. If your competitor has a page that competes for a related search term but is not internally linked to from their main cluster hub, that is a weakness — it means their silo has a structural hole. You can exploit that hole by building a more tightly interwoven group of pages around those terms.

The most underrated free tool in this entire workflow is the humble “related:“ operator. Run `related:competitor.com/path/to/core-page` and Google will surface other pages it considers topically similar. Compare that list to your crawled URLs. The intersection is what Google themselves identify as your competitor’s cluster. The difference — pages that appear in the related results but not in your crawl, or vice versa — flushes out orphaned or semantically misclassified content. You can also use `site:competitor.com` with a sorted list of your own seed keywords to check where the competitor ranks for terms you care about, but that only gives you ranking data, not structural data. The structural data comes from the internal link matrix. If you have Screaming Frog’s crawl, you already have the link graph. Throw it into a simple node-edge list in Google Sheets, then use a form of cosine similarity on the co-occurrence of outbound links from each page. Two pages that link to the same three or more other pages are encoding a shared topical relationship. That shared relationship is the cluster signal you are after.

Manual competitor analysis is not about seeing what you’re missing; it’s about seeing the pattern of what your competitor has chosen to build. The best part is that this entire process leaves you with a reusable spreadsheet that functions as a mini data pipeline. Once you’ve reverse engineered one competitor’s topic clusters, you can apply the same queries to five more competitors. The overlaps and divergences among their cluster maps reveal unexplored subtopics that none of them have fully covered, which is where you’ll find your white space. Free reporting tools are often dismissed as too crude for serious SEO, but the crude ones force you to use your brain. That manual reasoning, combined with the precision of search operator wrangling, produces an insight density that no paid dashboard will ever give you. So open up Sheets, drop in a few `IMPORTXML` hooks, and start decoding the competition one semantic breadcrumb at a time.