Drop in a Search Console or Ahrefs export. Every keyword is grouped by meaning, mapped as a living topic network, and handed back as a page-by-page plan — without a single keyword ever leaving your browser.
Each circle is one content topic, sized by how much it holds; the small dots around it are your actual keywords, grouped by meaning. Click any topic to open its page brief. Scroll to zoom, drag to explore.
This is a real run on a 500-keyword coffee list — the same result you get from the sample-data button above, in about ten seconds.
No API keys, no uploads, no queue. A sentence-embedding model runs inside your own browser tab, which is why it is free and why your keyword list stays private.
Export queries from Search Console, Ahrefs, Semrush or Keyword Planner. The keyword and volume columns are detected automatically, whatever the separator.
Each keyword becomes a 384-dimension vector, so terms that mean the same thing cluster together even when they share no words at all.
Every topic is one page, with its search intent, monthly volume and keyword list. Export the plan as CSV, or the map as a PNG for your report.
Most ways of grouping keywords ask you to name the groups first. This one reads them out of the keywords, which is the difference between a ten-minute job and an afternoon.
Tag every keyword against a list of terms, pivot, repeat next month. Exact, auditable and genuinely fine up to a few hundred keywords — the formulas are here. It matches characters, so “SERP” and “search results page” never meet.
The paid tools group by overlapping search results, which is Google's own verdict and the most accurate signal there is. It needs a live search per keyword, which is why it is priced per credit and why most lists never get run.
Each keyword becomes a vector and the close ones group. No tag list, no credits, no upload — it runs in this tab. Free to 5,000 keywords, and free above that if you ask.
The file type is detected on drop. The same clustering runs either way — on your queries, or on the pages you already published.
Any export with a keyword column. You get the topic map and a content plan: how many pages to write and what each should cover.
Same file, plus position and landing page. Now the map also shows what already ranks on page two and which queries two of your own pages are fighting over.
A Screaming Frog export maps the site you actually have: topics, every on-page problem the crawl can prove, and pages covering one topic without linking to each other.
Nothing here fetches your site. You run the crawler, the file is read on your device, and no request leaves the browser — see how it works.
Nothing. There is no paid tier — I built this for my own research work, it costs me nothing to run, and it is more useful to me with people using it.
Got a code? Enter it here.
What people ask before they upload their first file.
Keyword clustering is the process of grouping a large keyword list into sets of terms that share the same search intent, so each group becomes one page rather than many thin, competing pages. It is the step between keyword research and actually writing content — see the full guide.
Yes. Files of up to 5,000 keywords are free with no account and no card. The analysis runs on your own machine, so there is nothing for a bigger file to cost. Larger sets are free too — email info@graphmykeywords.com and you get an unlock code back.
No. The language model runs inside your own browser tab, so your CSV is read, embedded and clustered on your device. Nothing is sent to a server, which also means no data-processing agreement is needed to use it at work.
Yes. Drop in a crawl export from Screaming Frog and the tool maps your existing pages into topics, lists the on-page and technical problems the crawl can prove — broken pages, thin content, duplicate titles, orphans — and shows pages covering the same topic that do not link to each other. It never fetches your site itself: you run the crawler, and the file is read on your device like any other.
Any export with a keyword column — Google Search Console, Ahrefs, Semrush, Google Ads Keyword Planner, or a plain list. The keyword column is detected automatically, and comma, semicolon and tab separators all work.
Word matching puts “cold brew ratio” and “how to make cold brew” together but misses that “iced coffee recipe” belongs with them. This tool compares meaning using sentence embeddings, so keywords group the way a reader would group them rather than the way a spreadsheet filter would.
This is one. Upload a CSV and every keyword is grouped by meaning — free with no account up to 5,000 keywords, and free on request above that. There is no tag list to write and no paid tier, because the grouping runs on your own device rather than on a server we have to pay for.
Yes, and for a few hundred keywords you probably should — here are the formulas, including the SEARCH and TEXTJOIN tag matcher most people are after. The limit is that a spreadsheet matches characters, so it cannot see that two phrasings with no shared words mean the same thing.
all-MiniLM-L6-v2, a compact sentence-embedding model that turns each keyword into a 384-dimensional vector. It downloads once, is cached by your browser, and then runs offline.
Yes. Export the clusters as a CSV to hand to a writer or drop into a content calendar, and export the topic map as a PNG for reports and client decks.
Paste the code I emailed you. It is stored in this browser only.