Websites & SEO
Group keywords into topic clusters by search intent
Upload a keyword list as CSV, XLSX or TXT, one keyword per row. Every keyword gets a topic cluster, a search intent label and a confidence value, so you know which rows to check first. Your manual changes are kept.
or drop it here
Price: 1 credit per run of up to 50 keywords · 20 credits for €6.99 · Pro includes 40 credits a month
Longer list? Split it into files of 50 rows. 500 keywords are 10 runs and 10 credits.
| Keyword | Cluster | Intent | Confidence |
|---|---|---|---|
| compress pdf to 2 mb | Compress PDF to a set size | transactional | 0.91 |
| reduce pdf size to 1 mb | Compress PDF to a set size | transactional | 0.88 |
| pdf compressor online free | Compress PDF online | transactional | 0.84 |
| reduce pdf file size | Compress PDF online | transactional | 0.72 |
| make pdf smaller on phone | Compress PDF on mobile | transactional | 0.63 |
| why is my pdf so large | Why PDF files get large | informational | 0.41 check |
Low-confidence rows are listed again in the review CSV. The keyword CSV keeps every input row with its cluster, ready to import back.
- Cluster workbook
- Review CSV with confidence
- Import-ready keyword CSV
How a run works
- 1
Add the list
Choose a CSV, XLSX or TXT file with one keyword per row, up to 50 rows. You sign in with a free account in the workspace.
- 2
Pick granularity and language
Tight, balanced or broad clusters. German, English or auto-detect for mixed lists. The cost of 1 credit is shown before the job starts.
- 3
Check and download
Look at the preview, then download the workbook and both CSV files. If a job fails, the reserved credit is released.
Grouping a keyword list into topic clusters
Your keyword list goes in
Upload a CSV, XLSX or TXT file. One run takes up to 50 rows within a 64,000 character cap, so a larger export has to be split first. Only the phrases you supply are clustered. Case and spacing are folded for the comparison, but no row is dropped and the original phrase stays in the workbook.
The clusters and the confidence column
You get a cluster workbook, a review CSV with a confidence value for every assignment, and a keyword CSV you can import back into your own tools. Start with the low-confidence rows, since those are the assignments most likely to need a change. Each phrase sits next to the cluster it landed in, so you can check a group at a glance.
Granularity, language and the limits of intent labels
Tight gives more and smaller groups, while broad merges close neighbours into larger topics. Language can be German, English or auto-detect for mixed lists, and it changes how phrases are normalised. The tool does not use live search results. Intent and cannibalisation labels are therefore estimates from the wording, and whether two phrases deserve one page stays your decision.
Questions before you run it
How many keywords can I cluster at once?
Up to 50 keywords (rows) per run, within a 64,000 character cap on the extracted text. Each run uses 1 credit. For a longer list, split it into files of 50 rows: 500 keywords are 10 runs and 10 credits.
Do I need an account?
Yes, a free account. Clustering runs as an AI job, so it uses credits: 1 credit per run, shown before the job starts. A pack of 20 credits costs €6.99, and Pro includes 40 credits a month. If a job fails, the reserved credit is released.
Does it use live search results to decide the clusters?
No. Clustering runs on the phrases you supply. Without SERP data the intent and cannibalisation labels are heuristic, which is why every assignment carries a confidence value.
Can I make the groups tighter or looser?
Yes. Cluster granularity offers tight, balanced and broad. Tight yields more and smaller groups, broad merges near neighbours into larger topics.
Are any input rows dropped during normalisation?
No row is discarded. Case and spacing are folded for the comparison, but the original phrase stays in the workbook next to the cluster it landed in.
Which languages are supported?
German and English, or auto-detect when an export mixes them. The language setting changes how phrases are normalised before they are compared.