About The Dictation Index
What the Index is, what keeps the ranking honest, and where it still falls short.
Updated
What it is
The Dictation Index ranks voice dictation software by what people say about it in public. It collects posts that mention a tool by name, labels each one positive, neutral or negative, and turns the counts into a score with a fixed formula. Every score is printed with its sample size, its margin of error and the dates its posts cover.
It also keeps a registry of 69 dictation tools with their platforms, pricing and processing model, checked against each vendor's own site.
Why it exists
Most comparison pages in this category rank on one writer's opinion and show no sample size. The aim here is a ranking where the method is fixed in advance, the inputs are public posts, and each tool's losses are printed next to its wins. Whether it succeeds is yours to judge, and the rest of this page is here to help you do that.
What keeps the ranking honest
- A fixed, published formula. Every post counts 10, 5 or 0. The prior, the minimum sample and the filters are on the method page.
- Margins and ties on every ranking. The top score's lead is inside the margin of error. Ranks 1 to 8 are a statistical tie, and every page that shows the ranking says so.
- Losses printed next to wins. Each ranked tool's page has a topic table. Even on the highest-scoring tool's page that means reliability (3 positive posts against 32 negative) and price (30 positive posts against 44 negative).
- No paid placement. No vendor can buy a position, a badge, a review or a mention. Listing is automatic for anything that meets the definition.
- Open data. The registry and every published number are in tools.json under CC BY 4.0.
Where it falls short
- You cannot re-run the counts. The post IDs behind each score are not published yet. Until they are, every score is our count.
- The labelling step is not documented. How each post gets its sentiment and topic labels, and how accurate those labels are against a human-labelled sample, has not been published.
- One source. Every post comes from X, which leans toward people who build and talk about software. 57 of 69 registry tools have too few posts there to be ranked at all.
- Affiliated posters are not filtered. Official vendor accounts are removed. Employees, investors, affiliates and paid creators posting from personal accounts are not yet identified, for any tool.
- The method embeds choices. The source, the filters, the 50-post minimum and the formula are all judgment calls. Different reasonable choices would move some tools.
Each of these is described in full under known limits.
How to read the Index
Use it as one input. The registry facts can be checked against each vendor's own site. The scores tell you what people posting on X said about a tool in a recent window. They do not tell you which tool is best for you.
Contact
Corrections, vendor replies, removal requests and questions go to [email protected]. The corrections page explains what to send and what happens next.