AI is being added to document tools everywhere, and it is worth separating what the category is heading towards from what any particular product does today.
So, plainly, before anything else: Doconut.app has no AI features. It renders your file so you can read it. There is no summarisation, no chat with your document, no AI annotation, and no OCR. If you came looking for those, this is not the tool, and the rest of this post explains what it does instead — and why the absence of one particular feature is a deliberate choice rather than a gap.
What AI Brings to Document Viewing
Products in this category are being built around four ideas. None of them is on Doconut.app:
- Semantic search: finding an answer phrased differently from the question, rather than matching a string.
- Automatic summarisation: the key points of a long report without reading it.
- Accessibility: text-to-speech and machine translation applied to a document.
- Suggested annotations: proposed comments and highlights for a reviewer.
All four assume something that is easy to skip past: the machine has to be able to read the document as text. That assumption is where most of these features quietly fail.
The text layer, and why it decides everything
A PDF exported from Word carries its words as characters. A PDF produced by a scanner carries a photograph of words — to software, a picture with no text in it at all.
Search on Doconut.app works on the text layer a file already has. Open a report exported from Word and searching finds what you are looking for; open a scan of the same report and searching finds nothing, because there is nothing there to find. That is not a limitation we are working around: there is no OCR here. We do not read characters off a picture, and we would rather say so than let you conclude the search is broken.
The same fact governs every AI feature in the list above. A summariser given a scanned contract has nothing to summarise unless something ran OCR over it first.
What Doconut.app does today
- Opens 70+ extensions — PDF, Word, Excel, PowerPoint, CAD, PSD, Visio, Project, email, EPUB, TIFF and the image formats.
- Renders them on our servers, so your device does not have to be fast.
- Searches the text layer where the file has one.
- Lets you zoom, move between pages, and print with Ctrl+P.
- Costs nothing, asks for no account, and takes one file at a time.
- Deletes your file when the viewing session ends.
The size ceiling belongs to the format rather than to the site: 25 MB for a PDF or a PSD, 20 MB for a spreadsheet or a deck, 15 MB for a Word file, 10 MB for a CAD drawing, 5 MB for text, XML and HTML.
The road ahead
Document viewers will keep absorbing AI, and there is a good argument that a viewer is the right place for it: it is where the document is already open. There is also an argument for caution about where the document goes in order to be summarised — a free tool that reads your contract in order to explain it has, by definition, read your contract.
Our current position is the conservative one. Your file is rendered and deleted, and nothing else happens to it.
Try it on a real file
Drop the document that is currently stuck on Doconut.app — no installation, no account, and no copy kept afterwards. If it is a scan and you need the words out of it rather than just to read them, you need an OCR tool, and we will not pretend to be one.