How do you fill out a PDF that is not fillable?

What “not fillable” actually means

A PDF is fillable only if the person who made it drew the boxes. Those boxes are a structure inside the file called an AcroForm: a list of named fields, each with a rectangle, a type and a value. Acrobat, a browser and a phone all read the same list, which is why a fillable form behaves the same everywhere.

A form with no AcroForm has none of that. It may still look like a form — rules to write on, boxes to tick, a line for a signature — but those are ink. They are drawn on the page the same way a logo is. There is nothing to click because there is nothing there.

Three kinds of file arrive this way, and it is worth knowing which one you have:

What you have	What is in the file	What it needs
A print-out saved as PDF	Real text, real vector rules, no fields	The blanks worked out from the rules
A scan or a phone photo	One image per page, no text at all	The page read as a picture first
A form the publisher flattened	Text and rules, fields deliberately removed	The same as a print-out

The third is the one that catches people out: the form was fillable, and the publisher exported it flattened, which welds the boxes into the page.

The four ways to fill one

Print it, write on it, scan it back. Still the most common answer, and still the worst. It costs a printer, a scanner and a fresh set of artefacts on the page, and the result is a photograph of a document rather than a document.

Add text boxes by hand. Every general PDF editor can do this. You click, you type, you drag the box until it sits on the line. It works, and on a two-page form with forty blanks it is twenty minutes of dragging.

Convert it to a fillable form. Acrobat’s Prepare Form and the online tools that copy it try to find the fields for you and then hand you an editable form. This is the right shape of answer. The cost is a subscription in Acrobat’s case, or an upload in everyone else’s.

Read the page and fill it in place. The same detection, without the conversion step and without the upload: render each page, find the blanks, put a real input over each one, and write the answers into the file on export. That is what FormworkPDF does, and the rest of this page is about how well it works.

How the blanks are found

A printed form says where to write using a small number of shapes, and it has used them for a century. Three carry almost everything:

• A rule. A solid horizontal line with room above it. The rejections are what make it usable: ink sitting on the line is underlined text, another rule 18 points away is a table, and a line under a centred heading is a decoration.
• A leader. The row of dots that has marked a blank since before forms were typeset by machine. A machine set them, so they are evenly spaced and the same size — which is exactly what separates a leader from the last line of a paragraph.
• A boxed cell. Faint column lines break up in a scan, but the point where a horizontal rule stops survives. Those endpoints are the column edges, and a pair of them crossed by two rules is a cell.

The page is rendered off-screen, thresholded against a local average rather than a fixed cutoff — scanned paper is grey, and a fixed cutoff reads the whole sheet as ink — and searched for those shapes. Each blank that survives becomes an ordinary field: it gets an input, a place in the tab order, validation, and a name taken from the words printed beside it.

On a scanned IRS 1099-MISC — one image, no text layer, no fields — that takes about 300 milliseconds and produces 33 boxes, every one of the four columns included.

How well it works, measured

Detection is either scored or it is a claim. FormworkPDF is scored against 41 published forms: 9 that ship a real AcroForm, which is its own answer key, and 32 that the IRS published as print, from a 1913 Form 1040 to office scans of the 1980s.

	Fast	Accurate	Best
Text fields found	86%	93%	96%
Tick boxes found	52%	95%	100%
False positives	459	28	44
Time per page	0.3 s	~3 s	~10 s
One-off download	nothing	37 MB	97 MB
Works offline	always	after the first run	after the first run

Fast is pure geometry: nothing is downloaded, so it runs on a plane, and it is the only tier that handles a 200-page document without a wait. Accurate and Best are trained detectors that additionally tell a tick box from a text box. The app asks which you want before each scan, because the right answer depends on the document in front of you.

The misses are worth reading too. Form W-9 scores 40%, which sounds bad until you look at the marked page: its SSN and EIN are comb fields, one widget per character, so nine red boxes are one blank a human types into once. The score counts per widget. How that is scored goes through the answer key and the two bugs that only appeared at scale.

Filling it without uploading it

Everything above runs in the tab. The file is turned into an ArrayBuffer, read by pdf.js for rendering and by pdf-lib for writing, both inside the page you are already looking at. There is no request carrying your document, because there is no server to carry it to.

That matters most on exactly the forms that are not fillable. A blank W-4 is public. Your W-4, with your address and your dependants on it, is not — and the usual arrangement, where a free tool holds your file for an hour and then deletes it, is a promise about a copy that was made. Not making the copy is a different kind of answer.

1. Open the PDF. If it has fields, you get them immediately.
2. If it has none, choose Find blanks and pick how hard to look.
3. Type. Tab walks the page the way an eye does, top to bottom and along each line.
4. Sign it by drawing, typing, or photographing your signature on paper.
5. Export. The answers are written onto the page and the file is flattened, so nothing can be edited back out.

If the result is too big for the portal you are sending it to, shrink it on the way out.