AcroForm is the original interactive form format baked into the PDF spec: a flat dictionary of field objects with types like /Tx (text), /Btn (button, which covers checkboxes and radio groups), /Ch (choice, for dropdowns and list boxes), and /Sig (signature). Each field has a partial name in /T and an optional user-facing tooltip in /TU, and most of the tax, immigration, and HR forms you run into (W-9, W-4, I-9, 1040, and their many cousins) ship as AcroForms. This matters because the form values are stored as real data in the PDF, so when a payroll system or an IRS intake pipeline reads the file back, it gets structured values instead of having to OCR a picture of text.
Manual editing on this page is fully client-side: the PDF is parsed in your browser, you click into text fields, tick checkboxes, pick radios, and type. The AI path (free account, 20 free starter credits on signup) walks the field tree, reads the field names, tooltips, and any parent group hierarchy (the /Parent chains that turn a set of radio buttons into a single logical choice), then takes whatever info you pasted and maps it to the matching fields. If your W-9 has a field named topmostSubform[0].Page1[0].f1_01[0] with a tooltip of "Name", the AI knows to put your legal name there. For button fields, it picks the correct /V export value so checkboxes are actually checked and the right radio in a group is selected.
The output is a real AcroForm PDF, not a rasterized copy. Values are written into the field dictionaries, appearance streams are regenerated where needed (and NeedAppearances is set when a viewer needs to redraw them), and annotation flags are preserved. That means the file opens in Acrobat, Preview, or a browser viewer as a filled form, and any backend that introspects the field dictionary sees your values. If you later want to flatten it for archival, that is a separate step you choose, not the default.
Yes. The exported file is still an AcroForm. We write values into the field /V entries, regenerate or invalidate the appearance streams (/AP) so the visible text matches the stored value, and set NeedAppearances on the AcroForm dictionary when a downstream viewer has to redraw them itself. Button fields get their /V set to the correct export value so checkboxes show as checked and the chosen radio in a group stays selected. You can reopen the PDF, edit the fields, and save again without anything being flattened behind your back.
XFA is a different thing. It is an XML-based dynamic form format Adobe layered on top of PDF, often used for forms that grow rows or recalculate fields on the fly. Some older government and enterprise forms are XFA-only or hybrid XFA/AcroForm. We handle AcroForm. If a PDF is pure dynamic XFA with no AcroForm fallback, our editor will open it but the dynamic fields will not be recognized as AcroForm fields. In practice most US tax forms ship as AcroForm (or hybrid where the AcroForm layer is what matters), so this is usually fine, but it is worth knowing where the line is.
Not as a built-in export today. Under the hood the SDK already walks the field tree and logs field names, types, and values during diagnostics, so the data is there. A proper JSON export of field name and value pairs (and for buttons, the selected export value) is on the near-term list because developers keep asking for it for testing and for bulk workflows. If you need it sooner, reach out and we will prioritize it.