Enter paper forms from photos

You end up with every filled form stored as a paper response — read from a photo, checked by you, without typing the answers in.

Before you start

• Your organisation is on Pro, and you have the editor role or higher on the survey.
• The forms were printed with Print a paper form. Every page carries a code; it is how a photo finds its survey, version and page. A form printed any other way cannot be read.
• Best: you printed numbered copies. Each copy carries its own number on every page, so photos and scans of a whole stack sort themselves, in any order. With one copy to photocopy, every photocopy looks the same, so you tell Mapsurvey which photos belong together (see below).
• You have the forms as photos (any phone) or scans (a flatbed, an office copier, a PDF).

Photograph the forms

• One page per photo, the whole page in the frame, flat on a table, in even light. Keep the two QR codes at the bottom in the picture; one is enough if a thumb covers the other. On a map page keep all four black corner squares in the picture: they are what puts the marks on the map.
• Order does not matter. Numbered copies sort themselves; for a response you open yourself, every page finds its place by the page number printed on it.
• A scanner is fine too: a PDF of the stack, or one image per page.

Steps

1. Open the survey’s Responses page, open the arrow next to Enter a response and choose From photos of paper forms. The Paper responses page opens. It shows how many numbered copies were printed and how many have come back.
2. Add the photos, one of two ways:
   • Upload a stack. Drop the files on the page, or choose them. Dozens at once is fine, and so is adding a few more while the first ones are being read. A multi-page PDF is split into pages for you. Pages of numbered copies go to their copy’s response, whatever order they came in. With one copy to photocopy, upload one PDF per filled form: each PDF becomes one response.
   • New response. Press it, then drop the photos or the PDF of one form into its card. Use this for a single form, or for photocopied forms photographed page by page. On a phone, press Take photo to shoot pages straight from the page — one page per shot, tap again for the next.
3. Wait for the cards to fill in. Each response shows a box per page: read, to check (the reader was unsure about something), reading…, or not photographed.
4. Press Check on a response, or the Check Copy N button at the top, which always names the next one ready. The survey opens page by page of the printed form, with the photo of the page on the left and the questions printed on it on the right, already filled in. Fields the reader was unsure about are marked in amber; fields it saw left blank are grey.
5. Check each page against the photo, correct what is wrong, press Next. The page boxes at the top jump to any page. On the last page press Finish: the form is stored as a paper response and its photos are deleted. Check next opens the following one.

When a photo is missing or in the wrong place

• A page was not photographed. The check says so where the photo would be. Press Take photo or choose files right there; the page is read and the fields fill in. You can also fill that page by hand, or press Next to leave it blank.
• The same page twice. The later photo wins. The earlier one is kept under the response; Use the earlier photo swaps them back.
• A photo dropped into the wrong card. A numbered page always goes to its own copy, and the card you dropped it on says where it went.
• Set aside, on the Paper responses page. A photo lands here when nothing can tell whose it is: a single photo of a photocopied form, a page whose code cannot be read, a page of a response already stored, or a page of another survey. Each one says why. Pick the response and page number and press Attach, or remove it.

What the reader is good and bad at

• Ticks, crosses and filled boxes: good. A faint pencil mark or a box on a fold: it will say so and ask you to check.
• Handwritten text: transcribed as written, in its language. Expect to fix a word now and then; the field is marked amber when the reader was unsure.
• Numbers in boxes and ratings: good.
• Marks on the map are read too. A dot, a line or an outlined area drawn on the printed map, with its letter and number written next to it (A1, B2), appears as a pin on the map of the check where the respondent drew it. The map page of the check looks like the paper: the printed frame as a live map, then the same table of marks with the same rows and columns, filled with what the respondent wrote in each row. Click a row and the photo scrolls to its mark; drag a pin to correct a place; a row that is empty on paper offers place it on the map in case the reader missed one. For this the photo must show the small black squares beside the map (six, printed in the margins next to its corners and halfway down its sides) or the four corner squares of the page: they tell Mapsurvey how the paper lies in the picture, and a page that is bent or photographed at an angle is read as it lies. The more of them in the shot, the better; a photo of the map alone is fine as long as the squares beside it are in. Without any of them the map frame itself is used, less precisely; a photo that shows none comes back with the map questions empty, and you place the marks by hand. Forms printed before November 2026 have only the page corners.

Where things go

• A paper response is stored exactly like a response typed in by hand: source Paper form, your name as the person who entered it. The response details link back to it as Copy 37 (or Response 4 for a form without a copy number), and exports carry the copy number in the paper_copy column, so you can find the paper original.
• Photos are kept only until the response is stored, and at most 30 days. Delete on a response you do not want removes it with its photos; its copy number can be uploaded again. A response nobody touches for 30 days is deleted the same way.
• Each page is read by the same AI that drafts surveys; the page image is sent to the provider for the duration of the reading. The codes are read on our server, not by the AI. See the trust page.