Importing questions from another tool

If you already wrote the survey somewhere else, you should not have to type it again to find out whether InviziPoll suits you. Import questions takes an export from your current tool, shows you every question it recognised, and turns your confirmation into a draft poll you can edit before anyone sees it.

Where to find it

Import questions sits in the admin sidebar. You can also reach it from the first step of the poll wizard, and from the empty state on an empty poll list.

Import is available on every plan, including free workspaces and workspaces whose trial has ended. The AI bias review that can follow an import is a paid step; without it, the flow simply ends one step earlier and your questions are saved exactly as written.

What you can bring

Formats are detected automatically. If detection guesses wrong, override it from the format selector on the preview screen and the file is read again.

The CSV shape is a header row beginning with question, then one row per question with its answer options:

question,option1,option2,option3
How clear was the reorg announcement?,Not clear,Somewhat clear,Very clear

Files up to 2 MB are accepted, and an import produces at most 50 questions. If your export holds more, the first 50 come across and the preview tells you how many were left behind, so you can split a long instrument into two polls deliberately rather than discovering the gap later.

Pasting instead of uploading

Google Forms and Microsoft Forms have no question export, so there is no file to give us. For those, and for any form you can only see on a page, use the Paste from anywhere tab: select the form, copy it, paste it in. Page furniture, numbering and "Required" markers are stripped as it is read.

A paste is between 20 and 20,000 characters and produces at most 25 questions. A longer form is refused rather than trimmed, so you are never handed a short poll and left to notice the gap yourself. Split it and paste twice.

On a workspace with AI authoring, a model reads the paste and it costs one AI credit. On every other workspace, including free plans and ended trials, the paste is parsed in your browser for free. Either way you land on the same preview. See What happens to what you upload or paste for where the text goes.

What survives the crossing

InviziPoll polls are built from eight question types: single-select multiple choice, multi-select multiple choice, numeric scale, NPS, ranking, word cloud, open text, and date or time. Your source tool almost certainly has some type that is none of those, so a little translation happens, and the preview is where you check it.

1. Choice questions arrive as choice questions. Typeform multiple choice, picture choice and dropdown; SurveyMonkey single and multiple choice; Qualtrics multiple choice. Option labels come across in order. A question with fewer than three options is padded to three placeholders, because a poll needs at least three; a question with more than eight (ten for multi-select) keeps the first of them.
2. Rating questions arrive as scales. Typeform rating, opinion scale and NPS keep their number of steps. Qualtrics sliders keep their configured minimum and maximum. A source NPS question becomes a 0 to 10 scale rather than an NPS question, deliberately: a 0 to 10 rating might be a recommendation question or might be a satisfaction rating, and guessing would relabel your question for you. The promoter and detractor figures are computed for any 0 to 10 range, so the score is on the results page either way. Change the type in the builder if you want the question labelled as an NPS.
3. Matrices arrive as a grid. A Qualtrics or SurveyMonkey rating matrix becomes one scale per row, with the first and last column headings kept as the scale's end labels, and the rows carry a shared group so the builder, the respondent page and the results view show them as the table you authored. A matrix whose columns are choices rather than a rating run becomes one single-select per row instead, with the column labels as its options, and is grouped the same way. Each row is still a complete question underneath, which is what keeps exports, cohorts and trends working per row. See matrix and grid questions.
4. Ranking questions arrive as ranking questions. Typeform, SurveyMonkey and Qualtrics rank-order questions keep their items and their order, because order is what a ranking question is. A ranking whose export carried no readable items is skipped and listed, rather than stubbed as something else. See ranking questions.
5. Free-text questions stay free text. A long-answer box arrives as open text and a short-answer box as a word cloud, both keeping their original wording. Nothing about them is turned into multiple choice.
6. Date and time questions arrive as date or time questions. The input mode comes across. Exports rarely carry a timezone, so UTC is assumed and the preview says so: set the real one in the builder before publishing.
7. Descriptive text is dropped, not converted. A statement, a page of instructions or a SurveyMonkey presentation block is not a question, so it is skipped and listed with everything else that was skipped.
8. Everything else becomes a placeholder. A file upload, a signature field or a question type with no equivalent arrives as a multiple-choice question carrying the original wording, with the source's own answer labels where it had any and three placeholder options where it had none. The preview says which of the two you are looking at and why. Rewrite it or remove it before publishing.

Some Qualtrics element types are skipped outright rather than stubbed: text entry, descriptive blocks, page breaks, timing and metadata, constant sum, side-by-side and drill down. A rank order element is skipped only when the export carried no items to keep; when it has them, it comes across as a ranking question. The preview lists each thing it skipped and why.

What does not come across at all

Answer data, first and unconditionally. Also: branching and display logic, required-answer settings, images and media, page and section structure, piped text, randomisation, quotas, and any respondent or contact list. Formatting inside question text is stripped, so a bolded phrase arrives as plain words.

Long question text and long answer labels are shortened rather than dropped: anything over 200 characters for a question, or 60 for an option, is cut at a word boundary and marked with an ellipsis, and the preview tells you which questions were shortened so you can reword them yourself.

Rebuild branching in the poll builder after the import lands. Everything else is either a setting you choose fresh on the InviziPoll side or something the product deliberately does not hold.

Why answers are never imported

Responses in InviziPoll are encrypted in the respondent's browser, and the server stores only ciphertext. There is no key on our side and therefore no way to write a historical answer into a poll such that it would ever be readable in results. Importing them would mean either storing plaintext answers or storing something we could never show you.

Keep your old tool's exports where they are for historical reporting, and use InviziPoll from the next round onward. See Viewing results for how a new poll's results become readable.

What happens to what you upload or paste

Two different things happen, and which one depends on how you brought the questions in.

An uploaded export file is read once, in the request, by InviziPoll's own parsers. The recognised questions go straight back to your browser for the preview. The file is not stored, its text is not stored, and it does not leave our servers.

Pasted text on a workspace with AI authoring is read by a model, which is the only way to make sense of a blob of text copied off a page. The text you paste is sent to Anthropic, the AI provider InviziPoll uses, which reads it and returns the questions it found in it. It is used to answer that one request and not to train any model. The paste is not stored on our side either, and the model is instructed to transcribe the text rather than act on anything written in it.

Pasted text on any other workspace never reaches our servers at all: it is parsed in your own browser.

In all three cases nothing is saved until you name the draft and confirm.

Our audit log records that an import happened: who ran it, which format was detected, how many questions came back, how many warnings there were, and how large the upload was. For a paste it also records the model used, whether a credit was charged, and the size of the paste in bytes. It never records any of the content, of either the paste or the questions.

After the preview

Name the poll, pick a team folder if your workspace uses them, and confirm. You get a draft poll, not a published one, so nothing is visible to respondents until you publish it yourself. From there it is an ordinary poll: see Creating your first poll for the rest of the wizard, and Workspace poll templates if you want to save the result as a reusable starting point.

Turning import off

Import can be disabled for a workspace on request. It is not a self-service setting: ask InviziPoll support and an operator turns it off for your organization, after which the import entry points disappear and the endpoint refuses.

There is also a rate limit of 30 imports per workspace per hour. A real import session never reaches it.

FAQ

Can I import our historical responses or results? No, and this is not a roadmap item. Responses are encrypted in the browser and the server holds ciphertext only, so there is nowhere for readable historical answers to go.

My import shows placeholder options on some questions. Did something break? No. Those are the few question types InviziPoll genuinely has no equivalent for, such as a file upload, or choice questions that offered fewer than three options. The original wording is always kept, and the source's own answer labels are kept too wherever the export carried any, so a placeholder means there was genuinely nothing to keep. The preview says which case each one is.

Can I import more than 50 questions? Not in one pass, and a paste caps lower still at 25. A long file brings the first 50 across and tells you what it left; a long paste is refused outright rather than trimmed. Either way, split the instrument and import twice, which also tends to make for a poll people finish.

Can I import from Google Forms? Yes, by pasting. Google Forms and Microsoft Forms have no question export to download, so open the form, select the questions, copy them and use the Paste from anywhere tab.

Does importing use my AI allowance? Uploading a file does not. Reading an export is free on every plan, including free workspaces and ended trials.

Pasting does, on workspaces that have AI authoring, because a model reads the paste. One paste costs one AI credit, and only if it finds something: if the text turns out to be a results export, an empty page, or a form longer than the 25-question cap, you are told so and nothing is charged. On workspaces without AI authoring, pasting is parsed in your browser and is free.

The optional bias review after either route uses AI credits and only appears on plans that include AI authoring.

Will an import overwrite an existing poll? No. Every import creates a new draft.