AI Validation
Write one sentence that says what a good answer looks like. While the respondent fills in the form, AI checks their answers and files against it and shows a note when something does not fit, so they can fix it before they submit.
ProThis needs the Pro plan or higher. Compare plans
What it does
An AI Validation is a block under the questions it checks. You write a rule in one sentence and name the answers it reads with
@. It catches mistakes a normal validation rule cannot: a document that is not what you asked for, two answers that
contradict each other, or an answer that is filled in but makes no sense.
The respondent sees a note under the question when something does not fit. They can fix the answer or submit anyway, because a check only warns and never blocks the submission.
You see the result on every submission: Passed, or Needs review with a short reason written for you. Automations get the same result.
What you can check
| What you want | A rule that checks it |
|---|---|
| The right document | @Invoice is an invoice for order @order_number, dated in 2026. |
| Names that match | The name on @Passport matches @Full name, and the passport has not expired. |
| Two answers that agree | @Work email uses the same domain as @Company website. |
| A plausible answer | @Company website is the website of a real company, not a social media profile. |
| The right kind of file | @Bank statement is a bank statement from the last three months, not a screenshot of an app. |
| Enough detail | @Damage description says what was damaged, where and when. |
| Answers that match a document you sent | @Total amount matches the total on @Your invoice. |
A rule can name a hidden field as the value to compare against, such as an order number or a
customer name your system already knows. In the first example, @order_number comes from the request, and the respondent never
sees it.
A rule can also name a Documents block above it, to check answers against the files you hand the respondent, including the ones a request adds. In the last example, the check reads the invoice in the block and compares it with the total the respondent entered. A rule that names a Documents block still needs at least one answer, and it does not run while the block shows no document.
Add an AI Validation
- 1
Insert the block
Under the question you want to check, click AI Validation in the Blocks tab, or type /ai. The block starts out naming the question above it. It cannot go inside a column or a repeating group.
- 2
Write the rule
Write one sentence that says what a good answer looks like. Type @ to name questions and Documents blocks above the block, and hidden fields. The check reads only what you name.
- 3
Try it in Preview
Open Preview and answer as a respondent would. Preview runs the real check, so you see the note respondents will see.
A rule that names nothing never runs
If the sentence names no answer the check can read, with no @ or only questions below the block, the block says so:
“Name a question or file above with @, or this check never runs.”
Write a good rule
- Check one idea per block. To check two things, add two blocks. Each gets its own note and its own result.
- Be concrete about names, numbers, dates and document types: “dated in 2026” checks better than “recent”.
Name every answer the rule needs. The check reads only the answers you name with
@.- Write in any language. The respondent's note follows the language the form is shown in.
What respondents see
The check runs once the respondent has filled in the answers it names and moved on. “Checking…” shows while it runs. Next and Submit wait up to 8 seconds for it.
The note is written for the respondent, in the language the form is shown in. It says what is missing and how to fix it. It never mentions AI or your rule, and never shows a hidden field's value. It says “the order number we have for you” instead.
- When the answers fit, a quiet “Looks good.” shows instead.
Changing an answer checks again. The same answers are never checked, or charged for, twice: coming back to a saved draft shows the earlier result.
A check that cannot run shows nothing. Without a plan or enough credits, or when the check fails, the respondent carries on as if there were no check.
Results on each submission
Every submission records each check as it stood on the submitted answers. On the request page, and in a submission's details on the Submissions tab, each check sits right after the last answer it read, with an icon for its result:
| Result | What it means |
|---|---|
| Passed (green check) | The answers fit the rule. |
| Needs review (amber warning) | Something did not fit. The reason, written for you in the form’s language, says what. |
| Not checked (dashed circle) | The check did not run on these answers: no plan or credits, it was still running at submit, an answer it names was empty, or a Documents block it names had no document it could read. |
An automation gets the same results with the answers, so it can send flagged submissions to a person to review. Developers: see Webhooks reference.
Files it can read
A check reads PDF, Word, Excel, CSV, plain text, Markdown and HTML documents, and JPEG, PNG, GIF and WebP images: up to five files per check, each up to 4.5 MB. When a file cannot be read (another type, too large, or damaged), the note asks the respondent for a clearer photo or the original file.
Documents from a Documents block count toward the same limits. When the check cannot read one of them, it does not run, because the respondent cannot fix a file you sent. The submission records Not checked instead.
Plans and AI credits
AI Validation is on the Pro and Business plans. On Free you can add the block, but checks do not run. Each check uses the workspace owner's AI credits, including the checks any member runs in Preview.
| A check that reads | Credits it uses |
|---|---|
| Text answers | Under 1 |
| A photo, or a one-page PDF | ~1 |
| Each further PDF page | ~0.4 |
When the owner's credits run out, checks stop until the owner adds more. Respondents notice nothing, the submission records Not checked, and the editor and an email tell the owner. See AI credits & usage.
What is sent, and what is kept
A check sends your rule, the answers it names and the files it reads to the AI model, and nothing else from the form. The model only returns a result and a note. It cannot act on anything written in an answer or a file, so the worst a planted instruction can do is change a check's result, which you see on the submission.
The result is stored with the submission. The note a respondent saw is kept with their link for up to 30 days, so the same answers are not checked again, and is deleted with the submission, the request or the form.