When a form only lets through the right kind of answer, the data lands in your base ready to use, and the weekly cleanup pass mostly goes away. Air Fill is built to catch bad input while someone is still filling out the form, so the phone numbers, order codes, and values that reach Airtable already have the shape your workflow expects.
What Airtable forms check, and what they don’t
Airtable’s forms do check a couple of things well. An email field confirms that an address looks like an address, and a URL field confirms that a link looks like a link. In the newer interface-based forms you also get a small set of rules: a character minimum and maximum on text fields, a range on number fields, and the ability to mark a field required. For a lot of forms that really is enough, and it covers the everyday cases without any setup.
Where it runs short is anything with a specific shape — a phone number that has to be ten digits, a postal code that follows a national format, an order code that starts with a fixed prefix, or a text field that should only ever hold one of the few spellings your automations are matching against. As of writing, Airtable forms have no way to express rules like these. There’s no pattern or regex option in the form builder, and although Airtable does offer REGEX functions, those live only in formulas, and formula fields can’t be placed on a form. So the form accepts whatever someone types, and the checking, if it happens at all, happens later.
The workaround, and why it wears thin
Most teams end up patching this after the fact, and both of the usual routes cost something. One is to send submissions through a separate form service that supports custom validation and then sync the cleaned data back into Airtable. That works, but now you’re paying for and maintaining a second tool plus the connection between it and your base, and your form lives outside the place where everything else does.
The other route is to let every submission land and catch the problems downstream, with an automation or a formula field that flags the rows that look wrong. Someone then works a cleanup view on a regular schedule, re-typing malformed numbers, reconciling values that were entered three different ways, and emailing people to ask what they actually meant. The base quietly becomes the place where mistakes surface rather than the place where they’re prevented, and the person untangling them is rarely the person who typed the bad value in the first place. Airtable has a broad surface to cover, and deep per-field validation is one of the corners a small, focused tool can fill in.
How Air Fill keeps bad data out
Air Fill starts by making the correct answer the easy one to give. Every field type gets the input control that actually fits it, so a phone field opens a phone keypad, a number field opens a number pad, and a date field shows a calendar you tap instead of a box you type a date into. A surprising amount of bad data simply never gets entered, because the keyboard itself steers toward the right format, and the email and URL checks you’d expect run on the fields that need them.
Beyond shaping the input at entry, custom validation is in development: per-field rules you set once so the form can hold someone to a specific format before anything is submitted, whether that’s a length, a numeric range, or a pattern your workflow depends on. A field that has to be exactly ten digits can be told to be exactly ten digits, and a code that always starts with the same prefix can be held to it. When it ships, a value that doesn’t fit will be caught right there in the form, with the error shown on the field itself and everything the person has already filled in kept exactly where it was, so they can correct the one thing and keep going. The checking is designed to stay quiet until there’s something real to check, so a field isn’t flagged while it’s still empty or half-typed. Nobody gets scolded for a value they haven’t finished writing, and a submission that runs into trouble holds onto what was entered rather than losing it.
The point of all of it is a base you can trust a little more. Records arrive already in the format your views, formulas, and automations are counting on, the form does the catching so a person doesn’t have to, and the evenings that used to go to cleaning up after a form that let the wrong thing through get handed back to you.
