The legal basics your AI-built app is missing.

An AI builder works from your prompts. You described screens, sign-up, payments and a dashboard, and that is what it produced. Policies, consent wording, an unsubscribe link and image descriptions were never in the prompt, so they are usually absent, or present only as a footer link that leads nowhere. Nothing in the app looks broken, which is why founders tend to find these gaps late, often when a payment provider, an app store reviewer or a careful customer asks for one of them.

This page is not legal advice. We are a software company, and we cannot tell you what the law requires of your business. What applies to you depends on where the business is, where your customers are and what the app does with their information, and those are questions for a lawyer. What we can do is describe each item in plain words, show you how to see whether your app has it, and separate the parts that are software from the parts that are legal judgment.

Read it as a list of questions to settle before launch. Having every item here in place does not guarantee that you are free of legal risk, and a missing item does not by itself mean you have broken a rule. The decision table further down sorts each one by who should do it.

Visitors, payment providers and app stores commonly look for a small set of pages before they trust a product. A privacy policy says what information you collect, why, who else receives it and how a person can contact you about it. Terms of service set out the agreement between you and the people using the app. A refund policy says what happens when someone wants their money back. A cookie policy lists what the site stores in the browser and what each item is for. The last one is the simplest: who runs this, under what business name, and how to reach a person.

Checking takes a few minutes. Open your app signed out and look in the footer and on the sign-up and checkout screens. Click every policy link. In the apps we see, some of those links go to an empty page, to the home page, or to a block of placeholder text with another company's name left in it. Then read what is there against what the app does. A policy that says you never share data is wrong if the app sends sign-up emails through a mailing service.

You can ask your AI tool to draft these pages, and it will produce something that reads like a policy. Treat that as a draft. The tool does not know how your business is set up, where your customers are or what your contracts say, and it fills the gaps with plausible text. Before a lawyer sees it, do the part only you can do: write a plain list of what the app collects, which outside services receive it, and how refunds work today. A lawyer reviewing a draft against that list is working from facts about your app instead of the tool's guesses.

Some trust problems come from the words on the page and the way a screen nudges a choice. AI tools generate these too. Ask for a landing page and you commonly get placeholder testimonials with invented names, a row of star ratings, a count of happy customers and a line claiming the product is the fastest or the most secure. They are there because landing pages usually carry them. If they go live, your site is making statements about real people and real results that are untrue.

Read every public page as a stranger would and mark each claim you could not prove if a customer asked. Reviews should come from real customers who agreed to be quoted. Numbers should be ones you can show a source for. Words like secure, certified or guaranteed should match something that exists. Remove what fails, and ask the tool to search the code for leftover sample text, because the same placeholder often appears on several pages.

Then walk through checkout with a test card. The price a person sees on the pricing page should be the price on the final screen, with any tax, fee or renewal stated before they pay. If the product renews, the screen should say when and for how much, and a customer should be able to cancel from inside their account.

Design tricks that push people into choices are often called dark patterns. Common ones are a pre-ticked box that adds a paid extra, a decline button in pale grey beside a bright accept button, a countdown timer that resets when the page reloads, and a cancel path that takes far more steps than signing up did. An AI tool may add these when asked to improve conversion, since it has seen many pages that use them. Look for them on sign-up, checkout, the cookie banner and the cancel screen, and ask the tool to give both choices equal weight.

Some of your visitors use a screen reader, some cannot use a mouse, and many read on a phone in bright light. An app built from prompts about how screens look often leaves them out. The basics below are a sensible place to start, and you can check each one yourself.

Text descriptions on images, often called alt text, are what a screen reader says in place of a picture. Ask your AI tool to list every image with no description, or with a file name as its description, then write a short sentence for each image that carries meaning. Pictures that are only decoration can be marked so a screen reader skips them. Buttons that show an icon and no word need a label for the same reason.

Color contrast is whether text stands out from what is behind it. Pale grey text on white, or white text over a photo, is a common result of a design prompt that asked for something minimal. The inspection tools built into your browser can report contrast for any piece of text, and free contrast checkers do the same from two color values. Check body text, the placeholder text in forms, and the disabled and error states.

Keyboard navigation is a test you can run in one sitting. Put the mouse aside and move through sign-up, the main task and checkout using the Tab key, Enter and the arrow keys. You should always be able to see which element is selected, reach every control and close any popup. Menus and dialogs built from custom components commonly trap the keyboard or skip it entirely. Whether an accessibility standard applies to your business is a legal question. Fixing what the test finds is software work, and it helps every person who uses the app.

The table puts the whole list in one place, and the last column is the one to read closely. Where it says you, the work is checking facts and changing the app, and an AI tool can usually make the change. Where it says a lawyer, the question is what your business has to do, and no tool or software company should answer that for you.