Skip to main content
Free AI For That

How We Check Free Offers

Last updated 20 October 2026

This page explains how Free AI For That ("FAFT", "we") checks free offers, in short. The full rules, including what qualifies and which sources count as evidence, are in our Verification policy. Where the two differ, the Verification policy applies.

1. What we check

A tool is listed only if it's an AI tool and its free offer lasts at least a month. A product with AI features is listed only for those features, and only if they're in the free offer. Any stated free amount counts if it lets you finish the tool's main job at least once and keep the result. Free offers only for students count; free offers only for other groups don't.

2. What counts as proof

Only two things:

  1. The maker's own pages (its website, documentation, help pages and official code repository), quoted word for word, with a link and the date we read them.
  2. Our own signup test, where a member of our team signs up for the free plan.

Review sites, blogs, news, app stores and other directories are never proof. What we can't confirm, we show as unknown, and unknown never counts as a pass.

3. Checking a new tool

  1. We confirm the official website and that the tool isn't already listed.
  2. We read the maker's homepage, pricing page and the help, FAQ, signup and documentation pages that describe the free offer.
  3. An AI model reads those pages under our written rules and reports each fact with the maker's exact words. Our software checks that every quote appears on the page. A quote that doesn't is rejected.
  4. Our software applies the rules: is it an AI tool, is the AI in the free offer, does the offer last at least a month, and which one kind of free it is.
  5. We record how to get it free: whether you need an account, what the signup asks for, the page to start from, the steps, and any catches. Each detail needs its own quote.
  6. We record what our ranking and awards use:
    • how long the free offer lasts, and whether credits refill;
    • whether the main feature works on it, and whether its limit is stated;
    • whether a card is needed;
    • the catches: watermark, quality cap, commercial use, training on your content.

We also record the free allowance and the cheapest paid plan. Each fact needs its own quote; anything not stated stays unknown.

  1. Nothing is left blank. Every fact the ranking uses, every part of the tool page and every part of "how to get it free" is either quoted, or marked "not stated on the maker's pages" with the date we looked. A new tool with a blank isn't put forward until it's complete.
  2. The listing text is written from the same pages only, in our own words, and a second check compares every sentence with those pages.
  3. A member of our team approves the listing before it's published. Paid listings also need our signup test first. When the maker's own pages disagree, or the tool may be a demo, shutting down or renamed, it isn't put up for approval until a member of our team has looked. A change of owner is noted; it changes nothing unless the free offer changes.

4. Re-checking every listing, every day

Every night our software reopens the pages we quoted for every listed tool and compares them with the saved copy:

  • Nothing that matters changed (about 9 in 10 tools): the offer is confirmed and the tool's check date moves to today.
  • Missing details (how to get it free, the catches, the free and paid allowances, the details page text) are filled in from the maker's pages, about 40 tools a day, each with its quote or a "not stated" mark. They go into the next daily update, which shows how many tools are complete.
  • "No card needed" is re-confirmed each night the maker's words are still on its page.
  • Something changed (a quote is gone, a number changed, or words like "trial", "beta" or "discontinued" appeared or disappeared): an AI model reads the new page under our rules and reports what changed, with quotes our software checks.
  • The page moved: we look for the new page on the maker's own site.
  • The page can't be read (blocked or down): we try again the next day. After three days in a row, a member of our team checks it. A tool is never removed because its page couldn't be read.

If more than 1 in 10 tools look changed in one night, we assume our own check has failed: that night's comparison isn't used (no check dates move, and no change to an offer is proposed from it), and we investigate. That day's update can still carry new tools, missing details and category fixes.

5. Changes and removals

  • Any change to a free offer, better or worse, must show on two checks at least a day apart, and a member of our team approves it before the listing changes.
  • When an offer ends or no longer qualifies, the tool is removed from our lists. Its page stays, saying why. Any pick or award it holds is withdrawn at once, the change is logged on Free plan changes with the maker's own words, and unused paid listing time is refunded.
  • Announced changes (for example "the free plan ends on 1 December") are logged on the day we see them. The tool stays listed until that date.
  • Readers and makers can report a change. We check the maker's own pages, never the report alone.

6. What AI does, and what people do

  • AI models read pages, report facts with quotes, and draft listing text. They can't publish anything.
  • Our software checks every quote against the page, applies the rules the same way every time, and keeps the records.
  • People approve every new listing and every change to a free offer, do the signup tests, and handle anything unclear.

7. Records and testing

For every fact we keep the maker's words, the page, the date, and a saved copy of the page. Every month we run our checker on at least 50 tools a person has checked by hand. It must agree with the hand check on every answer to "is it free, and what kind", or we fix it before relying on it again.

Our other rules: How we find tools · How we categorize · How we rank · How we pick awards. All five are on How we work.