The pricing transparency problem
94% of the AI tools in our source data claimed to be free. Auditing them by hand showed why that number is meaningless — and what we changed because of it.
When we imported the dataset behind this directory, 4,647 of 4,921 tools (94%) were marked free. That figure should not be possible in a market where most of these products are venture-funded SaaS. So we checked, and the explanation was simple: the source recorded "free" whenever no pricing had been captured. Missing data was being displayed as a fact.
This post documents what we found and the rules we now apply, because the same flaw sits in most AI tool lists you will read this year.
What the audit found
We pulled a sample of tools carrying the free label and opened each pricing page by hand. Four patterns covered nearly all of them.
| Pattern | What the listing said | What the pricing page said |
|---|---|---|
| Free trial, not free | Free | 7 or 14 days, card required |
| Freemium with a hard cap | Free | Free tier exists but is unusable past a few generations |
| Genuinely free | Free | Open source or fully free, no paid tier |
| No public pricing at all | Free | "Contact sales" |
Only the third pattern deserved the badge. The rest were somewhere between misleading and wrong, and the fourth is the most expensive kind of wrong — enterprise contracts presented as zero cost.
What we changed
We stopped treating the default value as information. Three concrete rules now govern pricing display across the site.
- Only verified tiers render a badge. That means freemium, paid and enterprise — 274 listings today, out of 4,921.
- Unknown pricing shows nothing. No badge, no "free", no placeholder. An absent badge means we have not verified it.
- Verified values are editable. Corrections are applied through our admin layer and take effect immediately, without waiting for the next full data import.
The visible consequence is that most listings on this site now show less pricing information than competing directories show. That is the point. A blank field you can trust is more useful than a filled field you cannot.
If a directory tells you 90% of AI tools are free, it is not describing the market. It is describing its own missing data.
Why this keeps happening across the industry
Three structural reasons, none of them malicious.
- Scraping rewards completeness over accuracy. A field with a default value looks better in a database than a null, and defaults are never revisited.
- Pricing pages are hostile to machines. Usage-based tiers, per-seat minimums, annual-only discounts and "contact us" walls do not reduce to one number.
- Nobody audits downward. Directories add tools continuously; almost none re-check the ones already listed.
How to read pricing on any AI tool listing
Portable habits, whichever site you are on.
- Treat "free" as a claim to verify, never as a fact — click through to the pricing page before shortlisting.
- Look for the card requirement. "Free to start" plus a card is a trial.
- Check the unit. Per seat, per generation, per minute of audio and per 1,000 words produce wildly different bills at the same headline price.
- Find the ceiling of the free tier. If it is 10 generations a month, it is a demo.
- For anything "contact sales", assume an annual commitment and a floor of several thousand.
What we still do not know
Being explicit about the gaps: we have verified pricing for about 6% of listings. The remaining 94% are unknown, not free. We are expanding coverage category by category, prioritising the most-viewed listings first, and we keep pricing corrections open — if you find one wrong, tell us and we will fix that record directly.
Related reading
By Editorial team
Weekly AI tools digest
New launches, honest pricing notes and the tools we actually kept using.
Prefer a reader? Subscribe via RSS.