Skip to main content
Ranking criteria

How we rank AI tools

Six criteria, applied the same way to every tool. We publish them so you can challenge any recommendation.

Real price

The actual monthly cost from the vendor's pricing page, not a promotional or annual-only rate.

Business fit

Does the tool do the job this business type needs? A great coding tool isn't a great fit for a real-estate agent.

Setup effort

How hard is it for a non-technical owner to get value? Tools that need a developer rank lower for non-technical audiences.

Reliability

Is the vendor established and the product stable? New, unproven tools are flagged as 'Test' rather than 'Switch to'.

Total stack cost

A tool isn't ranked in isolation — we check it fits the total monthly budget alongside the other picks.

No paid placement

Vendors cannot pay to move up. Affiliate commissions, where they exist, never change the ranking.

Being in the catalog is not the same as being recommended

Every tool carries one of four statuses. Only the first can be a primary pick in a plan — that isn't a policy we describe, it's a filter in the code that builds your stack, so a tool on watch cannot reach your plan even if it scores well on the six criteria above.

  1. RecommendedEstablished, verified, and eligible to be a primary pick in a plan. It has a clear job in a real workflow and we can explain what it can't safely do.
  2. Listed onlyReal and worth knowing about, but we don't put it in plans — usually because something else covers the same job better, or because its pricing is hard to pin down. Every newly added tool starts here.
  3. On watchPromising, and we're following it. New products stay here until they show enough operating history that recommending one wouldn't be a bet with your money.
  4. ArchivedNo longer maintained by its vendor, or superseded. Kept visible so an old recommendation doesn't quietly disappear on you.

What we record before a tool can be recommended

  • An official pricing page we can check, and the date we last checked it.
  • The date we last reviewed what the product actually does — separate from its price, because those change independently.
  • How confident we are in the evidence, and why: high, medium or low.
  • A concrete workflow and business type it belongs to. "It's good" is not a reason.
  • What it can't safely do, and what data should never go into it.
  • The total cost of running it — including hosting and upkeep for anything self-hosted, because free software isn't free to operate.

Each tool's page shows its own status and dates, so you can check any of this yourself rather than taking our word for it. Browse the catalog