An AI tools directory looks simple: a name, a short description, a price and a link. But every one of those fields compresses a decision, and a reader who misreads them can pay for features they cannot use, pick a product that does not run on their device, or trust a listing that went stale months ago. This guide walks through the fields in a typical directory entry on this site and explains how to use each one, and where a directory stops being useful and you have to try the product yourself.
A directory is not a review
Start with the most important distinction. Our AI Tools Directory lists products with factual information: category, pricing model, platforms and official links. A directory entry that has not had a hands-on review shows “Not rated yet” rather than an estimated score. That is deliberate. A description such as “records, transcribes and summarizes meetings” tells you what the vendor says the product does. It does not tell you how accurate the transcripts are for your accents, your jargon or your audio quality. Our testing methodology explains what a rating means when one does appear.
Pricing labels
Entries use four pricing labels: free, freemium, paid and open-source. Each hides different questions.
Free should mean no payment is required for the core product, but check whether it is free forever or free during a launch period. Freemium means a free tier exists with limits. The limits are the real product for many people: caps on minutes, documents, messages or model quality can decide whether the free tier is usable at all. Paid usually means a subscription or usage-based billing; look at whether the price is per seat, per month, or per token or credit. Our guide to how AI pricing works covers credits and tokens in more depth. Open-source means the code or weights are available, but running them can still cost money for hardware or a hosting provider, and licenses vary on commercial use.
The optional “price from” field is a starting point, not a quote. It typically reflects the cheapest paid plan, billed in the most favorable way, and may exclude taxes, extra seats or usage overages. Always open the vendor’s pricing page, for example Otter.ai pricing or Grammarly plans, before you decide.
Platforms
The platforms field lists where the tool works: web, Windows, macOS, Linux, iOS, Android, API, browser extension, CLI, IDE plugin or self-hosted. Read it with your own setup in mind. A “web” tool works in any browser but may not integrate with your desktop apps. A “browser-extension” tool may only work in one browser and often reads page content, which has privacy implications. “API” means you will need to build something or use a no-code connector. “Self-hosted” means you operate it.
If a listing includes both a mobile app and the web, that does not guarantee feature parity; many products ship new features on one platform first.
Subcategory and what it implies
Each tool sits in one subcategory such as writing, meetings and notes, design or coding assistants. The label shows the primary job. Many products now span several, so a transcription tool may also generate summaries, action items and chat over past meetings. Use the subcategory to find peers, then compare feature lists directly rather than assuming the label captures everything.
Pros, cons and alternatives
Where an entry lists pros and cons, treat them as starting points for your own test. An empty list does not mean the product has no weaknesses; it means nobody has recorded any yet. The alternatives field helps you build a shortlist of two or three products to trial side by side. For a method, see our guide on how to evaluate an AI tool alternative before switching.
Last verified and added dates
Two dates matter. “Added” tells you when the entry first appeared. “Last verified” tells you when someone last checked that the basics, such as the website, category and pricing model, were still correct. AI products change quickly: names, plans and free tiers are revised often. If the verification date is more than a few months old, treat the pricing and feature claims as likely out of date and confirm them on the vendor site.
Claims a directory cannot settle
Some questions only a trial can answer: how accurate the outputs are on your material; how long responses take at peak times; whether the product keeps your data and uses it for training; and how well support responds. Frameworks such as the NIST AI Risk Management Framework offer a vocabulary for thinking about these risks. Treat superlatives such as “the most accurate” or “the fastest” in any listing with caution unless the vendor shows how it measured them.
A five-minute routine
- Read the description and subcategory to confirm the product does the job you need.
- Check platforms against the devices you use.
- Compare the pricing label with the vendor’s own page and note the limits of any free tier.
- Look at the last verified date and decide how much to trust the details.
- Pick two alternatives and run the same real task through each.
What to watch next
- More directory entries gaining full reviews and ratings.
- Vendors changing free-tier limits as inference costs shift.
- Standard ways of disclosing data use and model providers in listings.
- Consolidation, as meeting, writing and coding tools add overlapping features.
Key takeaways
- A directory entry gives facts and vendor descriptions, not a quality verdict.
- Pricing labels hide limits; open the vendor page before you buy.
- Platforms and integrations decide whether a tool fits your workflow.
- The last verified date tells you how fresh the details are.
- Always test two or three finalists on your own material.
Frequently asked questions
Why do some entries have no rating?
Ratings come only with a full published review. Entries without one show no score rather than a guessed one.
Is “free” really free?
Check the vendor’s page. Many free plans are capped on usage or features, and some are time-limited.
How often should I recheck a listing?
Verify pricing and terms again whenever you are about to pay, and treat entries not verified within a few months as possibly outdated.
