How We Review AI Tools
Most AI tool reviews online share two problems. They go stale within weeks, because AI vendors rename plans and move prices constantly. And many of them describe “tests” that never happened, complete with suspiciously precise benchmark numbers and screenshots that could have come from anywhere.
We built AI Tools Examiner around a different promise: every claim we make can be traced to a source you can check yourself. This page explains exactly how that works, what our scores mean, and where the limits of our method are.
What our reviews are based on
Our research starts with primary sources, in this order:
- Official pricing pages. We record every plan, its monthly price, its annual-billing price where offered, seat minimums and usage limits.
- Help-center and product documentation. This is where vendors spell out what a plan really includes: message caps, credit systems, minutes per meeting, integrations and admin controls.
- Changelogs and release notes. These tell us what shipped recently and what was quietly removed.
- Security, compliance and legal pages. Data retention, training on customer data, SSO and HIPAA availability, and refund terms.
- Verifiable public specifications. Supported platforms, file limits, languages and integrations that the vendor publishes.
When a vendor’s pricing page can’t be read directly, which happens when pages block automated access or only show prices after you click a toggle, we cross-check the figures against several recent independent sources. Those prices are marked with an asterisk in the pricing tracker so you know to confirm them before buying.
What we don’t do
This is the part most review sites skip, so we’ll be explicit.
- We don’t claim hands-on testing unless it happened. If a future article includes our own testing, it will say what we tested, on which plan, on what date and how. Absent that statement, assume the review is research-based.
- We don’t invent benchmarks. You won’t find “Tool X was 37% faster” on this site unless the number comes from a cited, public source.
- We don’t fabricate quotes or user reviews. No anonymous “Sarah, marketing manager” testimonials.
- We don’t use fake screenshots. Illustrations on this site are original graphics. We don’t use vendor logos beyond plain text names.
- We don’t let money touch the scores. Affiliate commissions fund the site, but a tool’s affiliate status is never an input. See our affiliate disclosure.
The scoring rubric
Every tool gets five sub-scores from 1 to 10. The overall score is a weighted average, rounded to one decimal place.
| Criterion | Weight | What we look at |
|---|---|---|
| Features | 30% | Breadth and depth of capabilities for the tool’s core job, model access, integrations, platform coverage, admin and collaboration features |
| Pricing | 20% | Entry price, availability of a usable free tier, price ladder between plans, hidden meters, seat minimums, billing flexibility |
| Ease of use | 20% | Onboarding as documented, clarity of plan limits, number of surfaces (web, desktop, mobile, browser extension), learning curve of the core workflow |
| Support | 10% | Documented support channels by plan, help-center depth, status page, response commitments, community resources |
| Value | 20% | What you get per dollar compared with direct alternatives at the same price point |
Why these weights
Features get the most weight because a cheap tool that can’t do the job is no bargain. Pricing and value are separate on purpose: pricing measures how fair and transparent the price structure is, while value compares what you get against competitors. A $200 plan can have a modest pricing score and still be strong value for the right user. Support carries the least weight because, for most AI tools, documented support is similar across vendors at the same tier, and it matters most for team buyers.
How sub-scores are set
Sub-scores are relative within a category. A 9 for features in meeting note-takers means “near the top of what meeting note-takers offer,” not “as many features as a general chat assistant.” That keeps scores comparable when you read a roundup or a comparison.
We store every tool’s sub-scores in a single data file. Verdict boxes, comparison tables and roundups all read from that file, so a tool can never be scored differently on two pages. When we change a score, it changes everywhere at once.
What the numbers mean
| Score | Meaning |
|---|---|
| 9.0 to 10 | Category leader with few meaningful weaknesses |
| 8.0 to 8.9 | Excellent; the right pick for many buyers |
| 7.0 to 7.9 | Good, with clear trade-offs to weigh |
| 6.0 to 6.9 | Usable, but most buyers have better options |
| Below 6 | Hard to recommend at current pricing |
How we handle pricing
Pricing is where AI tool content goes wrong most often, so we follow strict rules:
- List prices in US dollars, before tax. Regional prices and promotions vary; we note promotions only when they’re broadly available.
- Monthly versus annual is always labeled. In our tables, “Monthly” means billed month to month. “Annual” means the per-month equivalent when you pay for a year up front. The difference is often large: for example, Grammarly Pro is $12 a month billed annually but $30 billed monthly.
- Seat minimums and meters are called out. A “$25 per seat” team plan with a two-seat minimum is really a $50 minimum. Usage-based pricing, like per-resolution fees for customer support AI agents, is shown with its unit.
- Every price has a date. The “last verified” date on every page and tracker row tells you when we last checked. If you find a newer price, tell us and we’ll re-verify.
How comparisons and roundups work
Comparisons put two or three tools side by side using the same rubric. We name a winner in each category, then give a final verdict that explains which tool fits which kind of user. A tool can lose most categories and still be the right choice for you, and the verdict says when that’s the case.
Best-of roundups start with the job to be done, such as running a small business, supporting customers, or writing marketing content. We shortlist tools that are built for that job and publicly priced, then rank them using the rubric, with adjustments explained in the text when a tool’s overall score doesn’t reflect its fit for that specific use case.
How we choose which tools to cover
We prioritize tools that are widely used, publicly priced, and actively maintained. We also favor tools with public affiliate programs when choosing between equally relevant products to cover next, because that’s how we fund the site. That affects which tools we write about first, never how they score. Some of the highest-scoring tools on this site, including ChatGPT, Claude and Gemini, pay us nothing.
Updates and corrections
AI tools change constantly, so our articles do too. When a vendor changes pricing or plans, we update the data file, which updates every page that mentions that tool. When we make an error, we correct it and, if it changed a recommendation, add a note to the article.
Who does the work
Every article has a named author who is accountable for it. John Marshall covers chat assistants, writing tools and pricing. Craig Leslie covers team tools: meeting assistants, customer support AI and marketing platforms. Read more in our editorial policy.
Where to start
- Compare current plans across 27 tools in the AI pricing tracker.
- Browse best AI tools by use case.
- See head-to-head comparisons.
FAQ
Do you test every AI tool hands-on?
No, and we say so. Our reviews are based on official documentation, pricing pages, changelogs, help-center articles and published specifications. If an article includes hands-on testing, it states exactly what was tested and how. We never imply testing that didn’t happen.
How is the overall score calculated?
The overall score is a weighted average of five sub-scores: features 30%, pricing 20%, ease of use 20%, support 10% and value 20%. It is rounded to one decimal place and calculated automatically from the same data file everywhere it appears.
Can vendors pay for a higher score?
No. Vendors cannot buy coverage, placement or scores. We earn affiliate commissions from some links, but affiliate status is not an input to the rubric, and several top-scoring tools have no affiliate program.
How often do you update prices?
We re-verify prices on a rolling basis and whenever a vendor announces a change. Every review, comparison and tracker row shows the date its price was last checked.
What does the asterisk in the pricing tracker mean?
It marks prices we could not read directly from the vendor’s pricing page, usually because the page blocks automated access or hides prices behind a toggle. Those figures are cross-checked against multiple recent independent sources, and you should confirm them on the vendor’s site.