Every post on AI Tool Analysis answers a practical question for someone about to pay for an AI tool: what does it cost, what do you get, and which plan fits you? This page explains where our facts come from and how we keep them current.
We are a research site. We don’t run lab tests or our own benchmarks, and we don’t describe hands-on results. What we do is read the vendors’ own pages closely, compare plans and rivals side by side, and explain what the fine print means for an ordinary user.
Where our facts come from
We sort sources into three groups and use each one for a different job:
- The vendor’s own pages: pricing pages, help centre articles, documentation, terms and launch posts. Every price, plan limit and feature claim has to rest on one of these.
- Major tech press and the vendor’s official blog or social accounts: useful for context, timelines and what changed, and for cross-checking.
- Blogs, video roundups, review sites and community threads such as Reddit: used only for opinions and user reports, always attributed (“subscribers on Reddit report…”), and never as the only source for a fact.
Every claim a post depends on needs two sources, and at least one of them has to be the vendor’s own page.
How a post comes together
- One reader, one decision. Each post starts from a specific reader and the choice they are making, such as a student deciding between free and paid Gemini, rather than from a template.
- Read and save the sources. Every page we quote is saved with its address and the time we read it, so the post can be checked against exactly what the vendor said.
- Log every fact. Each fact a reader will see, in the text, a table, a chart or an FAQ, gets a row in the post’s claims record: the value, the source, the quoted line it comes from, the date we checked it and when it should be checked again.
- Write from the record. Numbers in tables and charts come from the same rows as the text, and any sum or comparison shows its working.
- Check before it goes out. Automated checks confirm that every quoted line appears on a saved source, that sums add up and that a price is the same everywhere it appears in the post. A separate fact-check pass then reads the draft against the sources and lists anything that doesn’t match, and those items are fixed before the post is published.
Prices and plans
- We quote prices in US dollars as the vendor lists them, and we say whether a price is monthly or billed yearly. Prices in your country, taxes and app-store prices can be different.
- We keep consumer plans and API prices apart, because they are different products with different bills.
- When a plan is sold as “unlimited”, we look for the fair-use terms behind the word.
- When two of a vendor’s own pages disagree, we show both with their dates instead of quietly picking one.
Screenshots, images and charts
- Screenshots show the vendor’s own page, carry the date they were captured in the caption, and are never edited.
- Photos are stock photos from Pexels, credited where they appear.
- Charts are drawn from the same recorded facts as the text.
- We don’t add AI-generated images to new posts.
Keeping posts current
AI prices and plans change often, so a post is never finished. We monitor the pricing and plan pages of the main tools we cover, and a change has to show up twice before we act on it. Prices that several posts share are kept in one place and checked again against the vendor’s pages on a regular schedule.
When something changes, we make the smallest change that makes the post correct again:
- correct the fact in place, when one or two numbers moved;
- rewrite the sections that changed, when a plan or feature was reshaped;
- rebuild the whole post, when most of it is out of date;
- write a new post, when the old one no longer answers the question people are asking.
Scores and verdicts
Some posts give a score out of 5. It is our judgement of value for money for the reader the post is written for, based on the recorded facts. It is not a measurement, and we don’t publish user-rating counts because we don’t collect user ratings. A verdict such as “worth it for a student, not for a team” always comes with the reasons behind it.
What we don’t do, and our limits
- We don’t present benchmark scores as our own. Benchmark numbers come from the vendor or a named third party and say so.
- We don’t invent people. Examples are written as hypotheticals (“say you use it for…”), not as a real person’s story.
- We don’t accept payment for reviews or for favourable coverage.
- We can’t tell you how a tool will perform on your own work. We report what the vendor documents and what users report, and we say which is which.
- Prices can change between our last check and your visit. Check the date on a fact, and the vendor’s page, before you buy.
This method applies to everything we have published or updated since October 2026. Some older posts still describe hands-on use in the first person, from before we changed how we work; we are correcting them as each one comes up for update.
Found a mistake?
If a price is out of date or a fact is wrong, tell us through the contact page or email aitoolanalysis@gmail.com. We’ll check it against the vendor’s page. You can read more about who runs the site on our About page.