About AI Tools Police
AI Tools Police is an independent AI publication. We investigate AI tools on evidence: verified pricing, documented features, and what real users report. This page sets out what we cover, how we work, who runs it, and how we are funded.
What AI Tools Police is
AI Tools Police is an independent AI publication. We investigate AI products, the companies that sell them, and what both mean for the people paying — and we publish what the evidence supports. Our promise to readers is the whole job in one line: we investigate AI so you can decide with confidence.
The principle underneath that is signal over noise. This market produces far more announcements than facts. Landing pages promise capabilities that documentation does not confirm, prices move without notice, and free tiers stop being useful at a point nobody advertises. Our work is to find the part that changes your decision and state it plainly, including when the honest answer is that a tool is not worth its price.
In practice that comes down to four commitments on every page. Pricing is checked against the vendor's own live pricing page and dated, so you can see how fresh the number is. Features come from official documentation, changelogs, and vendor support pages, not from marketing adjectives. Experience is aggregated from independent platforms where real users write, including G2, Trustpilot, Reddit, and Capterra. And the verdict says only what the evidence supports: verified pricing, real limits, honest verdicts.
Our method is verification-based, and we say so openly. We confirm what can be confirmed against primary sources, we weigh independent user reports for the rest, and we label every claim by how we know it. Nothing is presented as more certain than the evidence behind it, and a gap in the evidence is published as a gap.
What we cover today
Today AI Tools Police carries 77 published tool investigations across 15 category rankings. The coverage is deliberately concentrated on the categories where a buying decision has real money behind it, plus a few free utilities readers ask us to check.
| Area | Rankings we publish |
|---|---|
| Video | Faceless video tools, AI video generators, AI avatar generators |
| Audio | AI voice generators, AI music generators |
| Images and design | AI image generators, AI design tools, AI headshot generators |
| Text | AI humanizers, AI content detectors |
| Work and business | AI resume builders, AI website builders, AI presentation makers, AI data analysis tools |
| Web publishing | AI bot checkers |
Every investigation carries a last-updated date, and the original publication date is recorded and never changed. A review is a dated snapshot, and we treat it as one: we return to a page when a tool's facts move, and the verdict moves with the facts. Some of our rewrites have lowered a rating rather than raised it, because that is what the evidence said.
We do not publish everything. We publish what matters. A category ranking goes live when we have enough verified evidence to order it honestly, and a tool joins a ranking on merit or not at all.
How we work
The full method is public. Each part of it is written down so you can check our work against our own stated rules.
Method
Every investigation follows the same route: research the tool, its category and its alternatives; write to a clear verdict that states who the tool is for and where it breaks; verify every number before publication; then publish it dated and maintain it. If a figure cannot be confirmed, it does not go on the page as fact.
How we investigate →Editorial principles
Eight principles govern what we publish: truth over speed, evidence over opinion, research over repetition, quality over quantity, transparency over marketing, independence over sponsorship, clarity over jargon, and correction over defensiveness.
Editorial manifesto →Evidence labels
We say how we know each claim, in one of five ways. Verified means confirmed against an official document, a live pricing page, or a policy. User-reported means the pattern appears across multiple independent user sources. Vendor-reported means it is the company's own claim with no independent confirmation. Estimated means it is our calculation from stated assumptions. Unknown means we could not verify it, and we would rather print that word than a confident guess.
How we label evidence →Sources
We rank sources before we use them. Primary documents come first: official documentation, live pricing pages, legal texts, and filings. Reputable independent press and published independent measurements provide context. User signal from review platforms shows experience patterns but is never treated as settled fact on its own. Promotional material is a starting point for discovery and never a final source.
Sources & citations →Scoring
Every review of a tool still in operation carries an overall score out of 5 plus four to six subscores chosen to fit the category, because the things that matter in a voice tool are not the things that matter in a website builder. Two dimensions are weighed everywhere: economic reality and trust. Scores are capped by evidence, so a top pick usually lands mid-to-high rather than flawless.
How scoring works →Corrections
Anyone can request a correction, reader or vendor. We grade them by severity: typos are fixed quietly, factual errors on price, features or availability get a dated note on the page, and an error that changes a score or a recommendation gets a prominent note and an updated verdict. Vendors may report a factual error at any time. They may not preview, edit, or approve a review.
Corrections policy →How we use AI
We are an AI publication that uses AI in production, and we disclose exactly where. AI assists with research and source discovery, drafting and structure, data cleaning and cross-checking, and language editing. It is never used to invent a source or a quote, to fabricate first-hand experience, or to publish anything without human editorial review.
AI transparency →Where this is going
Tool investigations are the foundation of this publication, not the ceiling. We have started rebuilding AI Tools Police into a fuller publication on the same evidence rules, and we are doing it in the open, section by section. This is the plan, in the order we are executing it:
| What | Where it stands |
|---|---|
| Reviews and rankings | Live today: 77 investigations across 15 rankings, dated and maintained. |
| Original research | Live today: our first study audits what 76 consumer AI products publish about their own pricing and cancellation terms, from vendor documents only. The paper, the full dataset, the codebook and the log of everything we got wrong are all published together. |
| Analysis | Live today: where Research measures, Analysis argues. Each piece states its position in one sentence a reader can disagree with, argues it from figures traceable to our own published data, and ends with the evidence that would change our mind — which is a required field, not a good intention. |
| News | Planned, and deliberately selective: only developments that change a reader's decision, never volume for its own sake. |
| Academy | Planned: structured, leveled learning for readers starting from zero. |
| Guides | Live today: 4 outcome-driven how-tos, each stating its time and cost before the first step and ending with a way to check the result yourself. |
Each section appears on this site when it is ready, not before, and this page changes as each one ships. The research is the nearest, because its raw material already exists: every review here checks plans, credits, and renewal terms against the vendor's own live pricing page, and across 77 investigations that has become a running record of how the AI tool market actually behaves.
Reviews keep running through all of it. They are the engine that funds the rest, and nothing about the wider ambition loosens the rules they are held to.
Who runs it
Mucahit Kaya
Founder & lead reviewer
AI Tools Police was founded by Mucahit Kaya, who runs it as founder and lead reviewer. He tracks the AI creator-tool space daily, and every review here digs into verified pricing, documented features, and what real users report rather than a rewrite of the marketing page. Editorial fact-checking is systematic and happens before anything goes live: pricing, features and claims are all checked, and a page that cannot clear that check does not publish.
This is a small operation rather than a newsroom, and we would rather state that than imply otherwise. The trade is a real one. We publish less than a large site would, and in return every verdict on this site has a named human standing behind it.
How we are funded
AI Tools Police is reader-supported. Some of our links to the tools we cover are affiliate links, which means we may earn a commission when a reader buys through them, at no extra cost to the buyer. That is the main way the publication pays for itself.
What that money does not buy is the ranking. Commission rate and affiliate status never affect a score, a ranking position, a "best for" pick, or whether a tool is covered at all. Good tools with no affiliate program are reviewed on exactly the same terms as tools that pay well, and the disclosure appears site-wide and next to affiliate links.
Two paid options are published on our submit page. A $49 one-time expedited investigation buys a place at the front of the queue and nothing else; the rating and ranking are decided exactly as they would be for a free review, and an expedited investigation can and does score low when the evidence says so. A $99 sponsored placement, running for 30 consecutive calendar days, buys a slot that is labeled as sponsored everywhere it appears and kept separate from the editorial rankings. Neither option buys a score or a position. The full policy, including our stance on conflicts of interest and on free access provided by vendors, is set out in our affiliate standards.
Contact
Write to info@aitoolspolice.com. That address covers corrections, tool submissions, and anything else you want to raise with us.
Corrections are especially welcome, and we would rather hear about a mistake from you than leave it published. If a price is stale, a feature has changed, or a claim on one of our pages does not hold up, tell us what is wrong and where. We fix it, we date the fix, and if the fix changes the verdict then the verdict changes too. Accuracy matters more to us than having been right the first time.