🔍 AI Tool Reviews: How to Tell Who Paid for the Ranking
Most 'best AI video editor' lists are written by a company on the list. Here is a two-minute check that tells you who paid, plus what the law requires.
Sam Whitaker · Developer & API Cost Writer
· 12 min read
⚡ TL;DR — quick answers
- How can I tell if a 'best AI tools' list is an advertisement?
- Ask three questions in order: does the publisher's domain match a tool ranked on the page, does any tool on the list actually lose, and does a Ctrl-F for 'we may earn' or 'compensation may impact' find anything. A page that ranks its own product, gives every tool a winning category and hides its money relationship is an ad with a headline. It can still be useful for the competitor sections, but its ranking carries no information.
- Is it illegal for a company to rank its own product first?
- No, not on its own. 16 CFR 465.6 targets the misrepresentation, meaning a business presenting a site it controls as an independent review source. A clearly disclosed self-review is legal, and MemoTune's list does exactly that by stating outright that MemoTune is its own product. What crosses the line is the same ranking styled as neutral third-party research with no ownership disclosure anywhere on the page.
- Does an affiliate disclosure mean the review is honest?
- It means the page met a legal minimum, nothing more. Disclosure tells you a financial relationship exists; it says nothing about whether that relationship shaped the order. GetVoIP is unusually direct about this, stating that compensation impacts the order in which products appear within listing categories. Treat a disclosed page as a downweight rather than a discard.
Key takeaways
- Search 'best AI song generator' and seven of the nine first-page results are published by a company that ranks itself on its own list, with two of those pages naming different, mutually exclusive winners.
- The fastest tell is not tone but structure: if every tool on the page gets a 'Best for' label and the top pick has no limitations section, the categories were reverse-engineered from the list rather than found by testing.
- 16 CFR 465.6 has banned misrepresenting a site you control as an independent review source since October 2024, and 465.1 defines a clear disclosure as one that is difficult to miss, so a disclosure you had to hunt for fails by definition.
- A disclosed conflict is still a conflict: read vendor listicles for the competitor limitations sections, which are usually the best-researched part of the page, and ignore the ranking entirely.
Search for "best AI song generator" and look at who owns the pages that come back. On the check I ran this month, seven of the nine first-page results were published by a company that appears on its own list. suno.com puts Suno at number one. memotune.com puts MemoTune at number one. Two pages, two winners, mutually exclusive, both ranking on the same query.
Nobody on that page is lying, exactly. Suno probably is the best AI music generator for a lot of people. The problem is that the ranking contains no information, because the outcome was fixed before the testing started, and nothing on the page tells a reader that.
Every guide about spotting fake reviews is written for a different format. Check the Verified Purchase badge. Watch for review clusters posted on the same day. Look for accounts with one review and no history. None of that helps against a 4,000-word listicle titled "10 Best AI Video Editors in 2026." So here is a method that does, and it runs in about two minutes.
The two-minute method
Three questions, in order, plus one optional. Answer each in ten seconds, add up the points, and stop.
- Who publishes this page? Does the domain match a tool ranked on the list?
- Who loses? Which tool does the author tell you not to buy, and for whom?
- Who gets paid when you click? Ctrl-F for the boilerplate.
- How old is this, really? Find one version number and one dated price.
Scoring runs 0 to 11 points, and the bands are simple. Six or more: discard the ranking. Three to five: downweight it, read the competitor sections, ignore the order. Zero to two: read it like a normal article and check the claims you care about.
By the numbers
- 7 of 9 first-page results for "best AI song generator" are published by a company that ranks itself on the list, including suno.com, memotune.com, musicgpt.com and musegen.ai (our SERP check, August 2026).
- Suno #1, Udio #2, ElevenLabs Music #3 on suno.com's own "Best AI Music Generator" hub, which calls Suno the best platform to create music with AI in 2026 and carries no ownership disclosure (Suno).
- 5 of 6 entries on Reap's "Best AI Video Editing Tools" have a "Limitations" subsection. The one that does not is Reap, ranked first, whose section is headed "What Reap does exceptionally well" (Reap).
- 9 tools, 9 "Best for" labels, 0 losers in Buffer's "The 11 Best AI Video Editors in 2026," from Vyra "Best for: Handing the whole edit to an AI" to Adobe Express "Best for: People already in the Adobe ecosystem" (Buffer).
- 4% of annual turnover is the floor the EU sets for maximum fines on widespread infringements, and undisclosed paid ranking is banned in all circumstances under Annex I point 11a of the UCPD (EUR-Lex).
- 21 October 2024 is when 16 CFR Part 465 came into force, including the rule on company-controlled sites presenting themselves as independent (Legal Information Institute).
Check 1: who publishes this page?
Read the domain in the address bar, then read the list. If they match, you are done thinking about the ranking.
The ten-second version has four parts. Does the domain name appear as a ranked entry? Does that same domain have a /pricing, /signup or /app route? What does the footer copyright say? Is there an About page that names a product rather than a masthead?
The live cases are not subtle. suno.com/hub ranks Suno first. reap.video/blog ranks Reap first. chatcut.io ranks ChatCut first out of fifteen, ahead of Descript, CapCut, Premiere Pro, DaVinci Resolve and Final Cut Pro. memotune.com ranks MemoTune first out of six. In three of those four cases there is no ownership disclosure on the page at all.
This tell is worth three points on its own because it is binary, it takes one glance, and it invalidates the single thing you came to the page for.
Check 2: who loses in this list?
A review distributes bad news. A catalogue does not. So ask the page a question it usually cannot answer: which tool here would the author tell me not to buy, and for which kind of user?
Buffer's list assigns a "Best for" to all nine tools it ranks. Nine tools, nine winning categories, nobody left over. When N tools produce exactly N winning categories, the categories were reverse-engineered from the list rather than derived from testing. That page is otherwise decent, which is the useful part of the example: it names a specific 81-second test clip and reports that one tool's edit pacing stayed slack when the author asked for snappy. The everyone-wins structure and real testing can coexist. The structure still tells you the list was built to include, not to choose.
Our own AI video editing tools compared exists partly because a list that recommends everything is a directory with adjectives.
The difference between a real con and a fake one
This is where most readers get played, because pages that suppress criticism rarely leave the cons section empty. They fill it with something that costs the vendor nothing.
A real limitation is falsifiable and expensive to admit. "Watermarked videos." "Desktop version lags behind the mobile app." Both of those are ChatCut's description of CapCut, and both are checkable in ten minutes. A fake limitation is a scope statement wearing a weakness costume. ChatCut's own listed drawback is that it is "not suited for complex narrative editing that requires fine-grained timeline control," which translates to: we are focused. Nobody was ever talked out of a purchase by that sentence.
The structural version of the tell is cleaner still. On Reap's list, Submagic, OpusClip, CapCut, Descript and Vizard each get a "Limitations" subsection. Reap gets "What Reap does exceptionally well." Same page, same template, one exemption. That asymmetry is the mechanism, and it does not require anyone to write a false sentence. A sponsored reviewer's honest option is silence about weaknesses, not invention.
Check 3: who gets paid when you click?
Ctrl-F, then read what you find. The strings worth searching are short and standard: "we may earn," "compensation may impact," "how and where," "in what order," "advertiser disclosure," "this post contains affiliate links."
They hide in predictable places. Footer. A link labelled "Disclosure." A collapsed grey banner sitting above the comparison table that reads as furniture. Amazon Associates dictates its own wording, so "As an Amazon Associate I earn from qualifying purchases" is a required string rather than a stylistic choice, which makes it easy to grep for.
The honest version of this language exists. GetVoIP's disclosure page states that the site is advertiser-supported and that "This compensation impacts how and where products appear on this site, including, for example, the order in which they may appear within listing categories," then adds that the order is not to be mistaken for a ranking or an endorsement. That is a comparison site telling you its table is an ad. It is one click from the article, and no guide in this niche tells anyone to make that click.
What the law actually says, in reader terms
If the statute language is heavy going, this walkthrough covers the same ground in plain terms — what a compliant disclosure has to say and where it has to sit:
Three rules matter, and all three are short.
16 CFR 255.5 says that when a connection between an endorser and a seller might materially affect the weight or credibility of the endorsement, and the audience would not reasonably expect it, that connection must be disclosed clearly and conspicuously. Example 11 in that section is the affiliate blogger earning commission per sale. That is not an analogy; it is the exact case.
16 CFR 465.1 defines clear and conspicuous as "easily noticeable (i.e., difficult to miss) and easily understandable by ordinary consumers." Read that as a reader's test rather than a lawyer's. If you had to hunt for the disclosure, it failed the standard by the standard's own wording.
16 CFR 465.6, in force since October 2024, makes it a deceptive practice for a business to materially misrepresent that a site it controls "provides independent reviews or opinions ... about a category of businesses, products, or services including the business or one or more of the products or services it sells." Note what this does not say. Ranking your own product first is not banned. Doing it while presenting the page as independent research is the violation. The distinction matters, and most marketing coverage of these rules skips it.
Outside the US, briefly. In the EU, Directive 2019/2161 added point 11a to Annex I of the Unfair Commercial Practices Directive, making undisclosed paid ranking a practice banned in all circumstances, with maximum fines for widespread infringements set at no less than 4% of turnover in the member states concerned. In the UK, the fake and concealed incentivised review provisions of the DMCC Act 2024 have applied since 6 April 2025.
Disclosure is not objectivity
MemoTune's page discloses the conflict in plain words: "We're upfront: MemoTune is our product. But we genuinely believe the best way to earn trust is to show you exactly how every tool performs." Then it ranks MemoTune first of six, ahead of Suno, Lyria 3, Beatoven.ai, AI Song Maker and Mureka.
A reader taught only "look for a disclosure" passes that page. The disclosure is real, legal and better than what most of the cluster does. It is also not neutrality, and no amount of good faith makes a self-ranking a comparison.
The practical move is to downweight rather than discard. Vendor listicles are usually researched hardest in the sections about competitors, because that is where the page has something to prove. Read those. Ignore the order. The same logic applies to the competitor cons on our own best AI music generators list.
The freshness check that only AI tools need
Model versions ship monthly. Suno v5.5, Stable Audio 3, Seedance 2.5. A ranking that names no versions is comparing tools that no longer exist.
So: find one version number and one dated price on the page. If the article carries "2026" in the title, no per-claim date, no version numbers and no "checked on" stamp for pricing, treat everything in it as unverified regardless of who wrote it. This is also the check that costs vendors the least to pass and that almost none of them bother with, which is why it separates pages so cleanly.
The scorecard
Copy this, keep it in a note, run it on the next comparison page you land on.
| Tell | Ten-second check | Points |
|---|---|---|
| Publisher is on the list | Domain matches a ranked tool | +3 |
| Nobody loses | Every tool gets a "Best for" | +2 |
| Top pick has no cons | Only competitors get limitations | +2 |
| No disclosure anywhere | Ctrl-F "we may earn" finds nothing | +2 |
| No versions, no dates | No version number, no dated price | +1 |
| No raw output | No clips, screenshots or prompts | +1 |
Six or more, discard. Three to five, downweight. Two or fewer, read normally.
Run it on the "best AI song generator" results and the distribution is stark. Seven of nine score at least three points on the ownership row alone. Reap's video-editing list picks up three for ownership, two for the missing limitations section and two for the absent disclosure, landing at seven before anyone checks a version number.
What a review worth reading looks like
The positive checklist is shorter than the negative one, and it matches the standard Google publishes for its own quality raters. Google Search Central tells publishers to "Discuss the benefits and drawbacks of something, based on your own original research," to share quantitative measurements, and, when calling something best, to "include why you consider it the best, with first-hand supporting evidence."
Translated into things you can see on a page in one scroll: raw output you can inspect, the prompts that produced it, version numbers, prices with a date attached, a named human with a history, and at least one thing that went wrong. That last one carries the most weight, because it is the only item on the list that costs the author something.
When the page fails, triangulate instead. The tool's changelog tells you whether development is alive. The refund and cancellation policy tells you what happens when you want out. Searching the tool's subreddit or Discord for the word "cancel" surfaces the complaints that never make it into a listicle. And a competitor's comparison page will name the weaknesses the vendor's own page omits. Our free unlimited AI tools reality check was built the same way, from published prices rather than claims.
Where this site scores on its own scorecard
Here is the part that would be dishonest to leave out. aivideosensei.com makes money from affiliate commissions. We pay for the subscriptions we test, including the credit platforms we render on, and affiliate links carry rel="sponsored" with a labelled callout box and a standing affiliate disclosure and editorial policy.
So I ran the scorecard on our own best AI video generators post before publishing this one, and it scored three points. Downweight, by my own rule. It loses two for the everyone-wins structure, since every tool gets a job it is best at and no reader is ever told to skip one, and one more for pricing without a checked-on date. It passes the ownership row, since we sell no video model, and it passes the cons row, since the top pick carries its own weaknesses, moderation false positives and no native audio among them. But we do earn commission from a tool that appears on that list, and the disclosure box sits below the ranking rather than above it. By the plain wording of 465.1, a disclosure you have to scroll to find is a disclosure that failed.
That one is on me, and it is getting moved above the table this week. A test you refuse to run on yourself is not a test. It is a marketing asset with footnotes.
Sources & further reading
Outside figures cited above. First-hand test results are our own and noted as such in the text.
- 16 CFR § 465.6 - Insider consumer reviews and consumer testimonials — Legal Information Institute, Cornell Law School
- 16 CFR § 465.1 - Definitions — Legal Information Institute, Cornell Law School
- 16 CFR § 255.5 - Disclosure of material connections — Legal Information Institute, Cornell Law School
- Write high quality reviews — Google Search Central
- Directive (EU) 2019/2161 — EUR-Lex, Publications Office of the European Union
- Advertiser Disclosure — GetVoIP
Frequently asked questions
▸How can I tell if a 'best AI tools' list is an advertisement?
Ask three questions in order: does the publisher's domain match a tool ranked on the page, does any tool on the list actually lose, and does a Ctrl-F for 'we may earn' or 'compensation may impact' find anything. A page that ranks its own product, gives every tool a winning category and hides its money relationship is an ad with a headline. It can still be useful for the competitor sections, but its ranking carries no information.
▸Is it illegal for a company to rank its own product first?
No, not on its own. 16 CFR 465.6 targets the misrepresentation, meaning a business presenting a site it controls as an independent review source. A clearly disclosed self-review is legal, and MemoTune's list does exactly that by stating outright that MemoTune is its own product. What crosses the line is the same ranking styled as neutral third-party research with no ownership disclosure anywhere on the page.
▸Does an affiliate disclosure mean the review is honest?
It means the page met a legal minimum, nothing more. Disclosure tells you a financial relationship exists; it says nothing about whether that relationship shaped the order. GetVoIP is unusually direct about this, stating that compensation impacts the order in which products appear within listing categories. Treat a disclosed page as a downweight rather than a discard.
▸What should I read instead of ranking listicles?
Four sources that are hard to fake: the tool's public changelog, its refund and cancellation policy, a search for the word 'cancel' in its subreddit or Discord, and a competitor's comparison page about it. Vendor listicles are still worth opening for the competitor limitations sections, which tend to be the most carefully researched part of the page. Skip the ranking.
The 5 best AI video finds, every week
New models, tested prompts, and what actually worked in our production — one short email a week. No spam, unsubscribe anytime.
Written by Sam Whitaker
Developer & API Cost Writer
Indie developer who reads the API docs before opening the UI and scripts every test he runs more than twice. Tracks cost-per-call and rate limits the way accountants track invoices.
Explore these topics
Every guide, comparison and prompt library we have on each.
