gotalt?

AI claims and reviews checker

Drag one link to your bookmarks bar. Click it on any live page, including a staging site, localhost or a page behind a login, and it lists the wording that usually needs evidence behind it: AI claims, claims of meeting a law or standard, guarantees, and the review and testimonial patterns the FTC's rules speak to. Each item comes with the rule's own words, a link to the source, and a note on what evidence is worth keeping.

Published . Last reviewed .

Free · No signup · Runs entirely in your browser. The page you check is never sent anywhere

Get the bookmarklet

What it checks

It reads the page as your browser has rendered it, so it sees content added by JavaScript and the page as you left it, logged in. It reads only: it never clicks, opens, scrolls or types on the page. It never says that a claim is untrue or that a review does not come from a real customer. It lists wording, quotes the rule, and says what evidence is worth keeping.

  • AI capability claims. Phrases such as "AI-powered", "100% automated", "AI lawyer", accuracy figures and claims to detect AI-generated content. Each is a claim about what a product can do, and the FTC has said there is no AI exemption from the laws on the books.
  • Claims of meeting a law or standard: claims such as "WCAG compliant", "ADA compliant", "100% accessible" and "lawsuit-proof", the kind of claim the FTC's accessiBe order is about, and claims such as "EU AI Act compliant".
  • Guarantees and income promises. Phrases such as "guaranteed income", "we guarantee results" and "passive income". The FTC's descriptions of its AI business-opportunity cases mention income promised through AI-powered tools.
  • Review requests and incentives. Wording that offers something for a review, with a note when a rating is named in the same text; requests that name a rating; and wording that routes some customers away from public review pages.
  • Claims about the reviews shown. "All reviews are shown", "verified buyer", "real reviews from real customers" and "independent reviews", each next to the rule text that speaks to it, and a note when the page also loads a review tool.
  • Testimonials. Testimonial-like blocks, found by markup and by class and id names, and how each credits its author: a full name, a first name or initials, a role only, or nothing. It also lists credits that name someone inside the business, photos whose file name or alt text suggests a stock or placeholder image, and wording about AI-written reviews, AI avatars and purchased followers.
  • Review tools. Signatures for 12 review tools, matched against script and frame addresses and element ids and classes. A tool being there says nothing about who controls it or how it is set up.
  • What it could not check. A cookie or consent banner that appears to be showing, frames from other sites and content that loads later are listed on every run, so a quiet result is never mistaken for a clean one.

Every finding has the wording found, the text around it, the rule text with a link to its source, a note on what evidence is worth keeping, and a Show on page button that scrolls to the element and draws a box around it. It reads at most 20,000 elements per page and at most 3,000 characters of text per block, finds at most 300 testimonial-like blocks, and shows at most 200 findings per category.

Wording the bookmarklet looks for, with examples it finds
Kind of wordingExamples it finds
AI-powered or AI-driven wordingAI-powered, powered by AI, our proprietary AI engine
First, only, best, or better than peoplethe world's first AI, better than humans
Fully automated or no human involved100% automated, fully automated, no human involvement
Accuracy or success figure99% accurate, accuracy of 98%
AI content detection claimAI content detector, detects AI-generated text
AI presented as a lawyer, doctor or other professionalAI lawyer, replaces your attorney
Accessibility standard or law named with "compliant" or similarWCAG compliant, ADA compliant, WCAG 2.2 AA compliant, one line of code, compliant in 48 hours
"Fully compliant" and similarfully compliant, 100% accessible, stay compliant
Protection from lawsuitslawsuit-proof, protects you from ADA lawsuits
EU AI Act "compliant" or similarAI Act compliant, Article 50 ready
Income or profit promisepassive income, earn $5,000 per month
Guarantee of a result100% guaranteed, we guarantee results
Something offered for a review, in wording that names a ratingLeave a 5-star review and get 10% off
Something offered for a review10% off when you leave a review, a gift card in exchange for a review
Review request that names a ratingLeave us a 5-star review, only positive reviews
Wording that routes some customers away from public reviewscontact us before leaving a review, send only happy customers to Google
Claim that all reviews are shown or none are removedall reviews are shown, unfiltered reviews, we never remove reviews
Claim that reviews are verified, real or from real customersverified buyer, 100% real reviews, real reviews from real customers
Claim of independent or unbiased reviewsindependent reviews, unbiased comparison
AI-written, generated or purchased reviews and testimonialsAI-generated reviews, buy reviews
AI avatar, presenter or actor wordingAI avatar, digital human, AI UGC
Followers, likes or views offered for purchasebuy followers, buy Instagram likes
Testimonials with no name showna testimonial credited to Anonymous, A happy customer or nothing
Testimonial credited to someone inside the businessa credit such as Jane, our founder
Testimonial photo: file name or alt text suggests a stock, placeholder or generated imageshutterstock_1234.jpg, customer-3.jpg, ai-generated-face.png

The phrases are a fixed list of English wording. A phrase is listed wherever it appears, so a how-to article that quotes "WCAG compliant" is listed too; the panel says that explaining a rule or quoting someone else's words is not a claim about your product.

Why these checks: what the sources say

Each line below was checked against the primary text on October 22, 2026. This tool lists wording. It does not say whether a rule applies to you, whether a claim is accurate or whether a review is from a real customer, and nothing on this page is legal advice.

Sources behind the claim checks, verified October 22, 2026
SourceWhat it saysWhat the check does with it
FTC press release, September 25, 2024 (Operation AI Comply). FTC Chair Lina M. Khan: "The FTC's enforcement actions make clear that there is no AI exemption from the laws on the books." Five matters were announced, including DoNotPay, which the FTC said marketed an AI service as a "robot lawyer", and three business-opportunity matters that promised income through AI-powered tools. Lists AI capability wording, "AI lawyer" wording, income promises and guarantees, with a note to keep the evidence behind each.
FTC press release, April 22, 2025 (accessiBe final order). The order "bars accessiBe from representing that its automated products can make any website WCAG-compliant", and bars claims that the products "can ensure continued compliance with WCAG over time, unless it has the evidence to support such claims". It also bars misrepresenting that "statements made in reviews, blog posts, or articles about its automated products are independent opinions". Lists claims of meeting an accessibility standard, claims of protection from lawsuits, and claims of independent reviews.
FTC press release, August 28, 2025 (Workado final order, vote 3-0). The order concerns accuracy claims for an AI content detection product and requires "competent and reliable evidence showing those products are as accurate as claimed". Lists accuracy figures and claims to detect AI-generated content. This tool makes no such claim: it never decides whether anything is AI-generated.
16 CFR 465.2(a), Consumer Reviews and Testimonials Rule. It is a violation of the rule for a business to write, create, or sell a consumer review, consumer testimonial, or celebrity testimonial that "materially misrepresents, expressly or by implication: (1) That the reviewer or testimonialist exists; (2) That the reviewer or testimonialist used or otherwise had experience with the product, service, or business that is the subject of the review or testimonial". Lists wording about verified or real reviews, AI-written or purchased reviews, and testimonials with no name shown. The rule does not ask for a name; a name or a dated record makes a testimonial easier to check later.
16 CFR 465.4 and the FTC's questions and answers. The rule bars compensation or other incentives "conditioned expressly or by implication on, the writing or creation of consumer reviews expressing a particular sentiment, whether positive or negative". The FTC's answer: "The rule does not prohibit giving incentives for reviews, as long as there isn't an express or implied requirement that the reviews have to express a particular sentiment." Lists wording that offers something for a review, and says when a rating or tone is named in the same text.
16 CFR 465.5(a). It is a violation for an officer or manager of a business to write or create a consumer review or testimonial about the business "that fails to have a clear and conspicuous disclosure of the officer's or manager's material relationship to the business". Lists testimonial credits that name a role inside the business, such as a founder or owner.
16 CFR 465.6. It is a violation for a business to materially misrepresent that "a website, organization, or entity that it controls, owns, or operates provides independent reviews or opinions". Lists wording that claims independent or unbiased reviews, and names any review tool loaded on the same page.
16 CFR 465.7(b) and the FTC's questions and answers. The rule concerns a misrepresentation that displayed reviews "represent most or all the reviews submitted to the website or platform when reviews are being suppressed (i.e., not displayable) based upon their ratings or their negative sentiment". On asking only happy customers for reviews, the FTC says: "The rule does not contain a specific prohibition against such conduct. But this practice could violate the FTC Act." Lists "all reviews are shown" style claims and wording that routes some customers away from public review pages.
16 CFR 465.8. The rule covers "fake indicators of social media influence that they knew or should have known to be fake". Lists wording that offers followers, likes or views for purchase.
FTC press release, August 14, 2024 and the FTC's questions and answers. The release says the rule will "deter AI-generated fake reviews". The rule text itself has no AI wording. The FTC's answer on avatars: "The rule has no blanket prohibition on the use of AI-generated avatars in marketing." Lists wording about AI-written reviews and AI avatars, and checks whether the page also says the people shown are not actual customers.
16 CFR 255.2(c) and 255.1(h), Endorsement Guides. Advertisements presenting endorsements by what are represented to be actual consumers "should utilize actual consumers in both the audio and video, or clearly and conspicuously disclose that the persons in such advertisements are not actual consumers of the advertised product". An example in 255.1(h) describes a website that "accompanies the testimonials with stock photos" instead of "images of the actual endorsers". Lists AI avatar and actor wording, and photos whose file name or alt text suggests a stock or placeholder image. It never looks at the picture.
FTC press release, December 22, 2025 (warning letters). The FTC sent warning letters to 10 companies about possible violations of its Consumer Review Rule, and the release cites "civil penalties of up to $53,088 per violation". The letters are not formal findings. Nothing is scored. The figure is here so the size of the exposure is not a surprise.
Commission EU icons page (last update October 13, 2026). "The use of these icons does not establish legal compliance by itself." Lists wording that says a site or product meets the EU AI Act, with a note to keep a dated record of which duties apply to you.

Offering something for a review is not banned by itself

The rule bars an incentive that depends on what the review says. The FTC's answer, quoted above, is that giving incentives for reviews is not prohibited as long as nothing requires a particular sentiment, and its page adds that failing to disclose an incentive could also violate the FTC Act. So wording that offers a discount for a review is listed with a note to keep the request as sent and to check that the offer does not depend on the rating. When the same text names a five-star or positive review, the panel says so, because that is the pattern 465.4 is written about.

Asking only happy customers

Review gating is a plain-language name for steering unhappy customers to a private channel and happy ones to a public review page. The FTC's answer quoted above finds no specific prohibition in the rule and says the practice could still violate the FTC Act. The panel lists wording that does this and leaves the question to you and your adviser.

AI-generated reviews and AI avatars

Part 465 has no AI wording. It speaks of reviews and testimonials that misrepresent that the reviewer exists or had experience with what is reviewed, and the FTC has said it will deter AI-generated reviews of that kind. An AI avatar is not banned by the rule, according to the FTC's answer; the Endorsement Guides speak to whether the people shown in an advertisement presented as customers are actual consumers. The panel lists the wording and checks whether the page also says the people shown are not actual customers.

AI claims need evidence like any other

The FTC's AI cases are mostly about claims, not labels: what a product can do, how accurate it is, whether it makes a site meet a standard. The panel lists the phrases and asks you to keep the evidence behind each, as of the day the claim was made.

Install it in your browser

  1. Show your bookmarks bar

  2. Drag the link onto the bar

    Drag the "GotAlt claims and reviews check" link from the box above onto the bar. In Firefox you can also right-click it and choose Bookmark Link.

  3. Or make the bookmark manually

    Dragging needs a mouse or trackpad. Without one, press "Copy bookmarklet code", add a bookmark for any page (Ctrl+D, or Cmd+D on a Mac), edit that bookmark (in Safari, Control-click it to edit its address; see Apple's instructions), call it "GotAlt claims and reviews check", and paste the code into its address or URL field. Paste it there, not into the browser's address bar: some browsers remove the javascript: start of the code when you paste it into the address bar.

  4. Click it on any page

    Open the page you want to check and click the bookmark. A panel opens at the side of the page. Click the bookmark again, press Escape, or use the Close button to remove it. "Run the checks again" re-reads the page, which is what you want after a cookie banner is dealt with, after you scroll, or after a review widget has loaded.

Or press "Try it on this page" above to run it on this page. It will list the example phrases quoted here, because it reads wording: that is what it is for.

It sits beside our AI label and chatbot disclosure checker and our accessibility bookmarklet. They are separate bookmarklets, and you can install all three.

Nothing leaves your browser unless you click

The bookmarklet reads the page in your tab and shows its results in a panel in the same tab. It makes no network requests, loads no other scripts and writes nothing to storage, and it has no analytics, so we do not know that you ran it. The links in the panel (the sources on ftc.gov, ecfr.gov and the European Commission's site, this page and the Site Image Check) only load if you click them. They open in a new tab and send no referrer.

The Site Image Check link is the one that carries something: the address of the site you checked, such as https://example.com, never the page you were on or anything on it, so that the order page can fill it in, and a short tag saying the click came from this bookmarklet. If you do not click it, we never receive either.

A bookmarklet runs with the same access as the page you click it on, so only use ones you have read. This one is a single readable file: gotalt-claims-check.js. The bookmarklet is the same code with its comments, spare spaces and line breaks taken out, so it fits in a bookmark. Its rules are written once, in claims-rules.js, and copied into the file, so it never downloads a second script and a site that blocks scripts from other places cannot stop it loading part of itself. Whether a site's security settings let a bookmarklet run at all depends on your browser; see the limits below.

What it can and can't do

What this bookmarklet can't do

  • It does not judge claims, reviews or testimonials. It reads wording. It does not say whether a claim is accurate, whether a review or testimonial comes from a real customer, or whether a person in a photo is who the page says. A name that looks like a full name does not show that someone is real, and a missing name does not show that they are not. Names are recognized by capital letters, so one typed in lower case is listed as no name.
  • It does not say whether a rule applies or is met. Whether Part 465, the Endorsement Guides, the FTC Act or a state law applies to you, and whether your wording or records are enough, is for you and your adviser. It scores nothing and a quiet result means only that these checks found nothing.
  • It never decides whether anything is AI-generated. A photo file name or alt text that looks like a stock or generated image is a hint about a name. The check does not look at the picture, and it cannot see or hear a video.
  • It reads a fixed list of English phrases. Other wording, other languages and claims made only in images or video are not read, and a phrase is listed wherever it appears, including in an article that explains it.
  • Testimonials are found by markup and names. Blocks are found by blockquote, review markup, and class and id names that contain testimonial, review or quote. A slider or widget built another way can be missed, and one that loads later is seen only if you run the check again. Frames from other sites and closed shadow DOM cannot be inspected; open shadow DOM and frames from the same site are.
  • Review tools are found by their scripts and names. A tool being on the page says nothing about who controls or moderates it.
  • It sees one page, once. It reads at most 20,000 elements, 3,000 characters per block and 300 testimonial-like blocks, and shows at most 200 findings per category. If there are more, it says how many it left out.
  • Some sites can block it. A site's security settings (a Content-Security-Policy) can stop bookmarklets running, and browsers differ here: Firefox applies the site's policy to bookmarklets, which is a long-standing open Firefox bug report. A site with a modal dialog open can also cover the panel. Our automated tests run it in a Chromium-based browser.

Questions

Does this tell me whether my claims or reviews break an FTC rule?

No. It lists wording and quotes the rule text. Whether Part 465, the Endorsement Guides or the FTC Act applies to you, and whether your wording or records are enough, is a question for you and your adviser. This is not legal advice.

Can it tell whether a review or testimonial comes from a real customer?

No, and it does not try. It reads wording and markup: whether a testimonial shows a full name, a first name or initials, a role only, or nothing. A name that looks complete does not show that someone is real, and a missing name does not show that they are not. The rule does not ask for a name; it is about what a testimonial misrepresents.

Why does it list "AI-powered" if no rule bans it?

Because the wording is a claim about what a product does, and the FTC has said there is no AI exemption from the laws on the books. The panel asks you to keep a dated description of what the AI does and the test results behind the wording, so that you can support it if asked.

Is it against the rules to offer a discount for a review?

Not by itself. The FTC's questions and answers say the rule does not prohibit giving incentives for reviews, as long as there is no express or implied requirement that the reviews express a particular sentiment, and that failing to disclose an incentive could also violate the FTC Act. The panel lists the wording, says when a rating is named in the same text, and asks you to keep each request as sent.

What is review gating, and does the rule ban it?

It means asking unhappy customers to contact you privately while happy ones are sent to a public review page. The FTC's answer on asking only happy customers for reviews is that the rule has no specific prohibition against it, but that the practice could violate the FTC Act. The panel lists wording that does this and leaves the question to you and your adviser.

Does Part 465 ban AI-generated reviews or AI avatars?

The rule text has no AI wording. It covers reviews and testimonials that misrepresent that the reviewer exists or had experience with what is reviewed, and the FTC's August 14, 2024 release says it will deter AI-generated reviews of that kind. On avatars, the FTC says the rule has no blanket prohibition on AI-generated avatars in marketing, and the Endorsement Guides speak to whether the people shown in an advertisement presented as customers are actual consumers.

Why a bookmarklet and not a browser extension?

There is nothing to install from a store, and it only runs on the page when you click it. Extensions often ask for permission to read every page you visit. A bookmarklet only has the page you clicked it on. The trade-off is that a site's security settings can block a bookmarklet in some browsers, Firefox in particular; the limits above say more.

Does it work on localhost, staging and pages behind a login?

Yes, because it runs in the tab you already have open. Nothing is sent to a server, so GotAlt never needs to reach your page. That is the difference from our free scan, which fetches pages from the outside and so cannot see anything private. The one exception is the link near the end of the results to order a Site Image Check: if you click it, gotalt.com opens with the site's address in the link and nothing else from your page. That check reads public pages from the outside, like the free scan.

This checks one page, when you remember to click it

The Site Image Check reads up to 50 pages of your site for $49 (250 for $149), once, with no subscription. Its AI disclosure section lists, for the images the check opens, the AI markers each file declares (a C2PA manifest, an IPTC DigitalSourceType, XMP keywords such as contains-synthetic-performer) and whether the page shows AI-label text with the image (alt text, a figcaption or adjacent text), and, for each page, known chat widgets and whether AI-disclosure wording appears in the page's HTML. It never decides whether an image is AI-generated. Its Phrases to review section runs this checker's wording patterns on the pages it reads, up to 25 distinct phrases, each linked to the rule text. Testimonial credits and photos are read only by this bookmarklet, on the pages you click it on.

Check my site once from $49

A link from your site, newsletter or resources page helps other people find this free tool. Copy it here: