FTC AI claim sources
Use this page when a public AI claim promises accuracy, automated review, review authenticity, earnings, or business outcomes. The cases below show what evidence questions FTC materials raise before a buyer relies on similar wording.
This is a source-organized view of reviewed cases and claim patterns. It is not a current determination about a vendor, legal advice, a compliance certification, a company ranking, or a vendor approval list.
Source scope
What this source covers
Marketing and advertising claims about AI product performance, automation, reviews, testimonials, and outcomes.
Context and limits
FTC sources are most useful when a public AI marketing claim reaches consumers, small businesses, schools, buyers, or other customers. The FTC's AI enforcement focus has covered recurring wording categories including: numeric accuracy and detection claims that lack benchmark scope; automation or replacement claims that omit the remaining human review role; compliance or harm-filtering claims that overstate automated coverage; review and testimonial claims where AI generation or incentives were not disclosed; and earnings or cost-saving claims unsupported by actual customer-outcome distribution.
The table below also includes additional claim patterns that share the same evidence burden.
For each pattern, the key evidence question is whether records existed at the time the public wording was used—not whether a product works in general.
FTC materials do not address internal product capability, future versions, or claims that were never publicly visible to customers.
How buyers should use it
Use this page to match the wording you are reviewing to a documented pattern, then use the buyer question to decide what evidence to request before relying on the claim.
Do not use it to rate a vendor, decide a claim outcome, or treat this page as a complete view of every FTC AI matter.
Source timeline
Use these dated source points to understand which AI claim patterns this page can help you compare before you open a case detail.
-
December 19, 2023
Rite Aid facial recognition source describes an AI deployment without enough safeguards around false positives, image quality, employee use, and consumer impact.
AI safety claims need operational controls, not just model capability: testing, monitoring, staff limits, notice, and stop rules all matter.
-
September 25, 2024
Operation AI Comply groups AI lawyer, review-generation, and AI-linked business opportunity claims into one enforcement sweep.
Replacement, social-proof, and earnings claims each need claim-specific records that existed when the public wording was used.
-
November 26, 2024
Evolv source focuses on AI-powered security screening claims about detection, harmless-item filtering, speed, false alarms, and labor costs.
Safety-sensitive performance claims need item-level test scope, false positive and false negative rates, field conditions, and staffing assumptions.
-
January 13, 2025
IntelliVision source addresses facial-recognition claims about accuracy, bias, training data, and anti-spoofing capability.
Bias-free and anti-spoofing wording needs subgroup performance data, dataset coverage, tested attack types, and deployment-condition limits.
-
April 22, 2025
accessiBe source addresses claims that an AI-powered overlay could make websites WCAG compliant and keep them compliant over time.
Automated compliance wording needs criteria-by-criteria scope, excluded issues, manual review boundaries, and ongoing maintenance responsibilities.
-
August 28, 2025
Workado / Content at Scale source addresses a headline AI detector accuracy claim and the records needed to support future accuracy wording.
Numeric AI accuracy claims need benchmark design, content categories, model versions, sample size, and error rates that match the stated use case.
Claim patterns and evidence gaps
Match the public AI wording you saw to a pattern below, then use the buyer question to decide what evidence to request or whether to check the exact wording.
| Claim pattern | Evidence gap | Buyer question |
|---|---|---|
| Numeric accuracy, detection, or benchmark claims | Benchmark scope, content or item categories, model versions, sample size, false positive and false negative rates, and update cadence. | What dated test record supports the number, and does the test set match the content, venue, user group, or workflow we would rely on? |
| AI replaces a professional or removes human review | Task scope tested, qualified review involvement, known failure cases, limitation notices, escalation path, and non-use cases. | Which parts of the task were tested against qualified review, and where must a human still approve, correct, or stop the output? |
| Automated compliance, accessibility, safety, or harm-filtering claims | Named standard or harm category, automated coverage, excluded criteria, human review boundary, monitoring cadence, and complaint handling. | Which criteria or harm categories are covered by automation, and which still require manual testing, review, or remediation? |
| AI-generated reviews, testimonials, or social proof | Reviewer independence, AI-generation disclosure, incentive or compensation records, collection timing, and actual user-experience support. | What evidence shows the review or testimonial reflects a real user experience and is not AI-generated or incentivized without disclosure? |
| AI-linked earnings, ROI, growth, or cost-saving claims | Customer outcome distribution, total cost basis, time period, baseline comparison, refund conditions, and unsuccessful-customer context. | What median or typical customer result supports the claim after costs, setup, failed attempts, review time, and refund outcomes are counted? |
| Vague AI-powered data or targeting claims | Actual data source, consent record, model function, data-broker involvement, geographic accuracy, and whether the claim matches the current workflow. | What data does the AI actually use, what consent record supports that use, and what does the model do beyond ordinary targeting or list matching? |
| Bias-free, equitable, or group-neutral AI claims | Subgroup performance data, protected-class test coverage, dataset composition, evaluation methodology, error-rate parity across groups, and deployment-condition limits. | What subgroup performance data supports the bias-free claim, and does that data cover the specific user groups and decision types in the intended use case? |
| AI personalization or recommendation using consumer data | Data source and consent basis, model contribution versus standard list matching, data-broker involvement disclosure, accuracy limits by segment, and opt-out mechanism. | What data does the AI use to personalize or recommend, and are the source, consent basis, and AI contribution clearly different from ordinary database targeting? |
| AI health, wellness, or diagnostic capability claims | Clinical validation scope, tested population, regulatory clearance status, failure conditions, limitations compared to qualified clinical review, and excluded indications. | What clinical or regulatory record supports this capability claim, and does the marketing context match the cleared or validated use case and user type? |
Reviewed cases
Open a case to see the original source, claim type, evidence gap, buyer questions, wording boundaries, and update status.
-
Workado AI detector accuracy
- Source date
- August 28, 2025
- Checked date
- May 22, 2026
Risk pattern: Numeric accuracy without disclosed test scope or error rates
Evidence gap: Benchmark design and scope, content types included in testing, AI model versions covered, sample size, false positive rate, false negative rate, and how frequently the benchmark is updated as AI models change.
See evidence gap and buyer questions → -
accessiBe automated accessibility claims
- Source date
- April 22, 2025
- Checked date
- May 22, 2026
Risk pattern: Automated compliance promise without stated audit method or human review boundary
Evidence gap: Which WCAG version and success criteria the tool covers, the audit method and test scope, which issues remain outside automated detection, what human review and testing remains necessary, and how the tool handles dynamic content, third-party scripts, or ongoing maintenance.
See evidence gap and buyer questions → -
DoNotPay robot lawyer claims
- Source date
- February 11, 2025
- Checked date
- May 22, 2026
Risk pattern: Professional-replacement positioning without task scope or qualified review boundary
Evidence gap: Task scope and document types tested, the professional review standard used for comparison, failure handling and escalation path, disclosed non-use cases, and user notice about when to seek qualified legal advice.
See evidence gap and buyer questions → -
FTC fake reviews and testimonials rule
- Source date
- August 14, 2024
- Checked date
- May 22, 2026
Risk pattern: Testimonials or review counts used as social proof without disclosure of AI generation, compensation, or collection method
Evidence gap: Whether testimonials are from independent users, whether reviewers received any compensation or incentive, whether any testimonial content was generated or enhanced by AI tools, when reviews were collected relative to product use, and how the collection method is disclosed.
See evidence gap and buyer questions → -
Sitejabber review collection timing
- Source date
- November 6, 2024
- Checked date
- May 22, 2026
Risk pattern: Review ratings displayed without disclosure of collection timing or reviewer experience context
Evidence gap: When reviews were collected relative to product delivery or service completion, what specific experience reviewers were evaluating at the time of the survey, how the collection method and timing are disclosed, and whether displayed ratings reflect post-delivery experience.
See evidence gap and buyer questions → -
Rytr AI-generated testimonial service
- Source date
- December 18, 2024
- Checked date
- May 22, 2026
Risk pattern: AI-generated testimonial or review content presented as genuine user experience without disclosure
Evidence gap: Whether any testimonials produced by the tool were presented as genuine consumer feedback without AI-generation disclosure, what disclosure requirements apply to AI-generated review content, and how buyers can verify review authenticity on platforms that permit AI writing assistance.
See evidence gap and buyer questions → -
Rite Aid facial recognition safeguards
- Source date
- December 19, 2023
- Checked date
- May 22, 2026
Risk pattern: AI biometric surveillance deployment without documented accuracy testing, false-positive monitoring, or harm controls
Evidence gap: Pre-deployment accuracy testing, false-positive tracking, demographic impact monitoring, image-quality controls, employee training, vendor oversight, consumer notice, complaint handling, deletion rules for biometric data, and a clear stop rule when the system cannot control consumer harm.
See evidence gap and buyer questions → -
Evolv AI security screening claims
- Source date
- November 26, 2024
- Checked date
- May 22, 2026
Risk pattern: Safety-sensitive detection claim without item-level detection rates, false-alarm rates, or field-condition limits
Evidence gap: Detection rates by item type, false positive and false negative rates, sensitivity settings, field test conditions, school or venue deployment data, staffing assumptions, comparison method against metal detectors, and documented cases where items were missed or harmless items were flagged.
See evidence gap and buyer questions → -
IntelliVision facial recognition claims
- Source date
- January 13, 2025
- Checked date
- May 22, 2026
Risk pattern: Bias-free and anti-spoofing AI claim without subgroup testing, training-data support, or attack-scenario limits
Evidence gap: Subgroup performance by gender, ethnicity, and skin-tone groups; sample size and training-data documentation; benchmark design; error rates by deployment setting; anti-spoofing test methods; attack types tested; and limits for image quality, lighting, angle, and camera conditions.
See evidence gap and buyer questions → -
NGL AI content moderation claims
- Source date
- July 9, 2024
- Checked date
- May 22, 2026
Risk pattern: AI harm-filtering claim without disclosed detection scope, error rates, or child-safety limits
Evidence gap: Harm category definitions, training data and evaluation set, false negative and false positive rates, teen-user test coverage, language and slang coverage, escalation process, human review boundary, recurrence handling for repeat abuse, and update process for new harmful-message patterns.
See evidence gap and buyer questions → -
Cox Active Listening AI ad targeting
- Source date
- May 21, 2026
- Checked date
- May 22, 2026
Risk pattern: AI-powered targeting claim without evidence of data source, consent basis, algorithm function, or geographic accuracy
Evidence gap: Actual data sources used, whether voice data is collected or not, opt-in method and consent record, algorithm function, geographic targeting validation, data-broker involvement, customer disclosures, and whether the claim is updated when the product workflow changes.
See evidence gap and buyer questions → -
Air AI business growth claims
- Source date
- August 25, 2025
- Checked date
- May 22, 2026
Risk pattern: AI-linked earnings claim without customer outcome distribution, cost baseline, or refund-promise substantiation
Evidence gap: Customer outcome distribution, total startup and operating costs, baseline comparison, time period, refund eligibility and payout history, cancellation terms, customer-selection criteria, evidence that the conversational AI feature worked as described, and evidence separating that feature from coaching, sales, ads, labor, inventory, or market conditions.
See evidence gap and buyer questions → -
Automators AI ecommerce earnings
- Source date
- August 22, 2023
- Checked date
- May 22, 2026
Risk pattern: AI-boosted earnings claim without customer outcome distribution, workload disclosure, or cost basis
Evidence gap: Customer sample size, median profit and loss, total investment required, time to break even, passive-investor versus self-managed-store outcomes, customer workload, inventory and ad costs, ecommerce platform fees, refund or chargeback data, and a method for attributing any outcome to AI-powered tools rather than coaching or store operations.
See evidence gap and buyer questions → -
FBA Machine AI storefront income
- Source date
- June 3, 2024
- Checked date
- May 22, 2026
Risk pattern: Income promise tied to AI-powered software without full cost, customer sample, and failure-rate disclosure
Evidence gap: Customer result distribution, total investment range, inventory and platform fees, ad spend, fulfillment and labor costs, refund request outcomes, unsuccessful customer percentage, time to break even, evidence that the AI-powered software was active in customer stores, and a documented method for isolating the effect of repricing or storefront tools.
See evidence gap and buyer questions → -
Ecommerce Empire Builders AI income
- Source date
- September 25, 2024
- Checked date
- May 22, 2026
Risk pattern: AI-powered business opportunity claim without typical-results evidence, cost disclosure, or customer-loss context
Evidence gap: Typical customer result data by offer type, median net income, unsuccessful customer share, total cost for training versus done-for-you storefronts, refund rate, time period, store-operation workload, product and supplier assumptions, and evidence showing that AI-powered workflows materially changed results.
See evidence gap and buyer questions → -
FTC Content at Scale AI detector
- Source date
- August 28, 2025
- Checked date
- May 24, 2026
Risk pattern: Headline accuracy number without disclosed benchmark, test corpus, or error rate records
Evidence gap: Benchmark design and test corpus description, content types and categories included, AI model versions covered at time of test, sample size, false positive rate, false negative rate, test date, and a record the vendor retains that could be produced if the claim is questioned.
See evidence gap and buyer questions →
How to use this source page
- Match the source to the claim context. Marketing and advertising claims about AI product performance, automation, reviews, testimonials, and outcomes.
- Read the closest case. Compare the risk pattern and evidence gap with the claim you are reviewing.
- Check the exact wording. Paste the public URL or claim text into the checker to get evidence requests, buyer questions, and wording boundaries for the specific claim.
Source updates and corrections
If a source status, date, company response, or case context is outdated, send a correction. Submissions are private by default and reviewed before any update appears on this page.
Send a correction →