AI washing in SaaS: which claim patterns should buyers check?
Last reviewed May 30, 2026
Software-as-a-service products increasingly describe features as 'AI-powered', 'AI native', or 'built on AI' without explaining what the AI component does, what it replaces, or how it performs. This page helps you identify those patterns — vague feature labels, inflated automation claims, and compliance references not backed by an independent audit — and prepare the specific questions to ask before relying on any AI feature claim.
Evidence buyers verify
- A description of what the AI component does, distinct from rules-based automation, for any 'AI-powered' or 'AI native' claim.
- A list of which tasks require human review or correction and which are fully automated, for any automation or replacement claim.
- Third-party audit documentation with scope boundaries for any compliance, certification, or regulatory-ready claim.
Opens the checker for this claim type. Paste your vendor's exact wording there. Evidence questions only — not a blacklist or fraud detector. Not sure what a result looks like? See a sample receipt.
Sources this guide draws from
- ASA/CAP: AI as a Marketing Term Report ASA/CAP guidance· 2024
UK advertising standards analysis of how software products use 'AI' as a marketing term. Identifies vague AI-powered claims, evidence expectations, and patterns in SaaS product marketing.
- FTC v. accessiBe Inc. — AI Accessibility Tool Compliance Claims FTC enforcement· March 2024
FTC action against an AI SaaS tool that claimed to make websites automatically ADA-compliant. Source for AI compliance claims in SaaS products that overstate what AI-based automation can guarantee.
- FTC Crackdown on Deceptive AI Claims and Schemes FTC enforcement· September 2024
FTC summary of enforcement actions involving AI SaaS products with deceptive automation, accuracy, and capability claims.
Public claims with documented evidence gaps
"AI-powered platform — no manual work required"
Vague AI-powered- Source and date
- ASA/CAP AI as a Marketing Term Report · 2024
- Evidence signal
- Vague AI-powered claim combined with 'no manual work' automation language, without a description of what the AI does or what tasks still require human input.
- Evidence gap
- A buyer needs to know which tasks the platform completes without human input, which still require manual review or configuration, and what the AI component does as distinct from rules-based automation.
- Buyer question
- Which tasks does the AI complete fully automatically, and which require manual review, configuration, or correction by a user?
"automatically makes your website fully ADA and WCAG 2.1 compliant with a single line of code"
Compliance / Safety- Source and date
- FTC v. accessiBe Inc. · March 2024
- Evidence signal
- Compliance certification claim for a SaaS AI overlay, without audit scope, named certification body, or disclosure that compliance depends on additional developer work.
- Evidence gap
- A buyer needs to know which WCAG criteria the overlay addresses, which require developer remediation, whether the resulting site would pass an independent audit, and what liability the vendor accepts if a user files an ADA complaint.
- Buyer question
- Which WCAG criteria does the overlay meet, which require our team to make code changes, and can we get a written statement of what is and is not covered?
"AI writes, schedules, and publishes all your content automatically"
Automation / Replacement- Source and date
- ASA/CAP AI as a Marketing Term Report · 2024
- Evidence signal
- Full-automation claim for a content workflow without scope or accuracy limitations, implying the AI replaces human editorial judgment entirely.
- Evidence gap
- A buyer needs to know the error rate of AI-generated content, which steps require human review or approval, and what happens when the AI produces inaccurate or off-brand output.
- Buyer question
- What percentage of content published by this platform requires no human editing, and what types of errors has the AI produced in production for existing customers?
Match each claim pattern to the evidence buyers need
| Claim pattern | Evidence needed | Buyer question |
|---|---|---|
| AI-powered, AI native, or built on AI (without further description) | Description of the specific model or technique, what the AI component does in the product workflow, how it differs from rules-based logic, and how often it is updated or retrained. | What specific AI model or technique powers this feature, and what would the feature do differently if the AI component were replaced with rules-based logic? |
| Fully automated, zero manual effort, or no human required | List of tasks the AI handles without human input, list of tasks that still require human review, error rate in production, and escalation path for AI errors. | Which tasks still require a human to review or correct the AI output before it is used? |
| GDPR-ready, HIPAA-compliant, SOC 2 certified, or ADA-compliant (as a product feature) | Third-party audit report, scope of the audit, which systems and configurations are covered, and which compliance obligations remain with the buyer rather than the vendor. | Is this compliance certification scoped to the vendor's own infrastructure, and what does our team still need to implement independently? |
| AI learns from your data and improves over time | Description of how the model is updated, whether customer data is used to train a shared or customer-specific model, how improvement is defined and measured, and data retention terms. | Is our data used to train a model that also improves other customers' instances, and how is improvement defined and measured? |
| Integrates with everything or works with all your existing tools | List of supported integrations with version numbers, maintenance SLA, what happens when an integrated tool updates its API, and any limitations on data volume or format. | What is the integration maintenance process, and what is our recourse if an integration breaks after a third-party API update? |
| X% faster, reduces time by Y, or saves Z hours per week | Deployment conditions of the reference customer, baseline process, measurement period, whether the time saving includes setup and error-correction time, and whether conditions match the buyer's workflow. | Does the time-saving figure include setup, configuration, and correcting AI errors, or only the time the AI spends on the task itself? |
Evidence to request
- A description of what the AI component does, distinct from rules-based automation, for any 'AI-powered' or 'AI native' claim.
- A list of which tasks require human review or correction and which are fully automated, for any automation or replacement claim.
- Third-party audit documentation with scope boundaries for any compliance, certification, or regulatory-ready claim.
- The model update process and data usage terms for any 'learns from your data' or 'improves over time' claim.
- Deployment conditions and measurement method from a reference customer whose environment matches yours for any time-saving or ROI claim.
Questions to put in front of the vendor
- For this 'AI-powered' claim, what specific model or technique is used, and how is it different from rules-based automation?
- For this automation claim, which tasks still require human review, and what is the process when the AI output is wrong?
- For this compliance claim, who conducted the audit, what was the scope, and what compliance obligations remain with our team?
- For this 'learns from your data' claim, is our data used to train a shared model, and how is model improvement measured?
- For this ROI or time-saving claim, what were the baseline conditions of the reference customer?
- If we removed the AI component from this product, which features would stop working or work differently?
Wording boundaries to compare against
- Uses [specific model or technique] to automate [named task]; human review required for [named decisions] and when output confidence is below [stated threshold].
- Addresses [named WCAG criteria] using an overlay approach; additional developer changes required for criteria listed in the implementation guide.
- Reduced [named metric] by X% for a [company type] customer using [named baseline workflow]; individual results depend on workflow and configuration.
- Model is updated [frequency] using [named data source]; customer data is [used / not used] to update the shared model; data retention terms apply.