The 5 levels of AI involvement<\/h2>\n
The 5 levels, ordered from most-allowed to least-allowed. The ordering is based on consent, disclosure, and authenticity. Each level is a real industry pattern. Each level has different rules.<\/p>\n Level 1: AI-assisted.<\/strong> AI helps polish a customer’s words. The customer writes the testimonial. AI fixes grammar, removes redundancy, improves flow. The customer’s voice is preserved. The output is recognizably the customer.<\/p>\n Level 2: AI-translated.<\/strong> AI translates the customer’s words from one language to another. The customer writes in their native language. AI translates to English (or the target language). The customer’s voice is preserved. The output is recognizably the customer, in a different language.<\/p>\n Level 3: AI-rewritten.<\/strong> AI rewrites the customer’s words substantially. The customer writes a rough version. AI produces a polished version that goes beyond grammar. The customer’s voice may be partially preserved, partially replaced. The output may or may not be recognizably the customer.<\/p>\n Level 4: AI-generated.<\/strong> AI writes the testimonial from scratch. There may or may not be a real customer behind the testimonial. The output is AI-generated text. If there is a real customer, they may not have written any of the words.<\/p>\n Level 5: AI-fabricated.<\/strong> AI invents a fake customer and writes the testimonial in the fake customer’s voice. There is no real customer. The output is a fabricated testimonial attributed to a non-existent person.<\/p>\n AI-assisted testimonials are allowed and standard. The customer writes the testimonial. AI cleans up the wording. The customer’s voice is preserved. The output is recognizably the customer.<\/p>\n The consent question: the customer is aware that AI is involved. The disclosure is implicit (the customer wrote the original). The output is the customer’s words, polished. The ethical line is not crossed.<\/p>\n The trust value is preserved. The reader sees a polished testimonial. The reader does not know (and does not need to know) that AI was used for the polish. The customer knew. The customer approved. The output is the customer’s.<\/p>\n AI-translated testimonials are allowed. The customer writes in their native language. AI translates to the target language. The customer’s voice is preserved across languages. The output is recognizably the customer, in a different language.<\/p>\n The consent question: the customer is aware that AI is involved. The disclosure is implicit. The output is the customer’s words, in a different language. The ethical line is not crossed.<\/p>\n The trust value is preserved or even enhanced. The reader sees a testimonial in their language. The reader trusts it because the original was a real customer’s words. The translation is the AI’s contribution, not the substance.<\/p>\n AI-rewritten testimonials are grey. The customer writes a rough version. AI produces a polished version that goes beyond grammar. The customer’s voice is partially preserved, partially replaced. The output may or may not be recognizably the customer.<\/p>\n The consent question is the test. If the customer approves the rewrite, the testimonial is closer to Level 1 (allowed). If the customer does not approve the rewrite, the testimonial is closer to Level 4 (contested). The disclosure matters: was the customer told the rewrite is substantial?<\/p>\n The trust value is mixed. The reader who compares the rough version to the polished version can see the difference. The reader who only sees the polished version may trust it, but the trust is conditional on the customer’s approval. The ethical line is at the customer’s approval, not at the AI’s involvement.<\/p>\n AI-generated testimonials are contested. AI writes the testimonial from scratch. There may or may not be a real customer behind the testimonial. The output is AI-generated text.<\/p>\n The consent question is the key. If a real customer is behind the testimonial and approves the output, the testimonial is closer to Level 1. If a real customer is behind the testimonial but does not approve the output, the testimonial is closer to Level 5. If there is no real customer, the testimonial is Level 5 regardless of disclosure.<\/p>\n The trust value is conditional. The reader trusts the testimonial if the customer is real and the output is approved. The reader does not trust the testimonial if the customer is fictional or the output is unapproved. The ethical line is at the customer-and-output reality, not at the AI involvement.<\/p>\n AI-fabricated testimonials are universally rejected. AI invents a fake customer and writes the testimonial in the fake customer’s voice. There is no real customer. The output is a fabricated testimonial attributed to a non-existent person.<\/p>\n The trust value is zero. The reader who discovers the fabrication loses all trust in the brand. The legal exposure is high. The FTC has explicit guidance against fabricated endorsements. Google can de-rank pages with fabricated reviews. The platforms have explicit policies against synthetic testimonials.<\/p>\n The ethical line is unambiguous. Do not fabricate testimonials. Do not use a real name with a fabricated quote. Do not use a real customer’s name with an AI-generated quote the customer did not approve. The boundary is clear. The boundary is enforced.<\/p>\n Google’s policy: Google does not penalize AI-assisted content if the content is factually accurate and the customer is real. Google does penalize fabricated reviews and AI-generated content that misleads users. The schema markup must reflect the actual experience, not an AI-invented one.<\/p>\n The FTC’s guidance: the FTC’s Endorsement Guides require that endorsements reflect the endorser’s honest opinion, and that any material connection between the endorser and the advertiser be disclosed. AI-fabricated testimonials violate both. The FTC has actively pursued cases against companies using fake reviews.<\/p>\n Industry standards: most testimonial platforms (G2, Capterra, Trustpilot) require verified customers. AI-fabricated testimonials are grounds for removal. The platforms are investing in detection. The detection is improving.<\/p>\n The first trust risk is the discovery. The reader who discovers AI fabrication loses all trust in the brand. The loss is asymmetric: the brand spends years building trust and loses it in a single discovery. The risk is not worth the reward.<\/p>\n The second trust risk is the platform penalty. The platforms (Google, FTC, G2) are actively enforcing their policies. The penalty ranges from de-ranking (Google) to legal action (FTC) to removal from the platform (G2). The penalty is real.<\/p>\n The third trust risk is the dilution. The testimonials that ARE real get diluted by the ones that are not. The reader who encounters both cannot tell which is which. The trust in all of them drops. The dilution is permanent.<\/p>\n The first guideline is consent. The customer must know AI is involved. The customer must approve the output. The approval is documented. The documentation is the audit trail.<\/p>\n The second guideline is disclosure. The testimonial is published with the customer’s name and the customer’s role. The testimonial is not presented as if a real customer wrote it when AI did. The transparency is the trust.<\/p>\n The third guideline is authenticity. The testimonial reflects the customer’s actual experience. The numbers are real. The outcomes are real. The customer could verify the testimonial if asked. The authenticity is the value.<\/p>\n Testivo uses AI for the allowed levels. The AI helps polish the customer’s words (Level 1). The AI helps translate the customer’s words (Level 2). The AI helps identify the strongest quotes from a longer submission. The AI helps with curation, tagging, and management.<\/p>\n Testivo does not generate testimonials from scratch (Level 4). Testivo does not fabricate testimonials with fake customers (Level 5). Testivo does not produce AI content attributed to a real customer without that customer’s approval. The boundary is documented. The boundary is enforced in the product.<\/p>\n The position is not marketing. The position is product. The product does not have a feature for “generate a testimonial from scratch.” The product does not have a feature for “fabricate a customer.” The product’s AI features are limited to the allowed levels. The product is built to be trust-first.<\/p>\n If you have the AI-involvement decision framework and want the AI citability side, the testimonials and AI search<\/a> page is the other AI topic. For the format the AI would be polishing, the testimonial examples<\/a> page is the gallery of good testimonials that AI should help produce, not replace. For the broader definition of customer testimonials, the customer testimonials pillar<\/a> is the parent reference.<\/p>\n For the structured data that AI-assisted testimonials need to be machine-readable, the llms.txt<\/a> page is the reference. For the Testivo positioning more broadly, the testimonial software<\/a> page is the category overview.<\/p>\n![\"Five](\"https:\/\/testivo.tech\/blog\/wp-content\/uploads\/ai-levels-scale.jpg\")
Level 1: AI-assisted (allowed)<\/h2>\n
Level 2: AI-translated (allowed)<\/h2>\n
Level 3: AI-rewritten (grey)<\/h2>\n
Level 4: AI-generated (contested)<\/h2>\n
Level 5: AI-fabricated (rejected)<\/h2>\n
The 3 platform policies<\/h2>\n
The 3 trust risks<\/h2>\n
The 3 ethical guidelines<\/h2>\n
What Testivo does and does not do<\/h2>\n
The journey after this page<\/h2>\n