In a market where "AI-powered" adds an average 25-40% valuation premium, the incentive to exaggerate AI capability is enormous. The SEC has made AI washing an enforcement priority, and for good reason: the gap between AI marketing and AI reality is widening as more capital flows into AI-adjacent investments.
This article focuses specifically on due diligence — the structured process of verifying AI claims before committing capital. Whether you are a private equity firm evaluating an acquisition target, a venture investor conducting technical diligence, or a board member assessing management's AI strategy, these techniques will help you separate substance from spin.
Not all AI misrepresentation is outright fraud. AI washing exists on a spectrum, and understanding where a target falls on this spectrum is the first task of due diligence.
| Level | Description | Example | Risk level |
|---|---|---|---|
| Level 1: Genuine AI | Real ML models, proprietary training data, measurable impact | Custom NLP models trained on domain-specific corpora | Low |
| Level 2: AI-assisted | Uses third-party AI APIs with light customisation | GPT API calls with company-specific prompts | Medium |
| Level 3: AI-enhanced marketing | Rules-based systems rebranded as "AI" | If-then decision trees labelled "machine learning" | High |
| Level 4: No AI | Zero AI capability despite AI-forward marketing | Static algorithms presented as "AI-powered" | Very high |
The most common form of AI washing is not fabrication — it is exaggeration. Level 2 and Level 3 companies constitute the majority of AI washing cases. They have some automation or analytics capability but position it as proprietary AI to capture valuation premiums they have not earned.
Genuine AI capability requires genuine AI talent. During due diligence, request the organisational chart for the technology function and specifically identify:
If the company claims "AI-powered" products but employs fewer than three people with dedicated ML titles, the claim warrants deep scrutiny. AI is not a side project that a generalist software team builds during spare cycles.
Real AI systems leave infrastructure footprints. Ask to see:
A company genuinely training and deploying ML models will have substantial compute costs, version-controlled model artifacts, and systematic experiment tracking. If none of this exists, the "AI" is likely a thin wrapper around API calls or rules-based logic.
During a recent PE due diligence engagement, a target company claimed to use "proprietary AI" for demand forecasting. When the technical team examined the infrastructure, they found a single Python script running a linear regression model updated quarterly by a business analyst. Monthly cloud compute costs were £47. The company's AI claims had contributed to a £15 million valuation premium that was entirely unjustified.
Technical due diligence is necessary but not sufficient. Commercial red flags often reveal AI washing more efficiently than code review.
Genuine AI systems produce specific, measurable results. When a company describes its AI impact using vague language — "enhanced efficiency," "improved insights," "next-generation analytics" — without specific metrics, the underlying capability is likely thin.
Ask for three things:
Compare the prominence of AI in the company's marketing with the proportion of its technology budget spent on AI. If AI features prominently in investor decks, website copy, and press releases but represents less than 10% of the technology budget, the company is marketing AI rather than building it.
Companies with genuine AI capability have plans for what happens when models degrade, when training data becomes stale, or when the competitive landscape shifts. Ask: "What is your model retraining schedule? How do you detect model drift? What happens if your primary data source becomes unavailable?" Companies that cannot answer these questions do not operate AI systems at scale.
Do not confuse API integration with proprietary AI. A company that calls the OpenAI API, applies a prompt template, and returns the result is not an "AI company." It is a company that uses AI as a service. This distinction matters enormously for valuation because API-based approaches have no competitive moat — any competitor can replicate the integration in days.
Ask the target to list every AI/ML system in production, its purpose, the team responsible, and the measurable business impact. This document reveals scope and specificity simultaneously.
Have an independent ML engineer review the model architecture, training pipeline, data infrastructure, and deployment systems. This cannot be delegated to generalist consultants.
Cross-reference AI claims against cloud bills, compute usage, model versioning logs, and data storage growth. Infrastructure evidence is harder to fabricate than presentations.
Review LinkedIn profiles, publication records, and project histories of the AI team. Genuine capability is built by people with demonstrable experience, not by titles on an org chart.
AI washing creates a specific valuation risk: the premium paid for AI capability that does not exist. When the AI claims unravel post-acquisition — through failed integration, customer complaints, or regulatory action — the valuation premium evaporates and the acquirer absorbs the loss.
The practical response is to decompose the target's valuation into AI and non-AI components. What is the business worth without any AI capability? What incremental value does verified, genuine AI add? If the gap between the asking price and the non-AI valuation exceeds the verified AI value, the premium is unjustified.
The Opagio questionnaire provides a structured assessment of technology intangible assets, including AI capability, proprietary data, and human capital in the AI function. It is designed to complement traditional due diligence with systematic intangible asset evaluation.
AI washing is a valuation risk disguised as a technology problem. The due diligence protocol — capability inventory, architecture review, infrastructure verification, and talent assessment — provides a systematic defence. In a market where AI claims carry 25-40% valuation premiums, the ability to distinguish genuine AI capability from marketing is worth millions in avoided overpayment. Every investor and board member needs this skill.
Ivan Gowan is Founder and CEO of Opagio. He spent 15 years as a senior technology leader at IG Group (LSE: IGG), where he evaluated hundreds of technology claims from vendors, acquisition targets, and internal teams. Learn more about the Opagio team.
The SEC has made AI-washing an enforcement priority, and for good reason: billions in capital are being allocated based on AI claims that range from exaggerated to fabricated. After 15 years evaluating technology claims at IG Group, here is the 7-point checklist that separates genuine AI capability from performative marketing.
Read more →The EU AI Act takes effect in August 2026, making AI governance a compliance requirement — not just a best practice. This article provides a practical governance framework covering risk classification, oversight structures, accountability mechanisms, and audit processes for organisations investing in AI.
Read more →
PE operating partners managing 5-15 portfolio companies face a new dimension of value creation: assessing which portfolio companies are building genuine AI intangible assets and which are pursuing fashionable but value-destructive AI-washing. Here is the assessment framework that separates signal from noise.
Read more →Get the latest insights on intangible asset growth and productivity delivered to your inbox.
Book a free consultation to see how the Opagio Growth Platform can help your business.