Due Diligence Now Has to Catch AI-Washing, Not Just Check the Tech

When one company buys another, there's a stretch of work in the middle called due diligence – the buyer's investigation into whether the thing they're about to pay for is actually what it appears to be. Are the revenues real? Is the technology sound? Are there liabilities hiding in the contracts? Diligence is where a deal's assumptions get tested before the money moves, and the quality of that testing often decides whether the deal was smart or expensive.

In 2026, the questions diligence has to answer changed, because the thing being bought changed. And a lot of diligence checklists haven't caught up.

When AI becomes the reason for the deal

Start with a number that reframes the whole picture. According to PwC's 2026 M&A outlook, roughly a third of the 100 largest corporate M&A deals of 2025 cited AI as part of their strategic rationale. Not as a footnote – as a reason the deal happened. Buyers acquired companies to get AI capabilities, AI-driven growth, AI talent, or an AI position in their market.

Think about what that does to diligence. When AI is the reason a third of the biggest deals exist, the target's AI story stops being a technical detail to verify near the end and becomes one of the central things the price depends on. If the AI is real and durable, the valuation makes sense. If it isn't, the buyer is overpaying for a story. And stories, unlike audited financials, were never the thing diligence was built to stress-test.

This is why PwC doesn't hedge on the implication. Its 2026 guidance calls AI due diligence essential, and spells out what that means: assess the target's AI strategy and roadmap, estimate the impact of AI on the business over the next three to five years, evaluate the operating and capital requirements behind it, and test management's ability to actually execute the plan. That's a fundamentally different posture than "does the demo run." It's a judgment about whether the AI narrative driving the valuation will survive contact with reality.

The awkward symmetry

There's an irony worth naming, because it raises the stakes rather than lowering them. AI now sits on both sides of the deal table. PwC notes that AI tools are already being used to accelerate target screening and the diligence process itself. So you have diligence teams using AI to evaluate targets whose value is partly an AI claim.

That symmetry is easy to read as reassuring – everyone has better tools now. It shouldn't be. Faster screening doesn't make the core judgment easier; it just means more deals move through the funnel faster, which raises the cost of getting the judgment wrong. The tooling on both sides improved. The question of whether the target's AI is genuine or just well-marketed got no easier at all. If anything, it got harder, because AI also makes it cheaper to produce a convincing AI story.

What AI-washing looks like inside a target

The same dynamic that turned AI-washing into a regulatory concern in marketing makes it a valuation risk in a deal. "AI-washing," as a reminder, is describing something as more AI-powered or more capable than it really is. In a target company, it can take a few specific shapes, and naming them is the first step to catching them.

A company can present an "AI-powered" platform that turns out to be a thin wrapper over someone else's model – a service you could replicate with an API key, no proprietary data behind it, no real moat. It can show impressive AI-attributed revenue that, on inspection, is a handful of pilot customers who haven't renewed and may not. It can wave a roadmap full of capabilities that depend on a few key technical people who are already halfway out the door. In each case the surface looks like real AI value, and in each case the value evaporates under the right question.

And here's the problem: none of those show up if your diligence only confirms that the technology exists and functions. The demo will run. The product will work. They show up only when you ask the questions a wrapper can't survive. What is actually proprietary here – the model, the data, the workflow, or just the marketing language? Where does the AI-attributed revenue genuinely come from, and does it persist past the pilot phase? Can this team really ship the roadmap, or is the roadmap doing the work the product can't yet do? And could a competitor with a checkbook reproduce this "AI advantage" in a single quarter? Those are the questions that separate a real capability from a convincing one.

From afterthought to line item

The market is already starting to formalize this, which tells you it isn't a one-off concern. Commentary on 2026 diligence practice is beginning to list AI as a standard area of inquiry. China Briefing's guide to Hong Kong M&A this year, for instance, names data and AI governance among the core diligence focuses, sitting alongside long-established items like control, disclosures, and sanctions risk. I'll resist over-generalizing from a single market – that's one region's commentary, not a global standard yet – but the direction it points is telling. AI governance is migrating from "something thoughtful buyers might consider" toward a named line on the checklist.

That migration changes diligence scope in a concrete way. Things like model provenance (where the AI actually came from), data rights (whether the company legally controls the data its AI depends on), the durability of AI-driven revenue, and the gap between claimed and actual capability are becoming standard objects of inquiry. Not specialist add-ons brought in for the rare "tech deal," but part of the basic work of confirming that the thing the buyer is paying for is the thing that actually exists.

Where a boutique advisor earns the fee

This is exactly the kind of work that rewards judgment over headcount, which is worth saying plainly because it's where an independent advisor's value concentrates. Detecting AI-washing isn't a matter of running a longer checklist. A longer checklist is precisely what a well-coached target prepares to pass. It's a matter of knowing which claims to distrust and how to test them – the pattern recognition that comes from having seen, more than once, the difference between a genuine capability and a beautifully staged one.

A boutique diligence advisor who understands both the deal mechanics and the technology can ask the four questions above and, more importantly, read the answers – notice the evasion, feel the gap between the roadmap and the team, recognize the wrapper for what it is. A generic checklist run by a junior team will miss exactly that, because the target's whole story was built to pass the generic checklist. The judgment is the product.

The technology being good was always table stakes. In 2026, the diligence that actually protects a buyer is the diligence that can tell whether the AI story driving the price is true. Most checklists confirm the technology exists. The advisor's real job is to confirm the story does too – and to know the difference before the money moves.