How to spot 'AI washing': a practical guide to telling real AI from corporate spin

Somewhere between the release of ChatGPT and your next quarterly earnings call, every company on Earth seems to have become an AI company. The local accountancy is now "AI-powered". Your bank's app has gained an "intelligent assistant". Even the corner cafe has a sign promising an "algorithmically optimised" loyalty programme.

As The Guardian recently reported, firms are scrambling to rebrand themselves as tech-focused in what's increasingly being called "AI washing" — the practice of dressing up ordinary products, services and business processes in artificial-intelligence language to attract investors, customers or talent. It's the cousin of greenwashing, and it's just as misleading.

For Australian consumers, investors and workers trying to navigate this hype cycle, the question is no longer whether a company says it uses AI. It's whether the AI actually does anything. Here is a working guide to telling the two apart.

What AI washing actually looks like

AI washing covers a spectrum. At the mild end, it's a marketing team sprinkling "AI-driven" across a product page that describes what used to be called "a rules-based system" or even "an Excel formula". At the more egregious end, it's outright misrepresentation to investors — a serious enough problem that the U.S. Securities and Exchange Commission has begun bringing enforcement actions against companies overstating their AI capabilities.

The pattern is familiar to anyone who watched the dot-com era, when companies tacked ".com" onto their names and saw their share prices leap. The crypto and blockchain pivots of 2017–2021 played out the same way. AI is simply the latest narrative to be commercialised, and the incentive to overstate involvement is enormous: AI-themed companies have, for the last two years, attracted disproportionate venture funding, valuation multiples and media attention.

Five questions that cut through the spin

1. What problem is the AI solving — and was it actually unsolved?

Genuine AI deployments tend to be described in terms of a specific problem: predicting equipment failure in a wind farm, triaging radiology images, detecting fraud patterns across millions of transactions. Vague claims — "we use AI to deliver better outcomes for customers" — are a red flag. If a company can't tell you what the model does that a human or a simpler tool couldn't, the AI is probably decorative.

2. Where does the model come from?

Most companies claiming to "use AI" are calling someone else's API — usually OpenAI, Anthropic, Google or Microsoft. There is nothing wrong with that. But there is a meaningful difference between a business that has built proprietary models trained on its own data, and one that has wired a chatbot into its FAQ page. Ask: is the AI built in-house, fine-tuned on proprietary data, or simply licensed? The honest answer reveals a great deal about defensibility and depth.

3. What data is it trained on, and who owns it?

AI models are only as good as the data behind them. A genuine AI company can describe its training data — volume, source, licensing, refresh cadence. An AI-washed company gets uncomfortable when you ask. For Australian businesses, this question also has regulatory weight: the Privacy Act, the forthcoming reforms around automated decision-making, and the Office of the Australian Information Commissioner's guidance all hinge on what data is being used and how.

4. How is performance measured?

Real AI has metrics: accuracy, precision, recall, false-positive rates, latency, hallucination rates. AI washing has adjectives: "smart", "intelligent", "seamless", "next-generation". If a company can't produce a number that describes how well the model performs — or worse, doesn't appear to measure it at all — the AI is decoration, not infrastructure.

5. Who is accountable when it goes wrong?

This is the test most AI-washed firms fail. Genuine AI deployments come with a governance layer: human oversight, escalation paths, model risk committees, audit logs. Marketing-grade AI doesn't, because there's nothing to govern. If you ask "what happens when the model makes a mistake?" and the answer is a blank stare, you're looking at a sticker, not a system.

The investor angle: where AI washing becomes legally risky

AI washing isn't only a marketing nuisance — for listed companies it's becoming a disclosure issue. ASX-listed firms making material claims about AI capability are subject to the same continuous-disclosure obligations as any other forward-looking statement. Overstating AI integration to inflate a share price is, in principle, no different from overstating revenue.

Investors should look at the gap between what a company says in its glossy investor deck and what it says in its annual report's risk-factors section. Genuine AI adopters disclose AI-related risks: model drift, regulatory exposure, third-party dependency on foundation-model providers, data-governance liabilities. Companies that talk about AI on the front page but don't mention it in the risk section are usually doing PR, not engineering.

The geopolitical layer Australians shouldn't ignore

There's a second reason to care about distinguishing real AI from washed AI: the supply chain underneath it is becoming politically contested. Analysis from the Carnegie Endowment for International Peace on U.S.–China technological decoupling highlights how access to advanced chips, model weights, cloud infrastructure and even research talent is increasingly subject to export controls and national-security review.

For an Australian company genuinely operating at the AI frontier, this matters enormously — it shapes who you can buy from, where you can host, and which markets you can sell into. For an AI-washed company, it doesn't matter at all, because there's no real exposure. That asymmetry is itself diagnostic. If a firm claims deep AI capability but seems untroubled by chip availability, cloud-region restrictions or compute costs, the claim deserves scrutiny.

The advertising and brand-safety problem

AI washing also intersects with another corporate scramble: brand safety in digital advertising. The Washington Post recently reported on controversial digital ad placements leaving tech companies scrambling, a reminder that the same automated systems being marketed as "AI-powered" routinely place ads next to content advertisers would never knowingly endorse.

This is instructive. The ad-tech sector has been describing its targeting as "AI" for over a decade, and yet the basic problem of context — understanding what a piece of content is actually about — remains unsolved at scale. When you see a company claim its AI "understands" content, ask whether the same company's ad placements ever end up somewhere embarrassing. The honest answer is almost always yes, which tells you what the AI is and isn't doing.

A simple heuristic

If you remember nothing else, remember this: real AI shows up in operating costs, hiring patterns, and risk disclosures. AI washing shows up in press releases, landing pages and conference keynotes.

Companies that have genuinely embedded AI are spending serious money on compute, hiring machine-learning engineers and data scientists, retaining specialist legal counsel, and disclosing new categories of risk to their boards. Companies that are AI washing are spending money on agencies, rewriting their "About" page, and updating their LinkedIn headline.

The asymmetry is your friend. The next time a company tells you it's "AI-first", "AI-native" or "powered by AI", run the five questions above. If the answers are specific, measurable and slightly boring, you're probably looking at the real thing. If the answers are glossy, evasive and exciting, you're looking at a rebrand.

The AI revolution is real. So is the marketing department's interest in pretending to be part of it. Telling them apart is now a basic literacy skill — for investors, for customers, and for the rest of us trying to work out whose claims to trust.

Related on Bleen

Sources