For the past two years, barely an agency credentials deck has landed without the letters AI stamped somewhere across it. New tools, new practices, new “AI-first” positioning: the language has become so ubiquitous that clients could be forgiven for assuming the entire industry has quietly rebuilt itself around machine learning. A new audit suggests the reality is a good deal more modest.
An audit of 100 agencies conducted by iStories found that just nine could point to public evidence of coordinated, meaningful AI integration across their operations. The remaining 91 either made no visible claims at all or, more tellingly, talked a big game without leaving much of a trail to back it up. The finding, reported by Campaign Brief, draws a sharp line between what agencies say about their AI capabilities and what they can actually demonstrate to the outside world.
The claim-versus-proof gap
The distinction the audit draws is an important one, and it is worth sitting with. This is not a measure of how many agencies are using AI in some form. Almost all of them are, whether that means a strategist leaning on a large language model to speed up a first draft, a designer generating mood-board images, or a media team using the optimisation tools baked into the major ad platforms. That kind of ad hoc, individual use is now close to universal and largely invisible.
What the audit set out to find was something harder: evidence that AI has been woven into how an agency actually operates, in a way that is coordinated across teams and visible to clients and the market. On that far tougher test, the field thinned out dramatically. Nine agencies in a hundred could show it. The rest, by the audit’s reckoning, are either quietly getting on with it behind closed doors or dressing up scattered experimentation as strategy.
That framing matters because the advertising industry has always been fluent in the language of transformation. Agencies sell reinvention for a living, and it is not surprising that many have reached for AI as the latest proof point of relevance. The awkward truth the audit surfaces is that the marketing of AI capability has sprinted well ahead of the operational substance behind it.
Two ways to read the numbers
There are, broadly, two schools of thought on what a result like this actually tells us, and both have a point.
The more critical reading is that the industry has a credibility problem. If clients are being pitched AI-enabled thinking and AI-accelerated production, and only a small minority of agencies can show any coordinated evidence of it, then a lot of what is being sold is aspiration rather than capability. Procurement teams inside big advertisers have grown noticeably more sceptical over the past year, and an audit like this hands them a ready-made reason to ask harder questions before signing a scope of work. For agencies that have genuinely done the work, the noise from everyone else risks devaluing the real thing.
The more forgiving reading is that public evidence is a blunt instrument. Plenty of agencies are doing serious work with AI inside client engagements governed by strict confidentiality, and they are hardly going to publish case studies detailing how they use it. Integration that is genuinely embedded in a workflow tends to become invisible precisely because it is working, not because it is absent. On this view, an audit built on what is publicly visible will always understate what is actually happening, and the nine-in-a-hundred figure says as much about the industry’s caution around disclosure as it does about its capability.
Both things can be true at once. Some agencies are almost certainly overselling, and some genuinely capable ones are almost certainly under-showing. The value of the audit is less in the precise number and more in forcing a conversation the industry has been happy to avoid: if everyone claims to be AI-integrated, the claim itself stops meaning anything.
What it means for Australia
The audit lands squarely in the Australian market, and the timing is pointed. Local agencies are navigating a stretch of soft client spending, consolidation among holding companies, and clients who are increasingly bringing capabilities in-house. In that climate, AI has become one of the few pitch narratives that reliably cuts through, which is exactly why the temptation to overstate it is so strong.
Australian advertisers, for their part, have their own reasons to care about the gap. The industry has spent much of the past year worrying aloud about “AI slop”, the flood of cheap, generic, machine-made content that threatens to erode the value of creative work. An agency that can genuinely show coordinated, thoughtful AI integration is offering something quite different from one that is simply feeding briefs into a chatbot and hoping for the best. The audit gives clients a rough language for telling the two apart.
There is also a governance dimension that is unavoidable in the Australian context. As the federal government continues to work through its approach to AI guardrails, and as questions about transparency, data handling and disclosure sharpen, agencies that treat AI as a marketing slogan rather than an operational discipline may find themselves exposed. Clients in regulated sectors such as finance, health and government increasingly want to know not just whether an agency uses AI, but how, with what oversight, and with what protections around their data. Vague claims will not survive that kind of scrutiny.
For the nine agencies the audit credits with genuine integration, the finding is quietly useful. It suggests that doing the work properly, and being able to point to it, is still rare enough to be a real point of difference. In a market where AI language has been so thoroughly commoditised, demonstrable capability may end up being worth more than any amount of positioning.
What’s next
Expect the audit to feed directly into how agencies and clients talk to each other over the coming year. Advertisers running reviews are likely to start asking for evidence rather than assertion, and agencies that have invested seriously in AI will be keen to make that visible without breaching client confidentiality, a balance that is harder to strike than it sounds. The industry bodies that shape best practice will also be watching, because a gap this wide between claim and proof is the kind of thing that eventually invites either self-regulation or the external variety.
The broader lesson is simple enough. AI has moved from novelty to expectation faster than most agencies could build the substance to match, and the audit is a reminder that the market will, sooner or later, ask to see the working. For an industry that trades on persuasion, being caught claiming more than it can show would be a poor look indeed.
Sources: Campaign Brief
