PR's Uncomfortable AI Irony
A year ago, 37 per cent of Australian journalists said they were using generative AI in their own work. Medianet's 2026 Australian Media Landscape Report puts that figure at 54 per cent now: more than half of the profession, in twelve months.
Ask the same journalists what they think of AI turning up from the other direction, in a pitch from PR, and the number runs the opposite way. Seventy-eight per cent say an AI-written pitch lowers their trust in whoever sent it. Forty-eight per cent reckon they can pick one almost every time. Sixty-six per cent still rate PR professionals as an important source for stories, a figure that hasn't moved. What's moved is patience for how those stories arrive. That's the double standard sitting at the centre of the PR industry's own relationship with AI.
PR's own AI double standard
Here it is, worth admitting rather than pointing at from a safe distance: Stitch uses AI in its own workflow too. Research, first drafts, keeping track of what's already been pitched where. Most agencies do, whether they say so or not. Pretending otherwise would be its own kind of dishonesty, and in an industry that runs on trust, dishonesty is the one input that never comes cheap.
We wrote about the trust cost of AI-written pitches in more detail back in June, when this same report's headline figure first crossed our desk. What's changed since then isn't the trust problem. It's how many journalists are now using the exact tool they don't want to see on the other end of a pitch.
So what's actually different?
Not the tool. A journalist running a story through AI to draft a summary and a PR adviser running a pitch through AI to draft an opening line are, mechanically, doing the same thing.
What's different is whether a person who actually understands the story is still doing the thinking. Deciding which editor genuinely wants this, not just which one is on the list. Checking whether a claim survives contact with a fact-checker before it goes anywhere near a journalist's inbox. Knowing the difference between a pitch that's efficient and one that's just generic.
In practice, that looks unglamorous. It's deciding a client's figures won't survive a follow-up call and cutting the line rather than sending it anyway. It's knowing a regional daily wants a different note to a trade title, and that AI doesn't know that until someone tells it. None of that shows up in a word count. All of it shows up in whether the pitch gets a reply.
That's close to the same line regulators are converging on globally as they work out how to govern AI more broadly: the tool can inform the work, but a person has to be accountable for what actually goes out the door. In PR, where credibility is the product being sold, that line carries more weight than in most industries, not less. It's also the standard we hold our own AI use to: informing the work, never replacing the judgement call at the end of it.
Could you tell?
Scrolling Medianet and PR Newswire, watching what lands in journalists' inboxes, is part of how we keep across the state of pitching. The pattern the report describes is easy to recognise once it's pointed out: identical structure, the same three adjectives, a quote that reads like nobody actually said it out loud. Journalists are noticing it. Medianet's data says nearly half of them can tell almost every time.
We made a related argument in August, that credibility is a commercial asset, not a nice-to-have. This is the same case from the other side of the inbox.
So here's the question worth sitting with, whichever side of it you're on: when a pitch reaches you, could you tell if a person actually wrote it?
Sources
Medianet, 2026 Australian Media Landscape Report (March 2026) — medianet.com.au