What a Journalist Already Believes About Your Company — Before You Pick Up the Phone

2026-04-24

According to Muck Rack's 2026 State of Journalism report, 82% of journalists now use at least one AI tool in their work. A quarter of them use it specifically for research. Which means that before your press team answers a single question, a reporter may already have a mental model of your company built from whatever a public AI system surfaced about you.

The typical media relations playbook assumes the journalist starts from a neutral position. You get called for comment, you provide context, and you shape the story from there. That model is under growing pressure.

A growing number of reporters now run a quick query through Perplexity, ChatGPT, or similar tools before picking up the phone. The result is an informal briefing: what the company does, who leads it, what controversies have followed it, how it's generally positioned in its sector. Then they call you.

This isn't bias. It's a rational efficiency move. A journalist covering fifteen stories a week uses AI to compress hours of background research into minutes. The problem is what happens when that briefing is wrong.

The frame that shapes every question

Public AI systems draw on training data that can be months or years old. They generate confident, readable summaries without signalling whether the information is current or outdated. They may cite controversies that have long been resolved, repeat characterisations from a previous press cycle, or describe an organisation's strategy in language from two leadership regimes ago.

When a journalist walks into an interview with that frame, your answers get filtered through it. A question like "how have you addressed the concerns raised about X?" presupposes something the AI told them. Correcting the presupposition is harder than answering the question. You're not clarifying the record — you're dismantling a premise that was formed before the conversation started.

This is what makes media risk distinct from other AI perception risks. The public may never see the AI summary directly. But a journalist might — and it shapes every question they ask, which shapes the story they write, which shapes what the public eventually reads.

How errors compound

Published coverage becomes the source material that feeds the next round of AI summaries. A story written on a flawed premise can harden that premise into the record. Other journalists researching the same organisation later encounter the article, see it cited by AI as authoritative, and treat it as established fact. The next summary is built on the last story's errors, which were themselves built on an earlier AI summary's errors.

The pace at which this happens has accelerated. AI tools don't replace journalism — but they do accelerate how assumptions form and how they travel. An inaccurate characterisation that once might have faded quietly can now get woven into multiple coverage cycles before anyone flags it.

A different kind of media preparation

Organisations that navigate this well are the ones that treat AI perception as a form of pre-interview media preparation. Before any significant media interaction, it's worth asking: what does a public AI system say when someone queries this company right now? Is it accurate? Is it current? Does it reflect the organisation as it actually exists — or an earlier version that no longer applies?

The question used to be: what will journalists find when they search for us? Now it's also: what will AI tell them before they even start searching?

The answer to that second question is increasingly the one that matters most.