#42 AI and the Future of Journalism with Harriet Meyer

Steve:
We’re very excited to be joined by Harriet Meyer, an award-winning personal finance editor, journalist, and copywriter with over 20 years’ experience across national newspapers, magazines, and websites. Harriet still writes regularly. And if you haven’t read her series of how AI can help you with various aspects of your life in The Observer, I recommend you do so. It’s brilliant. But she’s also pivoted her career to become an AI trainer for journalists, content teams, and PR professionals. And that’s the subject we’re going to be discussing today, how AI is changing journalism.

Welcome to the Digital PR Podcast, Harriet. Thank you so much for joining us.

Harriet:
Thank you. I’m really excited about this. Delighted to be here. Thank you for having me.

Louise:
Great to be able to chat to you. Thought it might be a good opener question to basically ask, you know, not to labour the point, but you’ve been a journalist for a while, so you’ve probably seen a lot of how things have changed. We’re now in this age of AI. What have you seen that’s changed the most from a journalist perspective around how stories get commissioned, produced, distributed? What are the big changes?

Harriet:
Honestly, commissioning hasn’t changed as much as people think. I think we’re all rushing to try and catch up with where AI is now and we’re a long way from doing that, to be honest. If you’re a trusted journalist or if you’re receiving briefs and commissions, they’re pretty much standard as they used to be. Often you’ll just receive a line from an editor if you’ve worked with them for a while. The commissioning process isn’t some lengthy thing.

But where AI I think is changing things is more inside the workflow. So it’s really about it becoming a super powerful research and analysis partner so journalists can interrogate far more material far more quickly, and that genuinely helps with investigative and analytical work.

But as I said, newsrooms are still in the early stages of using this well, largely because editorial teams are under enormous pressure and behaviour change takes a lot longer than switching a tool on.

For some actual examples, the Financial Times scraped Russian adoption websites to help identify children abducted during the invasion of Ukraine and used image matching software to compare tens of thousands of faces against Ukrainian government records. That’s kind of like a really nitty-gritty use of AI.

Journalists have used ChatGPT to classify tens of thousands of federal complaints against Tesla workers, and it’s really in those sorts of things that I’m hearing more work being done in journalism around AI.

And of course there’s the ability to distribute stories in different ways. So in more direct audience relationships and more direct use of social media, TikTok, the creator economy is a big threat to media as well. The merging of those worlds and the distribution of stories is changing as well, I would say.

Even though we’re years on, more than three years since ChatGPT launched, we’re still very early doors.

Steve:
Coming back to that point, because I think, well, I know we’re very fascinated by your career arc, Harriet, in that you seem to spot that generative AI was going to be a big deal for journalism and the media industry kind of ahead of a lot of people. How did you do that? What was the kind of the penny-dropping moment when you realised it was going to be a big thing rather than just another fad?

Harriet:
Yeah, I think that is because I’ve been around for quite a long time. So I started in papers when it was all about the paper. I started at The Telegraph and I was writing several features a week and the online version of the paper was somewhere in a back cupboard somewhere and really not considered important and not the real journalists.

Then I saw the digital revolution and how that transformed and put pressure on media. I’ve seen how the industry has changed and the demands on journalism and journalists and the lack of time that we all have and how much pressure there is to produce more and more stories.

And then when ChatGPT landed at the end of 2022, I was in an editorial role and part-time editorial role and freelancing as well and doing lots of different work. I was getting a little bit itchy feet in terms of wanting a new topic to get into.

ChatGPT landed and I started dabbling with a few prompts. I saw the power of the early stages of this technology. I thought, God, this is the first version. It’s spitting out story ideas. It’s basically doing a rather crappy version of what we do.

Then I thought, well, I’m going to get ahead of this. This is going to be bigger than the digital revolution. It’s going to fundamentally change the way that we work, the way that we live.

I started reading everything and learning as much as I could at very early stages. Since then I’ve spent hours a day and listened to thousands of hours of podcasts and all sorts. I just became fascinated. It was just a new topic that I found really interesting and just learned as much as I could and experimented. It just went from there.

I think once you start to do that, you get a real understanding of the power of these tools and how they could potentially change things in media.

Steve:
How do you feel about it though? Because what I’m interested to know is that you were a very early adopter, but you seem incredibly positive about AI in so many ways. I saw you talk at a conference last year and you’re very eloquent and informed.

Do you feel positive about the direction of travel, or are there still fears or anxieties around it? Because it’s so mind-blowing what has happened and what is happening.

Harriet:
Generally positive, but I’m also naturally sceptical, which sounds odd. I’ve been a journalist for decades, I pick into things.

I am also terrified. It’s not that I’m just approaching AI with relentless positivity at all. I feel very conflicted about it. The tech giants, I wouldn’t trust them an inch.

What I do believe is that there are potentially phenomenal uses of this technology which can improve our lives. There is a big focus on the negative sides of AI, particularly in media. Understandably people are very fearful, very threatened by this tech that could replace our jobs and make us redundant effectively as humans.

But I also think it’s so transformative and it’s very exciting at the same time. It’s not black or white. There’s a real grey area with AI.

Where I try to approach it from is a very grounded perspective. Yes, it’s exciting being immersed in this tech and experimenting and working out how it can transform workflows. But at the same time I am very sceptical of the tech giants.

I do believe that a lot of what’s going on with the scraping of data is morally wrong. I really empathise with where journalists and PRs and creatives are sat with it.

I think that’s where I saw that I could try and bridge that gap between the old and the new worlds, because there was so much fear around it and anger as well.

So it’s trying to actually help people in this process of doing the AI training that I’m doing – seeing the mindset change. It is a mindset shift, and it’s helping people overcome that so they can at least find their place in what comes next.

Louise:
You mentioned how you do training, and some of that is training PRs in how to work with journalists, understand newsrooms, things like that. Taking AI slightly out of it, is there anything in general that you feel like PRs still misunderstand about how journalism works?

Harriet:
I know loads of great PRs and I’ve had lots of PRs who have a very clear understanding of how journalists work. But I think the biggest misunderstanding is probably quite how little time journalists have.

The pressure has increased, and because of that journalists are much more reactive than sometimes people realise.

Formulaic pitching isn’t going to cut it anymore. It has to be relevant. It has to be really speedy. There has to be trust there more than ever.

The availability issue is that you have to be really reactive. More so than ever, if a PR says “I’ve got a case study”, that case study has to be available to speak ideally within 24 hours. It’s not that it’s just getting faster, it really is faster.

This is where, if used strategically, coming back to AI, you can slot it into the workflows to actually help journalists, but you have to know how to use it well.

Steve:
Obviously you train newsrooms and journalists on AI and how to use it, and PR professionals likewise. In your opinion, where is AI most useful in PR and journalism workflows, and where is it not being used well? Where is it mostly noise?

Harriet:
There is a hell of a lot of noise around it, and to be honest we’re now in 2026 and it’s got worse.

Most people can ignore most of the noise. But it can genuinely support human thinking, strategy and things like holding context.

There’s something in the AI world where they’ve talked about moving from prompt engineering to context engineering. That really just means rather than using AI like Google, you’re giving it as much context and background information and data as possible.

Where I see the power of AI is its ability to analyse huge volumes of information and spot patterns.

Humans are actually quite bad at holding tons of background detail at once and AI isn’t. That’s where it adds value.

Where there’s lots of noise is around the idea of real end-to-end agentic workflows. You go on LinkedIn and see people shouting about massive end-to-end systems, “if you’re not doing three agentic deep research queries before breakfast you’re going to be unemployed”.

That’s not helpful. More often than not, particularly in the media, work is nuanced. Every piece of work might be slightly different. But AI can hold the context of a client or campaign in a way we just can’t.

Louise:
Do you think what you really need to start out on this journey is just purely a ChatGPT paid-for licence, for example?

What I often see is various different tools which are AI-powered, but they seem like wrappers around what ChatGPT can do. Is your advice to start with one model, or is there benefit in lots of specialised tools?

Harriet:
Tool overwhelm is a big thing.

Someone texted me the other day asking if I’d heard about a new tool being shouted about in newsletters. I hadn’t even got to that newsletter yet.

What you need to start off with is a paid account, because fundamentally if you’re using a free account you’re paying with your data. Paid versions have more features and you can use them more heavily.

The big large language models are ChatGPT, Claude, Gemini and Copilot. Those are your four big ones. Just pick one.

It might be the one your office currently uses. Quite a lot of teams have their own versions of ChatGPT or Copilot or Gemini.

Then really get to grips with it. The strategies and techniques I train teams in apply across all the models, but you have to learn to use them well.

If you’re jumping around chasing hype, one tool for this, one tool for that, you’ll very quickly make more work for yourself.

Steve:
That’s reassuring to know because I think both of us have commented on feeling overwhelmed just flicking onto LinkedIn.

Harriet:
I know. Nobody can keep up with AI. Even the people making the models can’t keep up fully.

That’s why my training has evolved from basics to more of a strategic approach and mindset shift.

Strip it back. What do you actually do day to day? How can AI help you? What are your workflows? How could they be reimagined?

If you try to keep up with LinkedIn and tool releases you’ll just drown in it.

Steve:
Really keen to get some examples. What are some ways agencies are actually using AI to enhance their work, not just speed things up?

Harriet:
One powerful example is using things like ChatGPT projects for campaign planning.

If they’re set up correctly with the right knowledge files and context, the model can hold all the background information about a client or campaign. Then it becomes very powerful.

Also simple things like repetitive tasks, but also scheduled tasks inside tools like Copilot, ChatGPT or Gemini.

You can set them up so you get updates emailed to you about journalists or topics you’re following.

You can also create personas, like a sceptical editor or a client persona, and have the AI push back on your ideas. That helps improve thinking.

You have to think of AI as another team member that knows everything and nothing. You have to guide it.

Steve:
This is something I’ve had to get used to, using custom GPTs with personas like “you are a journalist” or “you are a client”. Sometimes it gives quite scathing feedback.

Harriet:
Well a lot of people complain AI is too sycophantic, so at least you’re pushing it out of that!

Louise:
When you’re working with PR teams, what are the first things you set up with them?

Harriet:
You have to get the foundations right.

You start with setting up the model properly, custom instructions, settings, understanding the different models and features.

You also need to understand how AI works. People think they understand it, but they often don’t.

If you don’t understand the foundations, you’re not going to get the powerful outcomes later.

Steve:
Let’s scare ourselves a bit. What are the biggest risks with AI use among PR teams and editorial teams?

Harriet:
The biggest risk is trust. We’ve already seen fake stories and fake experts. If you don’t understand how AI works you might trust it too much. It can sound confident and polished even when it’s wrong.

Used badly, it can wreck trust instantly with audiences or clients. Human judgement and editorial judgement have never been more important.

Steve:
That’s reassuring but also sobering.

Harriet:
It should be.

Louise:
Who decides what responsible AI use is?

Harriet:
Ultimately responsibility sits with the user. You can have policies and guidelines, but you can’t fully control how people use tools. This is really a leadership issue more than a tech issue.

Steve:
That’s interesting when you think about the next generation coming into PR and journalism.

Harriet:
Yes, and it’s going to be messy before it settles. People are scared and unsure what they’re supposed to be doing.

That’s why communication inside teams is really important.

Louise:
What are the rules for junior journalists using AI?

Harriet:
Generally they can use it for research and analysing data, but not for writing copy.

AI might be used at the start of the process and the end, for research or headlines, but journalists still write the articles.

Steve:
Should PRs disclose if they’ve used AI when pitching journalists?

Harriet:
If you’re just using AI for formulaic announcements that’s fine. But if you’re producing story packages, you must fact check everything. I’ve heard horror stories of case studies that didn’t exist.

Louise:
What about journalists disclosing AI use?

Harriet:
Most national publishers aren’t disclosing AI use because everything published is still human-written. But some smaller publishers do label content as AI-assisted.

Steve:
Where are humans still outperforming AI?

Harriet:
Human qualities like intuition, connection, reading the room. AI can’t sit in a room with someone and find the thread of a conversation. When you’re writing or pitching, you should imagine yourself as the reader or journalist receiving it. That kind of human judgement is still incredibly important.

Louise:
For our final question, what changes do you think we’ll see with AI over the next year or so?

Harriet:
There’s a great quote from Roy Amara: “we tend to overestimate the effects of technology in the short term and underestimate them in the long term.”

Right now we’re still in early stages. The biggest shift will be mental, a mindset shift. Teams will start reimagining workflows and processes.

Steve:
Do you think governments will regulate AI properly?

Harriet:
I’m not very optimistic. Many people in government don’t know much about AI either. They’re learning like we are.

There will probably be some kind of pushback at some point, but the change is still coming. I think the coming year is going to be a bit crazy.

Steve:
That’s a perfect way to end the podcast. Harriet, thank you so much for joining us. How should people get in touch with you?

Harriet:
Drop me a line at [email protected] or visit my website harrietmeyer.com where you can find my journalism and AI training information.

Louise:
Fantastic. Thank you so much. It’s been an absolute pleasure talking to you.

Harriet:
Thank you for joining us. Thanks very much.