ChatGPT for travel marketing and PR: We put it to the test

ChatGPT. You know the name, you’ve seen the headlines. You’ve read post after post about this seemingly magical technology that can do anything you can do on a computer. You’ve been curious about it, anxious about it – you might have even played around with it yourself.

In this post, we’ve used ChatGPT to do some common tasks travel PRs, marketers and travel journalists would do, and presented our findings.

Like all new technology there are two sides to it: Travel brands can benefit from this tech right now; if it’s used wisely as a tool, and not as a shortcut to success. However, there are also some strong arguments against using it, many of them relating to ethics, information bias and emotion.

Let’s give you the full picture. We’ll inform you as best as we can based on our experience with ChatGPT, and give you some ideas about how to use it, how not to use it, and where this is all going in the future.

In its own words, ChatGPT is “an AI-powered chatbot developed by OpenAI, based on the GPT (Generative Pretrained Transformer) language model. It uses deep learning techniques to generate human-like responses to text inputs in a conversational manner.”

In other words, it is a conversational language simulator. It can also do other tasks like data sorting, and idea generation, depending on the questions you ask it.

What makes ChatGPT unique is that it can produce human-like responses. The output could be anything from poetry to press releases – in theory, anyway (song lyrics produced by ChatGPT are… well, see for yourself).

It works using machine learning, or ML. The algorithm was fed with reams upon reams of data – and it learned how to emulate human-like responses, in all manner of styles, from the data it was trained on.

This sort of training is a blessing and a curse, which we’ll get to in a little bit.

Far from it. ChatGPT is simply the latest in a long line of chatbots in OpenAI’s GPT series (GPT4 is in development).

Google is also getting in on the action, launching its own version, named Bard AI. And chatbots are touted as the future of everything – from homework to finance. But given Bard’s opening night flop, there’s a very clear sign hanging over AI chatbots right now;

Still, a lot of folks in SEO, PR and Content Marketing are convinced that ChatGPT is the answer to all their content problems (and if you’re a writer – maybe the cause of your demise).

So, let’s put ChatGPT to the test.

1. Write a press release
2. Write a blog post
3. Find a newsworthy hook for one of our travel clients that we could pitch to a journalist
4. Produce a factsheet on a tour operator itinerary

Here’s what we learned.

We asked ChatGPT to write a press release for the launch of Lanserhof Sylt. We launched their state of the art wellness retreat on Sylt a while back.

The press release copy was factually accurate, and the copy wasn’t too bad. However – the text was pretty much Lanserhof’s own wording. No huge surprise, as ChatGPT is trained on internet data up until 2021, so it would have had access to Lanserhof’s website.

When compared to our own press release for the launch, it was pretty similar – but lacked the sparkle and zing of a press-savvy human writer.

Functional? Absolutely. Nothing to write home about, though.

Verdict: 3/5 – functional but flat.

So far, ChatGPT seems to be good at producing copy that’s factually accurate.

But a huge part of PR is influencing audiences, so that theytravel with our clients. That means a considerable amount of our time is spent drafting opinion pieces, and putting together thought leadership articles on behalf of our clients.

How did it do?

ChatGPT’s effort was a total disaster.

It can’t do empathy, humour, irony, opinion – it’s not a creative engine. At best, whatever we got back was a colossal fail. Unusable, no matter how much we massaged it into shape.

Verdict: 0/5 – funny responses. But not funny haha.

Anyone can write a press release with some training – and although some clients think that is what PR agencies spend most of their time doing, it actually isn’t at all. It’s maybe 5% of our time, tops.

Most of our time goes into coming up with newsworthy hooks and creative ideas – or putting a new spin on an old story, so that we can pitch the client to the media in a new light. That’s because, understandably, travel brands don’t always have a lot of news. After all, you can’t refurbish your hotel restaurant every year!

Read more – “My PR agency always asks me for news, but we don’t have any!”

When we put ChatGPT up to the challenge, it admitted that this kind of response is outside of its abilities. Here’s the answer it gave us:

“…while ChatGPT can write a press release, it may not always find a newsworthy hook that is relevant and meaningful to your specific audience and context. The most effective press releases are those that are tailored to a specific audience and address a newsworthy event or announcement in a clear and concise manner. To ensure the success of your press release, it is important to carefully consider the news hook and the audience you are trying to reach, and to craft your message accordingly.”

And this is actually a great answer – far more useful than its attempts at creativity!

It brings us to this immutable fact; really understanding your travel brand’s audience is not going to come from old fashioned demographics. It’ll be based on interests and passions, on softer, less easily understood data.

Learn more about using data to find your audience and tell your travel brand’s story.

ChatGPT did okay here, because it didn’t try to “fake it ‘til you make it”. It gave us insight. This shows the true potential of chatbots; they can help us research facts quicker, and find holes in our own understanding – but right now, only the human brain can put together ideas in a new and unexpected way.

To evoke an emotion, you have to be capable of feeling emotions. To really understand our clients’ target audiences psychologically, we have to be able to empathise and mirror them.

Verdict: 2/5 – honesty is the best policy.

Pretty good, ChatGPT – but because the training data is limited to 2021, it’s not up to date. Personalised itineraries were unexpectedly decent, based on likes and interests, but the limiting factor of two year-old data doesn’t give us much help for anything new.

Verdict: 4/5 – with more data, this could be a huge timesaver.

Could ChatGPT fact check information, or help with research in travel journalism?

We tried it out – and yes it can. There’s a huge BUT coming up, though.

Remember we brought up bias at the top of this article?

The quality of ChatGPT’s output is dependent on the quality and accuracy of its training data. If it was trained on biased or inaccurate data, this will affect the quality and reliability of its outputs.

In all likelihood, it was fed highly biased, inaccurate data at some point in the machine learning process – because human-made content is also biased.

We are aware this now goes into the ethics of AI, but we actually have experience here – in the study of computational linguistics.

Human beings come with all sorts of cognitive biases, like recency and confirmation bias.

Our inherent biases come out in our behaviour – and in our data.

Basically, the information ChatGPT is trained on came from the internet, including from forums. While quality assurance and filtering will catch out the most obviously offensive stuff, none of the training data would have been checked for biases or prejudice, or factual accuracy.

We’ve seen the most shocking example of this happen before, in public. Microsoft’s Tay chatbot went from zero to toxic in 16 hours, because it was fed racist, misogynistic hate speech by a fringe incel group.

More insidiously, AI hiring tools have been found to have a preference– for straight white males, that is – because they were trained on past hiring decisions. Not on candidate merit or performance.

We need to be really careful here. Because most of the data in the world is biased in favour or against a person, idea or situation.

If anything, the presence of AI chatbots makes the role of journalists more important than ever.

Can ChatGPT help travel brands with influencer marketing and brand collaborations?

We’re always being asked; “who are the most important travel influencers?”

And we always answer “it depends”.

What’s right for one brand could be totally wrong for another. Same for brand collaborations. This comes back to the importance of knowing your audience.

Working with influencers and third-party content creators (either sponsored or unsponsored) can be an excellent way to get exposure on your brand. But, how do you know where to find the best influencers to bond and cultivate relationships with?

Read more – How to Work with Travel Influencers: The Ultimate Guide

This is where AIs like ChatGPT could help cut that legwork – by matching products and people. It could one day search the internet for cultivated audiences that are likely to appeal to another brand and values. It could help filter influencers, analyse who has high engagement and who doesn’t, or suggest people more aligned to your brand based on their history.

And there’s another side to this.

The AI-generated influencer, Lil Miquela, uses chatbot technology to create an entirely digital persona.

Despite the fact she doesn’t exist, millions of followers are totally on board – and brands like Prada have collaborated with the bot, allowing both to gain further exposure. Is this the right avenue for everyone? No.

But it sure is interesting to see the possibilities of AI unfold, and the power it can have when it’s done right. AI can even generate travel photos and win awards… no – really. Again, this all comes down to knowing the audience, and whether this kind of content is going to resonate.

Here’s ChatGPT’s answer:

“ChatGPT can be used to generate social media content, such as tweets, posts, and captions, that engage with customers and followers. The language model can also monitor social media for mentions of the brand and respond in real-time, helping to build a positive online reputation and manage crisis situations.”

And here’s our take: it can certainly be useful to write social media captions. We’ve tried it ourselves, and it works.

Here comes another big BUT…

Having been on the sharp end of crisis PR for travel brands more than once, we strongly recommend you stay well clear of ChatGPT for crisis management.

To gather data and information, and monitor? Yes – use it for that – but ultimately, you’ll need a human, empathetic touch. You need tact, care, and emotional intelligence.

Someone who understands the difference between issues management and crisis management. A death at a resort is a crisis. The ice cream machine breaking down is not.

Things can get very emotional. And we already know ChatGPT doesn’t do emotion, or even feign it, very well.

While ChatGPT can generate “good enough” content, it lacks the personal touch that comes from a human PR professional. Customers and influencers may be able to tell that they are communicating with a language model, which could harm the brand’s reputation and reduce the impact of its PR efforts.

Google has its own natural language model called LaMDA. It made headlines last year, when the person testing it for ethics and bias believed it to be a sentient, thinking, feeling being – a child-like entity that needed protection.

He was fired. And subsequently proven wrong (quite a few times, actually).

Regardless of whether humans ever achieve truly “General AI”, a scenario where a chatbot replaces Google is frightening to marketers.

We don’t see it that way.

Tools like ChatGPT will revolutionise how we work, but for the better – as long as governments and developers put strong ethical guidelines in place.

For a deep-dive into this, check out the Reith lectures on AI by Stuart Russell, on BBC Radio 4.

ChatGPT is powerful in so many ways. Chatbots like this will likely provide a starting point for PR, Content Marketing and journalism; compiling a first draft, putting together sources it can crawl on Google, and taking out the laborious research.

ChatGPT can excel in doing a rough draft or an outline.

But it completely fails in creative applications.

It can’t come up with newsworthy stories, or put a new twist on an old idea. It can look back, but it can’t create anything new – although it might inspire the right mind working with it.

It has no understanding of nuance, context – or audience insight.

It won’t be able to write a press release, or website copy without a human to help. It doesn’t understand nuance, history, context, or audiences. It’s our job to know how to adapt the tone and comms style to each brand.

It has no emotion.

As communications professionals, we understand how we want people to feel at the end of interacting with our brands.

This can’t be bottled. It can’t be replicated by ChatGPT. It may never be replicated by a machine.

No strategy.

Without a clear strategy in place, ChatGPT will fail to fly for most brands. Strategy requires a big picture approach; coordinating across multiple teams, people and personalities in most cases.

It also needs a human touch, to align it with the reality of the brand.

The outlook?

We will probably settle into a hybrid. People who are skilled in creativity, strategy, and emotional audience insights will be in demand. Journalists, PR professionals, and travel marketers will be required to put things into context and provide opinion.

And we’ll always need a human to lead us through a crisis.

But jobs that involve more menial or routine tasks will be replaced. This has always been the case with any new technology; from automated call centres, all the way back to the printing press.

Old jobs may change, and new ones will emerge. Large language models will enable new ways of working, and also lead to new and as yet unimagined jobs – but we’re not being replaced. Not yet, anyway!