Skip to content

Nothing gets forgotten

4 min

AI is increasingly shaping how people discover and understand organisations. Andrew Dawson, head of innovation at Edelman Intelligence, speaks to Rebecca Pardon about what this means for the future of reputation management. This article is from Communicate’s print edition. 

  • Corporate Affairs

People are increasingly using AI to decide what they think about a company. How much does that change the job of reputation management?

I think we’re at a really interesting moment. I’ve only ever done PR and communications, working in firms, at software companies and now back at a firm again, and I think about the change through a metaphor.

The way things worked in the 1990s and 2000s, with the advent of Google, was like a collective filing cabinet. You could say something like “best cheese shops” without articulating what you meant, because search engines couldn’t understand questions.

They would give you a filing cabinet of websites and it was your responsibility to form your own opinion. Now we have what I call “collective-conscious searching”. AI will ask what you mean by “best cheese shop”. More importantly, it’s no longer saying, “Here are eight or nine websites you should look at.” It’s saying, “Here are the three  shops you should go to, and here’s why they suit your specific tastes and budget.”

We’ve gone from social media, where brands, influencers and consumers could have a call and response, to people making decisions elsewhere. The machines sound wholly confident in the recommendations they’re making. That’s a frightening reality.

 

What makes it frightening?

There are people who are very aware of how machines work. But there’s also a group of people who take what they’re shown at face value. The machines are speaking to you because they want to please you.

They want to be helpful. What’s frightening is the machines’ impact on perception without understanding the consequences of their actions. They don’t understand: if I say this, someone might buy that, or they might feel this way.

We are turning over responsibility to a machine, or a set of machines, that doesn’t have the collective wisdom of lived experience.

 

Has the scale of misinformation and disinformation changed as a result?

Absolutely. I was working in the political sector during the last decade when the phrase ‘fake news’ became ubiquitous.

Back then, bad actors created misinformation, but they did it by hand. It was crude by today’s measure; a blunt instrument. What AI is able to do now is so much more, and that’s why safeguards are so important.

Ten or 15 years ago, to be truly malicious took a concerted effort. You might have needed a Photoshop expert and a whole series of people. Now any teenager with a cellphone can put together a fairly realistic image, or realistic-sounding audio, video or article, and send it around.

 

What does that mean in practice for the companies you work with?

I think it breaks into offence and defence. You have clients who say, “Can you help stem the tide? Can you help us understand what’s out there?” That’s playing offence.

It’s about understanding what’s happening and preparing yourself. People will ask, “Can you go out there and find all the videos that are fake?” But it’s got so good that even detection is difficult. And what if you miss one?

I like to think about a company’s profile in terms of true positives and negatives. What are all the narratives we know about you that we know to be true? Then look at everything else people are saying, confirm plausibility and isolate those issues.

Then there’s the defensive side. We have people who come to us because they hadn’t planned for a crisis and suddenly we have to sprint to get them up to speed.

 

Does AI also make it harder for companies to leave old crises in the past?

Yes. I think there was a world where we were allowed to forget and move on. AI makes it really hard to move on because it has this ability to go back. The same applies to corporate controversies.

An issue that happened ten years ago can suddenly become relevant again because the AI has retrieved it when searching for context to answer a specific question.

 

Can companies actually change what an AI system says about them?

I think it depends on the situation. In any crisis, the answer is “it depends”. I remember doing crisis work for  a consumer tech client in a previous role.

One involved washing machines that were combusting. The response was straightforward: put information about the recall out there and show the community what you’re doing to make things right.

The other involved a tweet about the brand. The question was: do we just inflame things by getting out into the world and making it worse? Ultimately, the answer was to let the storm pass.

There’s a perception that you can flip a switch and fix your reputation. But we’re still figuring out how these algorithms work.

What we have found is that the sources influencing LLM answers are different for every brand and sector. Comms teams need to start identifying new emerging channels or previously overlooked channels to pay closer attention to whether that’s Reddit, LinkedIn, short-form video or niche communities.

 

So how much confidence should communicators place in GEO?

I love the idea of GEO. Because LLMs tend to look for the most high authority sources when forming answers – news and earned media – optimizing for AI search means getting the right message published in the best places. Which is what good comms has always been about.  

What we need to be cautious about is treating LLM search like a crystal ball that we can manipulate. LLMs are potentially the best source of intelligence we have. I wouldn’t use that as a source of truth for what people are thinking.

I’d use it more as a collective pulse check on the world. The opportunity is not to treat the machine as an oracle, but to use the scale of what it can process to understand what is happening around you.