Why you shouldn’t trust AI-generated news stories

Spoiler alert: the article you’re about to read has been written by a person.

Not just any person, but a highly experienced writer with many years of experience distinguishing fact from fiction and fake news from real reportage.

As such, you can be confident that the sentences below are an honest attempt at representing the topic under discussion.

It’s become necessary to make these statements because the rise of AI has led to an explosion in algorithm-generated news stories which haven’t been written by a person.

In some cases, nobody has cast a critical eye over the algorithmic analysis of existing online information, vetting it for accuracy and identifying any obvious discrepancies or errors.

As a consequence, news is no longer as black-and-white as it used to be…

Black and white and cred all over

AI has the potential to positively influence many aspects of our lives, but its credibility as a news reporting tool has been seriously damaged in recent months.

The AI algorithms which scan existing news stories and then attempt to repackage them without breaching copyright have clearly been unleashed on the world with insufficient testing.

In short, they keep generating fake news.

In December, Apple published AI-generated news stories claiming the man responsible for killing American healthcare boss Brian Thompson had himself been shot.

Not only was this factually wrong, but Apple attributed reporting of the ‘event’ to the BBC, who promptly demanded the immediate deletion of this demonstrably fake news story.

Dander up, the BBC began looking more closely into fake AI news.

Last week, they published a scathing critique of four leading AI chatbots including OpenAI’s market-leading ChatGPT and Google’s rival Gemini.

The BBC fed these generative AI platforms new and factually accurate stories and then asked them a series of questions.

Over half of the responses had what the BBC describe as “significant issues”, while almost a fifth contained factual errors including incorrect numbers and dates.

Since accurate source information was being supplied, it’s worrying that the output was erroneous.

We’re not talking about subjective analysis here, such as differences in perspective. We’re talking about AI articles stating that Rishi Sunak and Nicola Sturgeon are still in office.

Experts were misquoted, facts were reversed (the NHS was quoted as recommending the opposite of its official policy), context was lost, and opinions were misrepresented as facts.

You can read the BBC’s analysis here.

Why does this matter?

It matters because trust in what we’re told by other people is at record low levels already.

In a post-truth age where some people believe their personal opinions outweigh known facts, it is imperative that media platforms report things as accurately as possible.

When AI-generated news stories get something wrong, it further diminishes consumer confidence – especially when errors are then falsely attributed to a legitimate media outlet.

Even if AI content is accurate, industries like journalism won’t survive if generative AI endlessly harvests published content, rewording it slightly and distributing it for free.

Almost everything you read online has to be paid for somehow. Content production would cease to exist without online advertising, affiliate links, paywalls, sponsorship or sales.

The tech giants who have basically stolen every piece of content ever published on the internet without acknowledging it (let alone paying for it) don’t seem to care about this.

You should care, though.

Because if you rely on generative AI search results rather than visiting source websites, or read AI articles instead of visiting news media sites, that vital source material will dry up.

All that’ll be left is a series of AI algorithms endlessly regurgitating the same content with diminishing returns – and dwindling accuracy, if recent performance is any benchmark.

There has never been data appropriation like this in human history – we are in the midst of an unprecedented technological experiment.

In the meantime, we’d recommend reading the news on established platforms, rather than relying on AI-generated summaries piped to your mobile device or web browser.

If there isn’t an unambiguous link to a reputable website or news outlet, be very cautious…