AI never forgets: your past PR crisis is now permanent training data that's hurting your brand reputation

By Mike Cook, Head of Search, AI and Emerging Platforms, News Corp Australia

"The internet never forgets." It's the cautionary idiom that many of us were forced to learn the hard way. And any brand with an established digital footprint is well versed in the stress of righting a digital faux pas.

This adage still holds merit - and arguably more weight - in the age of AI. Growth in LLM adoption brought with it a resurgence of traditional marketing triggers, including the impact of digital word-of-mouth. This time it's in the form of AI influence.

So what happens when AI influence leads to poor brand representation in LLMs?

News Australia research reveals 37% of LLM influence in the retail sector comes from external press and community forums. This extends beyond recent threads or articles from the latest news cycle. It's material used to train AI models. It's historical content that LLMs are still able to crawl, digest and regurgitate in whatever context is demanded from a user's prompt.

It's more unforgiving than the 24-hour news cycle. When a brand's PR crisis is picked up by the wider public or exposed in the media, it has real implications for long-term visibility across LLMs.

The truth that all digital marketers must become familiar with is LLMs remember.

A marketer's ambition to remedy a PR crisis unfortunately now includes the arduous task of rewriting AI knowledge maps.

It might appear that the answer is simple: flood the web, using all the age-old digital PR tricks to push fluff and drown out negative press. But LLMs don't just read page one. They synthesise historical data based on the volume, weight and consistency of the information.

The truth is that LLMs care about how your brand is represented holistically. It doesn't matter that your potential customer may never hear about an ill-received marketing campaign or scroll far enough to read a zero-star review—because a crawler will. In the age of AI, the power of brand reputation is transferred from the brand directly to the LLM.

This is not to say your brand needs to have a squeaky-clean digital footprint to guarantee favour within the LLM. In fact, data suggests a brand with a flawless digital history is suspicious to LLMs.

Not only do LLMs have a preference for 'middle-ground' reviews, but so do real shoppers. Research from Northwestern University found that purchase likelihood actually peaks when a brand or product sits between 4.2 and 4.5 stars. Furthermore, 82% of shoppers specifically seek out negative reviews to establish trust. When a footprint looks completely flawless, both consumers and AI models assume it's manufactured.

To rehabilitate after a PR fumble, brands must walk the tightrope between perfection and digital infamy. For LLMs specifically, the path to redemption starts with alignment between owned, earned and paid ecosystems.

LLM responses are built off probability-based logic; brands have to work to tip the odds back in their favour by providing a steady, authentic stream of balanced, high-quality data that spans owned channels, verifiable community consensus and expert commentary.

In a nutshell, a brand's digital footprint has to re-teach an LLM about how the brand should be perceived. The formula is simple in theory but requires continued vigilance: unite brand messaging, real consumer opinion and authoritative grounding, then rinse and repeat.

As etched in another idiom: "there are no shortcuts to any place worth going." This is true for LLMs. No, AI may not forget the blunder... but as is the axiom of machine learning, you can teach it to relearn the correct brand message.