Last week, I had to dig deep into a new client’s background — fast. They were in the middle of a substantial PR crisis, and time was not on our side.

Previously, my method involved resorting to Google, meticulously browsing through numerous pages of results. I would examine their website, news articles, social media presence, reviews, and even delve into obscure forum discussions. My objective was consistent: to gain a comprehensive understanding of their identity, operations, and any available public information that could influence their reputation positively or negatively.

Doing that kind of research the old way can take hours.

Today, AI has revolutionized this process, making it much more efficient. Tools such as ChatGPT, Claude, and Grok can rapidly distill public information, providing me with a quick overview in moments rather than hours. However, this expedited approach comes with a significant caution: we must also account for what these systems are articulating about individuals and organizations, as well as the methodologies they utilize to reach these conclusions.

Large language models (LLMs), which are the technology behind these AI tools, are trained using extensive datasets sourced from the vast online landscape. Consequently, your brand’s presence on the internet is no longer merely observed by individuals; it is now being analyzed and summarized by machines as well.

This changes the game for public relations.

Because while LLMs can be incredibly powerful, they’re still prone to hallucinations — a polite term for making things up. And if you’ve spent even five minutes with Google’s new AI Overviews (AIOs), you’ve seen it firsthand.

Some examples I’ve personally encountered in AIOs:

• That Gouda is the best-selling cheese in the U.S.
• That you should add non-toxic glue to pizza to keep cheese from sliding off
• That drinking urine is an effective treatment for kidney stones

Ridiculous? Absolutely. But it underscores a bigger issue: these systems can spread false or misleading information quickly and at scale.

Even with less extreme topics, hallucinations happen. I once asked Grok to summarize my background. It confidently told me I’d served in the Army Airborne. In reality? I was a Marine.

As more people rely on AI to answer questions they once typed into search engines, the accuracy and relevance of your brand’s presence in these models is becoming critical. Not only do you want to make sure the information is correct, but you also want your brand to show up at all. Ideally, you want to appear in answers about your industry, not just yourself.

So, how do you influence what these systems say? Unfortunately, it’s not as easy as feeding them your preferred narrative. If it were, AI tools would already be flooded by spam from low-quality marketers.

Instead, LLMs prioritize information from trusted sources across the web, and not all sources are weighted equally. Your company’s official website helps, but third-party credibility matters far more.

That’s why editorial media coverage remains the most powerful tool in modern PR — and it matters now more than ever. There are two core elements here: high-quality editorial features and press releases.

Editorial features — stories published by reputable media outlets that quote you or spotlight your work — carry the most weight. Why? Because they’re difficult to manipulate. Getting published requires a compelling topic, a unique perspective and often, relationships with journalists. You have to earn it. That’s exactly why LLMs treat this kind of coverage as a strong trust signal.

The more insight you share in those features, the better. If you’re quoted briefly, it suggests your voice is just one of many. But if your expertise shapes the bulk of the story, that sends a much stronger signal — both to readers and to the algorithms parsing it.

That’s also why it’s smart to pursue interviews and contributor content in addition to being quoted. These allow you to go deeper, share your thinking more fully and increase the likelihood that your perspective makes it into an AI summary.

Press releases still matter, too — but in a more limited way. They’re a paid channel, so anyone can publish them, but there’s still some editorial oversight. Editors at distribution services do basic fact-checking and screen for hyperbole before syndicating them to media outlets. The key is to make sure your press release is actually newsworthy. A strong release can also prompt journalists to cover your story further.

While LLMs pull data from various formats — text, audio, video — text-based articles still produce the fastest and most reliable impact when it comes to influencing AI responses.

In many ways, AI has transformed PR. But the fundamentals haven’t changed. You still need to earn high-quality media coverage. The difference is that now, those features are no longer just about reaching human audiences — they’re about training the machines that shape perception at scale.

The companies and individuals who recognize this shift and act on it now will gain a long-term advantage. Those who don’t? They’ll get left out of the conversation — by people and by AI alike.