LLM monitoring is the practice of tracking how your brand is represented in the answers produced by large language models like Gemini, ChatGPT and Claude. It captures the citations behind those answers, the sources feeding them, and your share of voice against competitors being referenced in the same prompts.
For PR and comms teams, this is quickly becoming as fundamental as media monitoring itself. As AI platforms become a mainstream way people research brands, knowing how you show up inside them is a means to reputation management.
The communications industry has spent the past two years talking about AI. Most of that conversation has centred on content strategy and on keeping "AI slop" out of your own messaging. Fair enough. But there’s a second part of the equation that gets far less airtime. It’s how AI shapes what people learn about your brand in AI-generated answers.
Think about your next customer, investor or journalist. There's a good chance they're asking an LLM a question before they ever pick up the phone or visit your website. Whatever that LLM tells them becomes their starting point. And right now, most PR teams have no free way to see what that starting point even is.
Traditional media monitoring tracks earned coverage: broadcast, print, online, social. It was built for that landscape, and Medianet has spent years refining how it works for clients. But AI search is a different beast. It aggregates coverage, synthesises it, and decides for itself what's worth repeating.
That's why we built LLM Monitoring in Medianet Labs. Rather than wait until a tool is fully polished, we'd rather build it alongside our clients and take feedback as we go. Labs gives clients exclusive early access to a growing suite of AI-forward tools designed specifically for comms and PR professionals, covering everything from brand visibility in LLMs to press release optimisation. It lives inside the same platform you already use for media monitoring and outreach, so there's no separate login and nothing extra to set up.
One of the first tools inside Medianet Labs is LLM Monitoring.
Earlier this year, Medianet ran Australia's first pilot study into how brands are represented inside large language models. The question was simple: when someone asks an LLM about a brand, where does that answer actually come from, and can PR teams trust what's being said?
Three findings stood out.
AI prefers owned content over earned media. In the Australian context, LLMs draw far more heavily on owned content such as company websites and press releases than they do on earned media coverage. This may vary depending on the prompt, the industry and the type of content required to inform an answer, but given the various restrictions many influential media houses impose on LLMs in Australia, it is your own content that can do the heavy lifting when informing the public. If your reputation strategy leans on third-party validation, that can still signal to LLMs what they crave the most: trust.
LLMs have a long memory. An issue from three or four years ago can surface with the same weight as something published last week. A human reader applies context and recency instinctively. An LLM doesn't. If it's indexed, it can resurface, regardless of how long ago it happened or whether it's still relevant.
None of this is visible through traditional monitoring. You can have a flawless media monitoring setup, catching every broadcast mention and every online article the second it publishes, and still have zero visibility into what AI platforms are telling people about your brand right now. That blind spot is being addressed by LLM Monitoring.
LLM Monitoring lets you track the kinds of questions your customers, journalists or stakeholders are likely to ask an LLM about your brand. It shows you the answer the AI gives, and just as importantly, it shows you the share of voice and sentiment compared with your competitors. That second part is the one most people don't expect.
Being able to see exactly how mentions are shaping what an AI says about your brand means you finally have something concrete to act on:
The dashboard gives you data points including share of voice, brand footprint, and total brand mentions, so you are seeing the problem and you're seeing the scale of it.
The tool itself is straightforward to use. This is not and will not be a new discipline you need to learn from the ground up. It's the same instinct that's always driven good reputation management: know where your brand is showing up, know what's being said, and know it before anyone has to ask you about it. The only difference is that it's now applied to a channel that simply didn't exist for PR teams until recently.
It's already shaping perception today, quietly, in the background of every ChatGPT or Gemini query someone runs before they contact your business or inform a decision. Comms teams who wait for this to become "standard practice" before paying attention will be playing catch-up on a narrative that's already been set. The teams who start monitoring now have the advantage of catching inaccuracies, outdated coverage, and competitor gains while they're still fixable.
This is also where generative engine optimisation (GEO) comes in: the practice of making sure your owned content is structured in a way LLMs can actually find, understand and cite. Since the Phase 2 research shows AI leans so heavily on owned content, that content needs to earn its place in an AI-generated answer. Clear, well-structured, fact-forward content on your own site has a real shot at being the source an LLM pulls from. Vague, jargon-heavy content usually doesn't.
In practice, this means treating your website and press releases less like brochures and more like reference material an AI can lift from directly. Clear headings, direct answers to likely questions, and specific, verifiable facts all give an LLM something concrete to cite. That's a different discipline to writing for a human reader scrolling a page, but it's one PR teams are well placed to pick up, because it's really just an extension of what good press releases have always needed: clarity, structure and credibility.
LLM Monitoring is one tool inside Labs, but it's not the only one.
PR Optimiser: Trained on decades of Medianet's proprietary data, this tool checks your press releases against the specific factors that drive journalist pickup, including structure, clarity and source credibility, before you hit send.
Medianet Arcade: A space for a bit of downtime as keeping across earned media, owned content and a brand-new AI channel is demanding work!
Coming soon: Purpose-built tools like Hansard Monitoring and the Media Contacts Agent are already in development and launching very soon.
Medianet Labs is live now, free to every Medianet client on every plan, from day one, at no extra cost.
Tell us what's useful. Tell us what's missing. Tell us what you wish these tools could do. Labs is as much a conversation as it is a product, and the people best placed to shape it are the PR and Comms professionals using it every day.