Matrix Chambers, Digitalis and DRD Partnership came together for their second event on AI and reputation, Protecting Reputation in an Era of Accelerated Harm. Chaired by Jon McLeod, the panel brought together three perspectives that clients increasingly need in the same room: Sara Mansoori KC on the legal remedies, Fred Duff Gordon on the digital landscape, and Claire Davidson on the communications strategy that ties them together. Across four core themes (authority, persistence, silence and intervention), they examined how the new information environment is changing the way advisers protect their clients.

1. Clients no longer have one reputation, they have several

AI has changed not just where people find information, but how much they trust it. People now act on AI-generated answers more readily than on traditional media, clicking, buying and resharing on the strength of a summary they have not examined the sourcing behind. And because those summaries vary by model, by prompt and by day, a client’s reputation is no longer a single thing that can be monitored and managed. It is several, and these multiple reputations exist all at once.

The format of AI answers makes this even more difficult. AI presents its findings in calm, confident prose that reads like a senior briefing note, flattening the credibility difference between a Glassdoor review and a piece in the Financial Times, and rarely showing its sources. Understanding what the models draw on is therefore the starting point for any strategy. Digitalis’ research points to earned media from credible third parties as the largest source, with Wikipedia the single most cited site. Sourcing trends are constantly changing, so a clear, agile understanding of the digital environment and where information about clients is coming from is essential to any reputation management strategy.

2. AI has not reinvented the rulebook, it has amplified them

For all the novelty, the panel’s most reassuring message was that many of the fundamentals still hold. The search index remains the engine behind generative AI, and most of what the models cite comes from the first two pages. What has changed is the scale. Material that once sat harmlessly on page 20 of a search can now be pulled into an AI answer and presented with the same authority as the front page. Controlling what sits at the top of the index, by displacing or mitigating the adverse narrative while strengthening the client’s own, is still where the most effective work gets done.

3. Correction is not closure

Perhaps the starkest point of the evening was that fixing a source does not fix the answer. LLMs predict responses from a vast body of text, so a surgical correction to a single article may barely register until the right text has changed in all the right places. A client can win at trial and find that nothing has changed in what the models say about them, and a bad actor needs only one prompt to revive an old story.

That asymmetry means clients need to plan for a marathon, not a sprint. The outcome itself also needs to be built to travel:

• A favourable judgment remains one of the most valuable assets a client can secure, as an authoritative statement the models will keep picking up.
• Agreed statements and statements in open court can work in the same way, provided the wording is right and they land on the right channels.
• Advisers need to keep track of which sources AI tools can actually use, particularly as publishers gain new rights to opt out of AI features.

4. Silence is no longer golden, but speaking too soon is just as dangerous

“Declined to comment” used to be a respected line. Now, it is an open door: AI will scrape up whatever it can find to fill the gap, because its aim is always to give the user an answer. There is a legal cost too, since an inaccurate statement left unchallenged can put a libel claim out of time while the allegation spreads beyond anyone’s control.

The answer is not a public fight. It is making sure that silence is never silence without a denial on the record, whether through a pre-action letter or carefully placed wording online. The panel were equally clear about the opposite risk. Orchestrated campaigns are now industrialised, operating at a scale and speed that would have been unthinkable a few years ago, and responding publicly before speaking to legal advisers can make a story live. Much of the most effective work now happens below the radar, ideally before publication.

5. Evidence is everything

With the courts of England and Wales yet to test many of these questions, the panel looked to international litigation for pointers. The practical lessons are clear:

• Screenshot and time-stamp harmful AI outputs as they appear, since they may not be reproducible later.
• Put providers and platforms on notice when information is false, and build a record of doing so.
• Document the loss: if a deal fell through because of information shared in an AI summary, that link needs to be provable.
• Look for breaches of platform terms, such as fake accounts or coordinated activity, which can offer a faster route to removal than a claim.

6. Advisers can no longer work in silos

The thread running through the whole discussion was that no single discipline can solve these problems alone. Many clients find themselves exposed because they have, in effect, outsourced their reputation to social media and third parties. Taking back control means identifying where the real risk lies, whether with regulators, banks or customers, and pairing tailored, direct engagement with a steady flow of owned content. That only works when legal, digital and communications advisers sit down together from the outset and work to one controlled message.