Multiple truths, one answer

By Christopher Cleves, Founder & CEO

The most consequential briefing you receive this year will come from someone you never met, never hired, and cannot ring.

Before an investor picks up the phone, before a regulator opens your filing, before a candidate decides whether to join, a growing number of them now put your name into a model and read what comes back. That answer was not written by anyone on your payroll. It was stitched together from whatever the model could reach – your site, old press coverage, Wikipedia, a subsidiary PDF, a five-year-old release – and then shaped by the model’s own view of the world.

The wrong question

A lot of the communications industry has spent 2026 debating whether AI counts as a stakeholder. Sessions have been devoted to it. Agencies have declared that it does. That is the wrong framing. A stakeholder has an interest in you. A model sits somewhere more powerful: between you and every stakeholder you actually have.

Almost everything an organisation publishes – earned coverage, its own website, its own channels – is now filtered, structured and re-presented by a model before most people ever see it. This is not a new channel and not a new audience. It is an intermediary, and one that already holds an opinion formed from its training data, its conversations with users, and whatever it can retrieve on any given day.

Multiple truths, one answer

On settled questions this matters less. On live ones – shifting policy, contested science, a restructuring, a dispute – it is decisive. There may be several legitimate readings of your position. Your stakeholder will be given one.

Answers do differ by model, by prompt, by who is asking, and by the week. That variation is only visible to the organisation that bothers to measure it. Each investor, journalist or policy adviser sees a single answer and, in ordinary use, has little reason to request a second one.

The defaults of the most widely used consumer products favour recency over settled understanding, simplicity over careful qualification, and a tendency to mirror the user’s own assumptions. None of those defaults is kind to the measured, caveated position your general counsel approved.

The intermediary ignores your org chart

Corporate governance draws clear lines: corporate affairs owns the narrative, legal owns the disclosures, digital owns the website. The model respects none of them. It pulls from subsidiary sites, forgotten PDFs, founder history, third-party commentary and old regulatory releases without caring who signed them off.

Two results follow. First, the last word on an issue is the word the model keeps repeating. If the most recent crawlable statement on a long-settled matter is an enforcement notice from five years ago, that becomes the story. Second, in most organisations no one owns the answer the models are giving. Ask who is responsible for a particular issue in the models and the most common reply is that no one had realised there needed to be an owner.

Silence is now a position

Authenticity used to be mainly about tone. At the speed the intermediary works, it is also about latency.

A significant message still has to pass through drafting, legal review, internal alignment and competing priorities. That takes days. The information vacuum is filled in hours, and it is filled by whatever is most dramatic, most shareable and most recent – the very material models are predisposed to surface.

Holding the line and waiting for a story to blow over was always a calculated risk. It is now a decision to let someone else write the answer your stakeholders receive. Proactive communication is no longer one option among several. It is the only posture that keeps an organisation inside its own story.

One announcement, several readings

No meaningful announcement lands with a single stakeholder group. Consider redundancies. Investors test cost discipline and the credibility of guidance. Media look for the human cost. Current employees test fairness; prospective employees test what kind of employer this is. The public judges decency; the regulator examines process. Each framing has to stand on its own, and none can contradict the others, because the intermediary will place them side by side.

Two things now matter more than before. The first is corroboration: consistent messages reinforced by independent third parties, because agreement across sources is what models appear to reward. The second is authority. Models show a clear preference for quotes from people they already regard as authoritative, and that perception is built from the existing media record. Authority cannot be manufactured in the week it is needed. It is a planning task, not a crisis task.

The supply problem

Underneath this sits a tension the industry has largely avoided. The intermediary leans heavily on earned media at the exact moment the economics of earned media are breaking.

Publishers face a poor choice. Give AI companies access, and their journalism shapes the answers while the traffic and advertising that once paid for it disappear. Withhold access and they vanish from the place where stakeholders now begin their research. Invisible, or visible and unpaid. For many newsrooms neither option sustains the fact-checked reporting the models still rely on for credibility.

The practical consequence for communicators is straightforward: the independent sources that corroborate your position are becoming scarcer and more valuable, and your relationships with them more important.

What it actually requires

The industry now has plenty of ways to measure this. Guides exist, tools track prompts, audits arrive every quarter or every six months. Measurement was never the difficult part.

The difficult part is acting on what the intelligence shows, inside the window that matters, across everyone who in some way stewards the company’s reputation: the general counsel, the corporate affairs lead, commercial leaders, the chief executive and whoever owns the website. It is half infrastructure – knowing what to say, to whom, in whose voice and through which channel – and half change management: getting those people to move together, quickly.

In practice, that means a short list of the highest-leverage changes each month, each with a clear owner, implemented that month and re-measured at scale, so the effect is demonstrated rather than claimed. The target keeps moving, so the discipline has to be continuous.


If you would like to understand what LLMs are saying about your organisation, or to start building the function to manage it, please get in touch at [email protected].

Related News & Insights

Multiple truths, one answer

Read more
curved architecture graphic

LLM Poisoning: Russia's New Playbook and the Risk to Your Reputation 

Read more
Analysis icon on the gradient background blue to orange diagonally

Trust is not a software problem

Read more

What Managing Reputation in AI Actually Looks Like in Practice

Read more

The first read on you is now an AI model

Read more

Reputation & LLMs: How These Models Are Trained on Your Content

Read more

Protect your reputation in AI-Search & LLM

Manage high-stakes issues and build a digital reputation that creates competitive advantage

See how
Copyright © 2026 Dablam Ltd (company number 15115628) | All Rights Reserved

14-15 Southampton Place, Holborn, London WC1A 2AJ
linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram