If Every Stakeholder is Coming to AI First, Then AI is Your First Stakeholder.

Straight to the Point

The media environment has changed, again. We went from a small handful of named gatekeepers, to a network era driven by the search engine algorithm, to a new intermediary: frontier AI models that answer hundreds of millions of questions a day without users ever clicking through to a source.
Product marketers moved first; a vendor category called Generative Engine Optimization has formed in eighteen months, and CMO budgets are following it. Corporate communicators have been slower to recognize that the stakes for reputation are higher than the stakes for product visibility.
Every Chief Communications Officer should now be running a recurring AI-perception audit of their company — across models, across stakeholder audiences, over time, against competitors, traced to source. The audit is already being run. The only question is whether you are reading it.

What We See

At this spring’s Google I/O, Sundar Pichai introduced what he called “the biggest upgrade to Search box in over 25 years.” The familiar list of blue links is being replaced by an intelligent box that dispatches background agents to monitor topics. AI Mode now serves a billion monthly users. AI Overviews serve two and a half billion.

Three months earlier, Condé Nast CEO Roger Lynch had told the Financial Times he had instructed his teams to “plan as if search traffic will be zero.” Google’s announcement read like confirmation.

Step back and the arc becomes clear. Mass reputation used to be managed inside a small room — a few network anchors, a handful of newspaper editors, the senior staff at three weekly magazines — people you could call by their first names. Then came the network era: the golden age of blogging, peak Twitter, the podcast for every preoccupation, and now the algorithmic short-form video. Anyone with conviction and a smartphone could find an audience. Even so, an algorithm sat in the middle. SEO was the first time CCOs and CMOs had to write for a machine.

What’s emerging now is different from both. AI is a new gatekeeper, at a scale the network era never reached. ChatGPT has roughly 900 million weekly active users. Gemini has 750 million monthly. And Pew Research found that when a Google AI Overview appears on a results page, users click a link to a source 8% of the time, against 15% when no AI summary is present. The answer is arriving before the click ever happens.

What It Means

Product marketers have been the early movers. Generative Engine Optimization — GEO, sometimes AEO for Answer Engine Optimization — now has a name, dozens of venture-backed startups scrambling for share, and a budget line. The CMO question is recognizable. The CCO question is bigger and more dangerous.

The audiences that hold a company’s license to operate — regulators, congressional staff, institutional investors, journalists, advocacy researchers, plaintiffs’ attorneys, prospective employees — use these tools too. Often as the first step. They ask the questions a search bar was never asked, because it could not answer them: Is this company a responsible operator? What’s the controversy with the new product? Are their safety problems trending up or down? The AI answers. And the user moves on.

We have spent the last few months measuring this issue. Two findings have already changed how we think about the work.

Across models, the picture is more stable than you’d expect — under two points of spread on a ten-point scale across the leading frontier vendors. Across users, it fragments. Same company, same model, eight stakeholder personas — and the answer moved by two to three points end to end. The Republican Senate staffer scored the company higher than the Democratic Senate staffer in every one of our eighteen audits — by a point and a half on average; over two points at energy companies. The environmental advocate and the sell-side analyst heard meaningfully different companies. AI is a mirror that understands its user’s interests and occupation. There is no single result. There is a constellation.

The second finding will embarrass most communications teams. The company’s own channels accounted for nearly half of everything the AI said — but the breakdown was telling. SEC filings, earnings transcripts and sustainability reports did most of the lifting. The corporate newsroom rarely cracks five percent. Executive op-eds barely register — under one-tenth of a percent across every client we have measured. The carefully turned tagline is often invisible to the models. The loose phrase from last quarter’s earnings Q&A may be the doctrine. And the unglamorous work of keeping a Wikipedia page accurate — work Purple has done with clients for years — has become one of the highest-leverage things a communications function does.

Across all of it, the pattern is clear: AI reads for data and structured facts, not corporate taglines. The new test for every piece of owned content is not just whether it is on message but whether it has the data groundedness that AI can utilize.

One more reality sits underneath both. The frontier models most people use by default — without web search — answer from a training-data picture twelve to eighteen months behind. Months into Trump’s second term, ChatGPT was still telling users Joe Biden was president. If a model can be off by a sitting president, it can be off by your CEO transition, your reset language. The AI knows the long version of your company’s story far better than your latest — making the multi-year narrative more important than the campaign you launched last spring, not less.

What To Do

Run the audit each quarter. An honest one has five components.

Across models. ChatGPT, Claude, Gemini. Where do they agree, where do they diverge?
Across stakeholder audiences. A generic reputation score is the wrong artifact. Map the constellation: what the regulator hears differs from what the Republican staffer hears, from what the institutional investor hears. The differences are where the strategic work lives.
Over time. AI portrayals are not static. They move as models retrain and as web-search retrieval flows into more responses. Track the trend, not the snapshot.
Against competitors. You will be measured against peers whether you commission the work or not. Better to know which battles you are winning.
Traced to source. Which of your content is propagating? Which of your best turns of phrase are doing no work at all? Tyler Cowen calls the AIs “your best informed reader.” For most communicators that still reads as a provocation. It shouldn’t. Who actually reads the 80-page sustainability report cover to cover? Claude does. The 10-K? Gemini does.

Every CCO is already being audited by AI, on behalf of every stakeholder who can type a question. The audit is comparative, recurring, and nearly free for the asker. The only question is whether you are reading it before they do.