There’s no shortage of advice right now about writing for AI. New frameworks, new tools, new acronyms arriving every quarter. And it’s easy to walk away thinking you need to completely rethink how you create content.
That’s not entirely true, but you do need to understand what’s actually changed.
AI-powered search doesn’t reward optimization tricks. It surfaces content that’s clear, specific, and genuinely useful to the person reading it. The content that performs best for real humans is exactly the content these tools are most likely to understand, use, and surface. That’s not a coincidence. It’s how they were built.
Why writing for people still wins
The bigger the organization, the more content gets written for internal audiences before it ever reaches a real one. Legal reviews it. Compliance approves it. Three stakeholders add their priorities to the brief. By the time it publishes, it’s been softened into something that satisfies everyone in the room and nobody outside it. Content written to offend no one turns out to be remarkably easy to ignore.
From SEO to GEO
For the better part of two decades, search optimization meant one thing: ranking. Get to page one, get the click, get the visit. The signals that mattered were largely technical: backlinks, keywords, page speed, domain authority.
Generative engine optimization is a different game. When a search tool synthesizes an answer rather than returning a list of links, ranking becomes secondary to citation. A patient searching for information about an atrial fibrillation diagnosis doesn’t get ten blue links anymore. They get an answer, drawn from sources a language model decided were specific, accurate, and useful enough to trust. Your page either contributed to that answer or it didn’t.
That shift has consequences for every enterprise organization managing content at scale. The question is no longer whether your pages rank. It’s whether they’re the kind of source a language model would trust enough to cite.
What that looks like in practice
The cancer care page is a useful place to make this concrete. Imagine a page for Saskatchewan Health Authority’s oncology program that reads like this.
Before
“Compassionate Cancer Care, Close to Home
Our oncology program is built on a foundation of compassionate, patient-centered care. Our multidisciplinary team of specialists works collaboratively to develop personalized treatment plans that address the whole person, not just the diagnosis. We are committed to delivering world-class cancer care with the warmth and support you deserve, every step of the way.”
There’s nothing wrong with the tone. It’s warm, reassuring, and exactly what a committee of stakeholders would approve. But read it again and notice what it doesn’t tell you. What cancers does this program treat? Where is it located? How do you get referred? What does the first appointment look like? The language gestures at care without ever describing it. For a patient trying to make a decision under enormous pressure, that’s a dead end. For a language model trying to construct a useful answer, there’s nothing here worth citing.
After
“Cancer Care at Saskatchewan Health Authority
Saskatchewan Health Authority’s oncology program treats breast, lung, colorectal, and blood cancers at locations across the province, including Saskatoon and Regina, with same-week appointments available for new patients.
What to expect:
- Initial consultation within 5 business days of referral
- Dedicated care coordinator assigned at first appointment
- Treatment plans developed by a multidisciplinary team within 2 weeks of diagnosis
- Services include surgery, medical oncology, radiation, and clinical trials
Who we treat:
- Adults newly diagnosed with cancer
- Patients seeking a second opinion
- People managing ongoing or recurring cancer
How to get started:
- Ask your primary care physician for a referral
- Call directly to self-refer
- Use the online appointment request form at saskhealthauthority.ca”
Every line is earning its place. The headline tells you what it is. The locations and timeline answer the questions that would otherwise send someone back to Google. The structure removes the barrier between a frightened person and the information they need to take a next step.
That last point matters more than most teams realize. Your content won’t always be read as a full page. It might surface as a snippet, a pulled quote, or part of a generated answer. When that happens, vague language disappears entirely. Specific, useful details are what carry through.
The information is probably already there
Most enterprise teams that come to us with a content problem think they need more content. What they usually need is better structure. The information is often already there. It’s buried in paragraphs written for annual reports, or wrapped in language that made sense to the internal team that produced it but means nothing to someone encountering the organization for the first time. Restructuring that content, pulling out what it is, who it’s for, what’s included, what to expect, doesn’t just improve readability. It changes whether the content can be understood at all.
Start with the questions you’re already being asked
If you’re not sure where to begin, the shortest path is through your own organization.
What do patients, customers, or prospects ask most often? What questions show up in intake calls, sales conversations, and support tickets? Those questions are a direct inventory of what your content is probably failing to answer. They’re also exactly what people are now typing into search, and exactly what AI tools are being asked to resolve.
When your content answers those questions clearly and specifically, it doesn’t just perform better in search. It does the job it was always supposed to do.
Where to start
You don’t need to rebuild everything at once.
Pick one page that represents a core service or capability, something important enough that getting it wrong has real consequences. Replace the brand language with specific information. Break it into clear sections. Answer the questions a real person would actually ask. Check that every claim is accurate and current.
One well-structured page built for a real reader will consistently outperform ten pages written for an internal audience. Start there.
A final thought
Search will keep evolving. The tools will change. The acronyms will multiply. What won’t change is this: a real person is always on the other end of the query. Someone trying to understand a diagnosis, evaluate a vendor, or make a decision that matters to them. The organizations whose content serves that person clearly and specifically will be the ones that get cited, trusted, and chosen.
