There's a dangerous temptation when people talk about "writing for AI": imagining you have to write for the machine, in some strange optimized dialect. It's the fastest route to producing bad content that, ironically, AI won't cite either. The truth is simpler and harder: the content LLMs like to cite is, almost always, content humans would like to read — just with a few structural touches that make life easier for whatever retrieves the information.
Why Clarity Is the Real Optimization
When an LLM retrieves content, it has to extract a reliable answer from a text. The more ambiguous, vague, or rambling that text is, the harder it is to extract something usable — and the more likely the model prefers a cleaner source. Clarity isn't a courtesy to the reader; it's what makes your content usable by the AI.
That changes how you think about writing. It's not about packing the text with technical terms to "look expert." It's about articulating ideas so that both a reader and a model understand, effortlessly, what you're claiming.
Actually Answer the Question, Early
Content AI cites usually has a simple trait: it answers the question it promises to answer, and answers it early. If the headline says "how to choose a CRM," the first paragraphs already deliver concrete criteria — not three blocks of intro about the importance of CRMs in the modern world.
AI retrieves passages. If the useful answer is buried in the eighth paragraph, after a lot of throat-clearing, its chances of being plucked drop. Put the substance where it can be found.
Structure for Extraction, Not Decoration
Subheads that describe what comes next. Paragraphs that handle one idea each. Sentences that assert things rather than just hint. When it makes sense, lists — but real lists, with substantive items, not empty bullets.
This structure isn't aesthetic. It creates "handholds" for retrieval. A model can safely isolate your answer to "what are the criteria for X" when it sits in a clear block, announced by a subhead, far better than when it's diluted inside a paragraph about something else.
Be Specific Where Others Are Generic
AI tends to prefer the source that says something concrete over the one that says something that could apply to any company. "Our product is innovative and customer-centric" gives the AI nothing to cite. "Our product cuts contract-closing time by automating step X" gives it something.
Specificity is both better writing and a better signal for AI. Concrete names, explained mechanisms, clear distinctions: all of it raises "citability" because it gives the model substantive material to use.
Demonstrate Real Expertise, Not Words About Expertise
Models are surprisingly good at telling content that demonstrates knowledge from content that merely talks about knowing. A text that anticipates the hard questions, acknowledges nuance, admits trade-offs, and takes reasoned positions signals authority in a way no adjective can.
In practice: instead of saying "we're experts in X," show the kind of reasoning only an expert in X would have. AI retrieves the substance, and the substance is what builds your reputation in the semantic space.
Keep Content Alive
For models that search live, recency counts. Content that's updated, revised, kept relevant has an edge over content published and abandoned. That doesn't mean rewriting everything weekly; it means treating your reference pages as assets that deserve maintenance, not posts that age on a shelf.
What Not to Do
Don't write to fool the model. Keyword stuffing, repeating terms artificially, creating lists just to "look structured," inflating the text with jargon: none of it fools today's models, and all of it worsens the human reader's experience. You end up paying the price twice.
Don't confuse volume with presence. Ten generic posts on the same topic, with the same words, contribute less than one genuinely good, specific piece. AI weighs by quality and relevance, not quantity.
The Simple Test
Before publishing, ask one question: if an expert read this, would they learn something or recognize the competence behind it? If yes, you've probably written something AI also values. If no, no technical optimization will save it.
In the end, writing to be cited by LLMs is writing well, with a little more attention to structure. And knowing whether it's working means looking at the real outcome: which of your pieces AI is actually citing, on which questions, and what the winners have in common. That feedback — which of your content becomes a citation — is what closes the loop between writing and being found, and it's what Genoma helps you see.