How to Write for LLMs: Lessons from a Content Marketer
Google’s latest shifts get a lot of attention, but what's been more interesting to me is how people are actually behaving around search.
LLMs (ChatGPT, Claude, Perplexity) have increasingly become the first stop when someone needs to understand something, especially early on. Not to “research” in the old sense, but to orient themselves. To sanity-check a question before they decide what to do next.
This changes the job of content in subtle ways.
It’s not that ranking no longer matters. SEO isn’t dead, people. It’s that ranking is no longer the “moment of impact.” More often, the moment that shapes perception happens earlier, inside a summary, a synthesized AI Overview answer, or a recommendation that never sends a click your way.
To stay visible, here's how your content needs to shift:
Keyword stuffing → Natural language
Traffic generation → Question answering
Ranking on Google → Being cited by AI
TL;DR It’s time to stop optimizing for clicks and start optimizing for answers. It’s what’s mattered all along: authority, originality, and human-first storytelling, and the same qualities that PR pros have been evangelizing. LLMs reward what humans already trust, so your SEO, content, and PR teams should be working together to earn visibility.
For example…
Traditional SEO:
A Google search for “best date night spots in Toronto” still returns a familiar list of articles, all optimized to win a position.
LLM Response:
Ask the same thing in an LLM, and you’re more likely to get a short, confident answer that sounds like a friend who already knows the city: “If you’re looking for romantic restaurants that blend ambience, cuisine, and memorable experiences, try Scaramouche Restaurant, Alo, Canoe, Osteria Giulia, or Restaurant 20 Victoria.”
✓ One delivers options you have to click through
✓ The other delivers the answer you were hoping to find
This distinction matters, because LLMs don’t behave like search engines. They don’t reward pages for matching strings of text or accumulating backlinks in isolation. They reward material that helps them answer a question clearly, credibly, and in the language people actually use when they’re thinking something through.
LLMs Don’t Think Like Search Engines
Unlike traditional search engines that rely heavily on keyword matching, backlink profiles, and domain authority, LLMs operate differently. They don’t return a list of ranked pages, instead they generate an answer based on semantic understanding, intent matching, and context.
Key Differences in How LLMs Process Content:
Semantic Understanding - LLMs care more about meaning than exact-match keywords. It’s like searching for “cozy Italian date night spot” versus “romantic pasta restaurant.” Different wording, same intent. LLMs understand that both are looking for the same kind of experience: candlelight, handmade pasta, and maybe a good bottle of red.
Intent Matching - LLMs prioritize usefulness. They aim to deliver the most relevant, accurate answer based on the question behind the question. The goal isn’t to show ten options, it’s to help someone feel confident in one or two. That’s a different bar than “ranking well.”
Contextual Response - Unlike search engines, LLMs can handle multi-step, conversational queries. Follow-up questions don’t reset the conversation the way they do in classic search. They narrow it. Refine it. That makes vague or over-generalized content hard to surface, even if it technically ranks.
Showing Up in AI Results ≠ Ranking on Google
One quiet misconception I still see is the assumption that if a page ranks on Google, it will naturally show up in LLM answers. Sometimes it does. Often it doesn’t. To earn visibility, your content must be structured to answer, not just rank.
For example: Perplexity uses its own web crawler (PerplexityBot) to index content. It references Google / Bing rankings as signals, but not as the final word. It uses native LLMs to generate summaries, not just serve links.
Understanding how LLMs retrieve, evaluate, and present information is the new edge in SEO. To show up in AI results, you need to think less like an algorithm, and more like a discerning customer.
I’ve found it useful to think about AI more like a curator than a crawler.
A good concierge doesn’t list every restaurant in the city. They scan the menu, understand what you actually mean by “best,” check whether a place has been talked about in the right circles, and then make a call. That’s closer to how LLMs operate than the old ten blue links ever were.
AI Curates, Not Just Crawls
There’s no exact way to track how AI tools rank responses (yet), but think of it like how a savvy diner picks a restaurant:
Step 1: Scan the Menu (Keywords)
You’re craving something good, so you search “best sushi restaurant Toronto.” AI tools, like diners, start by picking up on the main ideas — words like “best,” “sushi,” and “Toronto.”
Step 2: Understand What 'Best' Really Means (Semantic Analysis)
“Best” could mean highest rated, freshest fish, trendiest vibe, or fastest service. The AI doesn’t just take the word literally, it figures out what people really mean when they say it.
Step 3: Check the Vibes (Topical Relevance)
Now it cross-references: Is the place known for omakase? Has it been reviewed recently? Does it show up on local “top 10” lists? Are people raving about the uni? AI does the same, pulling in related signals like category fit, user engagement, awards, and reviews.
Step 4: Make a Recommendation (Relevance Scoring)
Finally, like a great concierge, the AI serves up what it thinks you’ll love, the content that best matches your intent, context, and appetite for credibility.
If you want your content to show up on the AI’s shortlist, don’t just serve keywords. You have to serve value, relevance, and clarity.
LinkedIn’s Favourite Debate: Is SEO Dead?
Which brings me to the LinkedIn debate that resurfaces every few months: is SEO dead?
Anyone who’s been doing this long enough knows the answer is no, but also not quite. Search traffic still matters, especially when people are comparing options, reading reviews, or doing deeper evaluation. If it’s driving real results for your business today, abandoning it because LLMs are trendy would be irresponsible.
What has changed is where SEO tops out on its own.
The overlap between SEO and LLM visibility is real. Clear structure, topical depth, trustworthy sources, and consistency still do a lot of the heavy lifting. The difference shows up in how the content is written and reinforced.
LLMs prefer writing that answers a question directly, in plain language, and then earns the right to go deeper. They surface material that sounds like it came from someone who actually understands the problem, not a page designed to satisfy an algorithm.
In practice, that’s pushed me to adjust less in what I cover and more in how I shape it.
If search traffic is a meaningful driver for your business today, don’t cut back on SEO just because LLMs are trending. Instead, focus on layering your strategy.
The good news? SEO and LLM optimization already overlap in meaningful ways. Both reward:
Clear, structured content
Topical depth and relevance
Authoritative, trustworthy sources
The difference is in the delivery. LLMs prefer direct, conversational, answer-first writing. You can keep your site structure, backlinks, and keywords, but make sure your copy is also easy for AI to parse.
This means:
✓ Clear headings
✓ Concise answers to real questions
✓ Consistent language around core topics
How I’m Adjusting in My Content Strategy for LLMs
LLMs may be complex, but they ultimately surface the same things humans value:
Clear, structured content
Helpful, natural sounding explanations
Accurate information and topical authority (you really know your domain)
Trustworthy sources
There are many best practices and tactics to choose from, but here’s how I’m adjusting my strategy…
Structure Content for Skimmability and Relevance
Okay, this isn’t new, but I’m doubling down on:
Clear headings and subheadings
Numbered lists, bullet points, and FAQs
Consistent use of H2s (for main topics) and H3s (for subtopics)
… anything that helps LLMs (and readers) instantly “get” what the piece is about. I also write headings as natural language questions to match how users ask, not just how they search. Longtail keywords still matter, but they need to reflect real human thinking, not robotic phrasing.
Integrate FAQs That Actually Answer the Question
FAQs are gold for both SEO and LLM optimization. I’m embedding them more intentionally by:
Using natural-language questions (e.g., “How is LLM optimization different from SEO?”)
Writing direct, two-to-three sentence answers
Leading with the answer, then adding context
Using structured markup (FAQ schema) to improve visibility
Linking to deeper resources if more detail is needed
LLMs love clear problem–solution structure, and, hey, so do your readers.
Write Like a Human Talking to Another Human
Ironically, the best way to show up in AI-generated answers? Don’t write for AI.
What excites me most is the return to narrative quality: human-centred clarity, originality, and relevance. Write for people. I’m simplifying technical explanations, ditching corporate jargon, and leaning into first-person language to make content feel like a real conversation.
If it sounds robotic, it’s forgettable. If it sounds human, it's findable.
Build Topical Authority, Not Just Volume
You can’t become a trusted source overnight. You earn it.
That’s why I’m prioritizing topic clusters, not random posts. I’m building content ecosystems that show depth and breadth, interlinked guides, expert interviews, playbooks, newsletters, and research-backed insights around core themes.
I’m also:
Quoting subject matter experts (our podcast helps with this)
Linking to authoritative sources
Tying content to trending topics and news
Sharing exclusive data to show real thought leadership
I’m repurposing relentlessly: newsletters become LinkedIn posts, LinkedIn posts become articles, articles become citations for LLMs to pick up.
The more high authority sources your brand appears in, the better your odds of being referenced in AI-generated responses. I’ll write more on this below.
Make Your Value Proposition Crystal Clear
LLMs don’t guess. If your messaging is vague or inconsistent, you won’t get surfaced.
So I’m making sure:
Our product’s purpose and audience are unmistakable
Messaging is aligned across blog posts, web pages, and social content
We consistently repeat key phrases around our brand and positioning
If you want to be known for “LLM optimization,” say it everywhere. That repetition helps LLMs associate your content with that niche.
Refresh, Don’t Just Create
I’m also spending more time refreshing existing pieces instead of publishing net-new ones for the sake of it. Some of the most valuable content on any site was written a year or two ago, before this shift really took hold.
That’s why I’m revisiting top performing pieces to:
Tighten structure
Refresh stats and examples
Update links, quotes, and citations
Add new sections where needed
It’s not glamorous, but it’s effective. It’s also essential if you want to stay credible.
One thing I think gets understated in all of this is how much brand presence outside your site matters.
One study found that LLMs are more likely to surface and cite content that includes stats, quotes, and external references — boosting visibility by up to 40% — because they signal authority, accuracy, and trustworthiness (the qualities AI models prioritize when choosing what to surface). As an added bonus, this approach also aligns with Google’s ranking factors, which reward brand authority and credibility in search.
What Still Works: Digital PR
LLMs don’t just look at what you publish. They look at where and how your brand shows up across the web. Media mentions, trade coverage, reviews, LinkedIn posts from credible people on your team, even the way customers talk about you in public forums all reinforce the same story.
If that story is consistent, specific, and grounded in real experience, it becomes easier for AI systems to associate your brand with a particular category or way of thinking. If it’s scattered or generic, the model has nothing solid to grab onto.
This is where digital PR and thought leadership still do real work, even if we’ve stopped calling them that. Not because they “drive traffic,” but because they create repetition and context in places LLMs actually learn from.
Foundational PR still matters:
Media features in relevant trade or mainstream publications
Press releases that speak to timely, credible milestones
High-authority backlinks from earned or paid placements
Thought leadership pieces published through respected outlets
Review sites and aggregators that mention your product in a trusted context
What’s Next: Creative PR Tactics
Some of the most effective LLM signals come from owned and earned presence on social channels, especially LinkedIn.
This is where your sales team’s personal brands come in:
Empower your sales team and execs to own thought leadership on LinkedIn
Share industry POVs, customer wins, and commentary that links back to your category
Repurpose internal insights into carousels, articles, and short posts
Create consistency between how your brand talks, and how your people talk
LLMs like ChatGPT and Perplexity index LinkedIn content. If your brand and product themes show up consistently across your team’s public-facing content, it strengthens the semantic association between your brand and your niche.
Why It Works
When LLMs crawl the web, they’re looking for:
Repetition of brand-topic pairings across domains
Quality backlinks and earned mentions
Trusted sources referencing your brand in context
Content that shows authority, recency, and topical consistency
The internet tells a story about your brand, whether you shape it or not. That story becomes fuel for LLMs. Mentions in high-authority publications, review sites, Reddit threads, or LinkedIn posts from your execs? AI notices. So do people.
Your owned channels—your website, your blog, your socials—need to reinforce that story. Thought leadership matters. So does consistency. The more you show up across trusted, credible sources, the more likely you are to show up in AI-generated answers.
So yes, get your media hits, but also build a team of credible voices.
Another TL;DR 😅 Optimize for usefulness, not clicks. Stop asking “Is SEO dead?” and start asking: “Is our content worth surfacing in the first place?”
The Format’s Changing. The Fundamentals Aren’t.
There’s something reassuring in all this, content that’s genuinely helpful still wins. Writing that respects how people think and decide still travels further. Structure that makes information easy to extract helps both humans and machines. When you optimize for usefulness instead of clicks, you tend to end up with something that works across more surfaces, not fewer.
The format is changing. The bar for clarity is higher. But the work itself feels familiar, especially if you’ve spent years close to customers, sales, and support, listening to how people actually talk when they’re trying to solve something. That’s still the signal worth paying attention to.
When you write for usefulness instead of just clicks, you’re not just future-proofing your content, you’re making it better for the people it’s meant to serve.