- Ranking in LLMs depends a lot on traditional SEO and how search engines break down and fetch queries in real time.
- Schema and other technical elements matter much less than many people think; your core content and topical authority still matter most.
- LLMs use fanout queries and “drift” to reword prompts and fetch related results, understanding this helps your pages get cited more often.
- Tracking LLM-driven traffic requires rethinking reporting, look for referral traffic, not just organic clicks.
You want to show up in LLM-powered answers, like those in ChatGPT or Perplexity. The truth is, if you rank in Google for informational and commercial queries, you’re already in the running to appear. These LLMs break each user prompt into 1 or more search queries, fetch real Google results in near real time, and construct an answer citing those sources. So, your job is (perhaps boringly) still about ranking in Google. Schema is usually not needed. Instead, focus on broad, useful coverage of your niche, keep your content up to date and relevant, and track how your pages perform as sources for answers, not just your organic clicks.
What Actually Gets Content Cited by LLMs?
Let’s start simple. Many SEOs overcomplicate things by treating LLMs as something magical. They aren’t. For most queries, ChatGPT, Perplexity, Claude, and others break your longer prompt down into typical short-form Google-style queries.
Then, they scrape the live top search results. You don’t need some secret technical hack to get your site mentioned. If you earn top spots for those split-out queries, your pages become the “source” in AI answers.
LLMs are not storing the internet. They run searches, then synthesize responses using current rankings.
This means:
- If you rank page one for “best small business CRM” and someone asks “What CRM should I use as a 200-person marketing team?” in Perplexity, your post is eligible for a citation.
- Pages can show in LLMs within minutes of being indexed by Google. Real-time crawling is the norm.
Don’t Waste Time Chasing Schema for LLMs
There’s a belief that marking up your content with schema guarantees citations in LLM answers. It’s just not true for 99% of sites. Schema can help Google create rich features for specific use cases, like product reviews or event times, but for general advice, tips, and rankings, LLMs still just use the regular visible text on your page. They don’t weigh technical markup heavily. I say this as someone who tried both ways and noticed no reliable difference for broad topics.
Understanding Query Fanout and Drift
If this stuff sounds complex, it’s not as tricky as it seems. When someone asks an LLM a question, say, “2025 SaaS SEO guide”, the model will:
- Break the query apart (“SaaS SEO guide,” plus often add a year, like “2025”)
- Run each piece as a distinct search
- Assemble the answer by pulling from the top organic results for each subquery
This process is sometimes called “query fanout,” and it’s why the words you use in your titles and H2s really matter. LLMs regularly change their queries, this shifting pattern is called “drift.” Some days, the tool will keep adding years or extra phrases; sometimes, it won’t. On top of that, user intent shifts over time, and the model adapts.
If your content covers both broad and long-tail variations (like including “2025” and “guide” variants), you’re more likely to appear in answers, no matter how the LLM queries drift.
What Does Drift Mean for Content Creation?
If you’re noticing a page ranking for dozens of long-tail variations in Google Search Console, many with zero clicks, you’re not alone. These are usually LLM queries, pages being fetched for answers but not clicked by users. It might feel odd to see all these impressions and almost no visits, but this is normal for LLM surface area.
How to See What LLMs Are Fetching (and How to React)
You might be wondering how to know which queries LLMs use and which pages they’re pulling. The best method is to use Search Console’s “Queries” report. Look for:
- Sudden impressions for very detailed, low-click queries
- Position: most of these show up between 1-10, even if you get no clicks
But impressions in Search Console aren’t the only thing to look for. LLMs don’t drive standard SEO clicks most of the time, they send referral traffic.
If you want to measure LLM traffic, set up referral source tracking in GA4 or Looker Studio, track conversions and events by source/medium. Check for “chatgpt,” “perplexity,” “claude,” etc. in your traffic reports. It’s a small stream for now, but can reveal which pages are showing up in answers and which engines are sending that traffic.
Referral traffic, not organic clicks, is your signal that LLMs are mentioning your site and people care enough to check the source.
Quick Table: Google Search Console vs. Analytics for LLM Traffic
| Tool | What You See | How to Use for LLM Insights |
|---|---|---|
| Search Console | Impressions, average position, queries (has many zero clicks) | Spot which queries and pages get surfaced often |
| Analytics / Looker Studio | Clicks, sessions by referral, goal/event completions | Track actual visits sent from LLM-generated answers |
How to Influence Which Pages Get Cited
Let’s get practical. If you want to maximize how often LLMs pull your brand into answers, try these steps:
- Check what you already rank for in Google’s top 10. Those are your likely citation candidates.
- Use LLMs (or public query analysis tools) to see what variations the models run. If you’re not being cited for a phrase, look at the answer text; do you cover the full question, including year, use case, or persona?
- Update your existing content with newly discovered subtopics, either as expanded H2s or, if there’s enough volume, as separate pages.
- Monitor how the snippet summary changes when you update a page and resubmit it to Google; LLMs will often pick it up within a day.
It’s not always obvious when you need new content versus simply expanding a section. If you already rank around positions 8-12 for a detail, try boosting it on the same page with a new heading or FAQ. If your main page covers a broad term and you notice a long-tail ranking in distant positions, a standalone page is worth trying.
Sometimes, being cited is just about making your coverage as complete as possible, not reinventing your site’s architecture.
Are Large Brands Favored by LLM Results?
It is common for answers to cite big sites: Reddit, Quora, Wikipedia, and news outlets. Some of this comes down to topical authority and breadth, not secret formulas. LLMs tend to surface sites Google already trusts for the broken-down fanout queries. Then there’s the forum boost: Google’s results (and, in turn, LLMs) often stack discussions and Q&A forums above lesser-known blogs, since community votes signal relevance and depth.
That doesn’t mean smaller blogs cannot break through. For emerging topics or deep long-tail searches, niche experts get included, sometimes quickly, if they address gaps in coverage or phrase their H2s and intros to match exact topics users prompt LLMs with.
How LLM Query Language Shapes Your SEO
LLMs are full of hesitation. If you ask a general question, their internal fanout queries almost always get more specific: professions, years, use cases. A generic “best” post won’t always surface unless you also anticipate these refinements. For a post about accounting software, for example, your heads might use: “Best accounting apps for independent consultants (2025).” This structure increases your chances of being cited because you match how fanouts split long prompts.
Should You Worry About Schema and Technical SEO for LLMs?
Let’s try not to overthink it. Technical tricks are not the answer here. Schema is mostly a waste of time for informational and review content. Sure, if your entire business relies on event listings, recipes, or flight statuses, schema might help you earn a knowledge panel or rich snippet. But for almost everyone else, it’s not moving the needle in LLM visibility.
Some agencies may still sell hefty schema packages. I understand why, it’s a tangible deliverable and easier to sell than storytelling or topical mapping. But you don’t need it for blog posts, how-tos, or product comparison guides. And I say this from experience running sites both with and without technical markup for years. There’s just no pattern supporting schema as a citation trigger in most cases.
Focusing on technical SEO tricks often distracts from building actually useful content that matches evolving search and answer queries.
What to Actually Focus On: Topical Authority and Real-World Relevance
If you want more citations, let’s say the obvious: build content that users (and by extension, Google) want to rank and reference. That sounds naive, but it’s the only real lever that’s moved rankings or LLM inclusions for me. To do this, try:
- Using Google Search Console to discover what longer keys people actually search (not just keyword tool guesses)
- Publishing content on the topics you want to be cited for, start with intent-rich pages, not just general guides
- Expanding coverage where you already have a strong “beachhead” (if you rank position 8 for a related topic, go deeper)
- Making sure your content answers both the broad and the very specific questions that show up as long-tail fanout queries
Most LLMs reward content that thinks ahead. Add a section for “2025 updates,” or “for teams of 100 to 300 people,” or whatever you see popping up as query drift. Do not just focus on what keyword tools suggest, check your impressions and landing page reports daily or weekly to see what the engines are actually querying.
Concrete Example (But Not the Usual Suspects)
If you run a gardening advice blog and want to show up for “best containers for balcony gardens,” look at related phrases showing impressions in console like “2025 balcony planning,” “container plants for small spaces,” and “balcony garden soil tips.” Add these as H2s, not separate posts. Then review engines like Perplexity; see which of your pages get cited for these subtopics and cross-link among them to reinforce relevance.
How to Reverse-Engineer LLM Fanout Queries
Sometimes you want to see exactly what underlying searches an LLM is running. You can do this by:
- Using browser dev tools to inspect the XHR calls made by ChatGPT or Perplexity as they collect citations
- Reviewing the “network” tab for queries or URLs loaded in real time (look for search payloads, not just the answer)
- Examining the structured data on your own site, but realize it rarely matters for broad question answering
Then, you can test. Add a new H2 or FAQ to your page for a subtopic you know is popular, submit it to Search Console, and fetch as Google. Wait a few minutes, then prompt Perplexity or Claude with the parent query, see if your update is cited. This feedback loop is fast and oddly satisfying. Sometimes it’s almost like you’re “teaching” the LLM new facts (you are, kind of), but it all depends on Google recognizing your update first.
Are LLMs Research Tools?
I have mixed feelings here. They are not research tools in the traditional sense. You can game them by publishing new claims or summaries on your blog, then watch LLMs regurgitate those points as facts days later. Why? Because most simply synthesize what Google ranks, with little validation. If you publish nonsense and get it indexed, LLMs can repeat it. This is dangerous because the knowledge base can get polluted fast. But it’s also an opportunity if you’re careful and honest, use it for good, not to flood the web with junk.
Testing Your Own Hypotheses: The Role of “Just Check It” Mindset
This is something I think is underappreciated. So many SEO strategies are passed around as gospel but never tested. Take meta descriptions: some people swear by them, others rank fine without. Try removing or swapping them out for a few pages. Watch what happens. The same goes for bullet point counts, image ratios, or the need for certain schema types. Most of the time, less is more. Don’t let superstition get in the way of what works for your niche.
When Should You Make New Pages vs. Expand Old Ones?
This is a judgment call. If your site ranks for a “root” query and you see closely related long-tails with good impressions, you’re likely fine adding more depth (new headings, FAQs) to the existing page. If you see distant long-tails pop up (high impressions, very different angle), that’s a sign to spin up a standalone post or guide and link it back to your beachhead.
Quick example: You rank page 3 for “indoor container garden soil” and page 1 for “balcony containers.” Try making a guide connecting the two or updating the page with the weaker ranking to get that bump. Sometimes just adding a relevant H2 gets you on page one. I have seen this work many times and I am not even always sure why the boost happens the way it does.
The Big Picture: What Actually Moves the Needle
- Track what’s working not just in organic, but in LLM-generated traffic; measure referral sources, not just search positions.
- Anticipate drift, update and expand your content to catch evolving fanout queries and new ways users frame questions.
- Do not waste hours tweaking schema for blog posts, focus on clear, specific answers and up-to-date coverage.
- Don’t take theories as fact. Test changes on your own pages and see the results in your reporting tools.
- Remember, LLMs are just reflecting what’s in Google and other engines. If you want to appear, focus on building real authority through useful, present-tense, and navigable content.
If you’re frustrated that big brands seem to dominate LLM answers, dig deeper. Find where query drift lets your expertise shine. The more specific you get, and the more you tune your coverage to the way people actually ask, not just how you wish they did, the more you’ll find the algorithms pick you up. It may take some manual testing. At times, it’ll feel like trial and error. And the landscape keeps shifting, so what works one week might slip the next. But that’s always been the game in search, hasn’t it?
Need a quick summary of this article? Choose your favorite AI tool below:
