Rank number one in Google and you might still be invisible where your next customer is actually looking. Only 12% of the pages that AI assistants cite also rank in Google's top ten, according to Ahrefs' 2025 analysis, and around 80% rank nowhere at all. The link everyone fought over for twenty years is no longer the thing that gets you named in an AI answer.
Your customers have started asking ChatGPT and Perplexity the questions they used to type into Google. The answer they get names a few businesses and quietly leaves out the rest. LLM SEO is the work of being one of the named few. This guide explains what it is, what to do in your first month, how to check your own visibility for free, and what the work costs. We run it as a paid service, so the plan below is the one we actually use.
What is LLM SEO?
LLM SEO is the work of optimising your business so AI assistants cite and recommend you inside their answers, rather than only ranking your links on a results page. The target changes from winning a position to becoming the source a model trusts enough to name out loud, in the sentence a buyer actually reads.
The old game was ten blue links. You fought for position, someone clicked, and the visit was yours. The new game is a single written answer, and the click often never happens. When Google shows an AI summary, people click a link just 8% of the time, against 15% when there is no summary, and only 1% click a source cited inside the summary itself, per Pew Research's 2025 study. Pew also found that around 18% of US searches produced an AI summary as far back as March 2025, and that share has only grown since.
How often people click, with and without an AI summary
- No AI summary shown
- 15%
- AI summary shown
- 8%
- A source cited inside the summary
- 1%
Being cited is the new visit. If the answer names three suppliers and yours is not one of them, you were not outranked, you were left out of the paragraph entirely. That is the whole job: get your business into the paragraph, and be the source the model reaches for when it writes the next answer on your topic.
None of this makes traditional SEO worthless. A page still has to exist, load and be crawlable before any assistant can quote it, so the technical groundwork carries straight over. What changes is the finish line, from a ranked link a person clicks to a fact a machine repeats on your behalf.
LLM SEO vs GEO vs AEO vs LLMO, are they the same thing?
Mostly they are the same thing wearing different labels, and the vocabulary has not settled yet. Each term points at one goal, getting named inside AI answers, so the differences are emphasis and who coined them, not method. Here is the clean map, because most guides only mention these acronyms in passing and never lay them side by side.
| Term | Stands for | Emphasis | Who uses it |
|---|---|---|---|
| LLM SEO | Large language model SEO | Getting cited and recommended inside AI assistant answers | Agencies and SEO teams |
| GEO | Generative engine optimisation | Broad visibility across generative engines | Marketers and researchers |
| AEO | Answer engine optimisation | Winning the direct answer, snippets and voice included | Content and technical SEOs |
| LLMO | Large language model optimisation | Shaping how models describe your brand and facts | Brand teams, still emerging |
Pick whichever term your clients already use and get on with the work. We tend to say LLM SEO because the buyers we speak to call the tools LLMs, but the tactics underneath do not change with the acronym. Anyone telling you these are four separate disciplines is selling four invoices for one job.
One caution is worth stating: the terms will keep shifting as the tools do, so do not build your strategy around a definition. A vendor who insists their acronym is the only correct one is usually protecting a product, not describing the work.
Why does LLM SEO matter now?
It matters because AI search is still small but growing faster than any other channel, and the buyers it sends are close to ready. ChatGPT reached about 900 million weekly active users by February 2026, up from 800 million in October 2025, per OpenAI's figures reported by TechCrunch.
ChatGPT weekly active users
- October 2025
- 800M
- February 2026
- 900M
Scale is only half the story. Generative AI platforms drew about 9.5 billion web visits a month on average in the year to May 2026, up 70% year on year, with unique visitors up 57% to 655 million, according to Similarweb's 2026 data. That is fast growth from a small base, which is the real shape of it.
Here is the counterweight, stated plainly. AI still drives less than 1% of referral traffic today, and organic search dwarfs it, per BrightEdge's 2025 data. Anyone selling AI search as your main channel this year is ahead of the evidence, and we would rather you knew that going in.
Quality is where it gets interesting. Adobe tracked US retail visitors from generative-AI sources closing the conversion gap, from 43% less likely to buy in July 2024 to just 9% by February 2025, while browsing 12% more pages and bouncing 23% less, per Adobe Analytics in 2025. These are people doing research, not idle browsers. Adobe, reported through Digital Commerce 360 in 2025, also found 38% of US consumers had used generative AI for shopping, and 73% of them treat it as their main product-research source.
The traffic is small today, but the buyers it sends are already reaching for their wallets.
The direction of travel is clear enough that marketers are moving on it. A 2025 Ahrefs study put an AI Overview alongside 34.5% lower click-through for the top organic result, which tells you the answer is absorbing the click rather than passing it along. A Scribewise survey from September 2025, reported by eMarketer, found 54% of US marketers planning to fully adopt a GEO strategy within three to six months, so the window to be early is closing.
How do LLMs decide who to cite?
They do not simply reuse Google's top ten. Assistants pull from third-party mentions, structured facts and trusted community sources, which is why only 12% of AI-cited pages also rank in Google's top ten. Ranking well helps you get fetched, but it is not the ticket into the answer on its own.
The overlap does vary by engine. Google's own AI Mode, which sits closest to its search index, shows about 54% domain overlap and 35% URL overlap with the classic top ten, per Semrush's 2025 comparison study. Standalone assistants like ChatGPT and Perplexity overlap far less, so there is no single set of pages to copy. Each engine cites its own crowd, and you have to win them separately.
How much AI citations overlap with Google's top 10
- AI assistants overall (Ahrefs)
- 12%
- Google AI Mode, same domain (Semrush)
- 54%
- Google AI Mode, exact page (Semrush)
- 35%
What fills the gap is the citation substrate: the Reddit threads, YouTube videos and third-party listicles where your brand gets named by someone other than you. This substrate moves fast. Reddit's share of ChatGPT citations fell from around 60% in early August 2025 to about 10% by mid-September, on Semrush's 2025 tracking, so betting everything on one platform is a mistake you pay for later.
The practical read is simple. Consistent brand mentions across the places a model already trusts will beat one perfect page on your own site. You want the wider web saying your name, in threads and lists and reviews, before the buyer ever thinks to ask.
How to do LLM SEO: your first 30 days
Work in a fixed sequence rather than a loose list: measure first, fix your own facts second, then earn outside mentions. This is the order we use when we run LLM SEO for a client, because it stops you polishing schema for a fortnight while the answers still never name you.
Week one is for measuring. Pick the ten questions a real customer would type, run them across ChatGPT, Perplexity and Gemini, and log who gets cited. This doubles as your free visibility baseline, which the next section covers in full. You cannot fix what you have not counted, so resist the urge to change anything yet.
Weeks two and three are for the fundamentals. Give each buyer question one strong, self-contained page with a clear heading and a direct answer near the top. Add Organization and Article schema so the facts are machine-readable. Then put real facts on the page, the prices, specs and dates a model can lift and quote, because an assistant cannot cite a claim it cannot find.
Off-page presence fills weeks three and four. Get named in the third-party lists and community threads the assistants actually cite, and earn brand mentions from people who are not you. When we do this for a client, the outside mentions move the needle first; tweaking your own schema looks busy but rarely changes who gets named. As for llms.txt, add it and expect little. The file is cheap to publish and tells AI crawlers which pages matter most, but adoption is early and the payoff is unproven, so treat it as housekeeping. If you want the deeper mechanics of getting ChatGPT to recommend your business, we cover them in a separate guide.
How do you measure your LLM visibility for free?
You do not need a paid tool to start. A fixed set of buyer prompts, run every month across the main assistants and logged in a plain spreadsheet, tells you most of what a tracker would, for the price of an hour.
Here is the method we hand to clients. Keep the same ten prompts each month so the numbers stay comparable. Run them in ChatGPT, Perplexity and Gemini separately, because each cites a different crowd and a blended score hides that. Log every brand and link the answer names, then count how many of your ten questions mention you at all. That figure is your share of voice, and watching it move month to month is the entire point of the exercise.
A paid tool earns its place later. Once you are tracking dozens of prompts across several engines, or reporting to someone who wants charts, tools like Profound, Peec or Semrush start saving real time. Before that, the manual log is honest ground truth and costs nothing to keep. This kind of repeatable monthly watching is exactly the job we hand to AI marketing agents that run the work for the clients we manage.
Whichever route you pick, record the date and the exact wording of every prompt. Assistants rephrase and reorder their answers constantly, and without a dated record you will end up arguing with yourself about whether anything actually improved. The log is dull to keep, and it is the part that pays off.
What does LLM SEO cost?
Costs fall into three bands: do it yourself with free prompts and your own time, tool-assisted with a subscription and light monthly work, or done-for-you on an agency retainer. We publish real prices where most guides in this space publish none, so you can size the decision before you talk to anyone.
The DIY band costs your time and nothing else. The ten-prompt log, the on-page fixes and a handful of community mentions are all within reach of a founder willing to spend a morning a month on it. Progress is slow, but it is genuine, and it teaches you how the engines behave.
Tool-assisted work adds a tracker subscription plus the hours to act on what it shows. You see more, across more engines, and you catch shifts like a citation source collapsing before they quietly cost you visibility. The work of fixing pages and earning mentions is still yours to do.
The done-for-you band is a retainer, where an agency runs the measurement, the on-page work and the mention-building for you. If that is the route you want, our AI SEO service lists what it costs. The paid counterpart, advertising inside ChatGPT, sits alongside it for when you want the answer slot faster than you can earn it.
Where to start this week
Start with the cheapest step that tells you the truth: run your ten buyer prompts through ChatGPT, Perplexity and Gemini and see whether the answers name you. Ranking is no longer the same as being cited. Fix your on-page facts, earn mentions where the models already look, and measure the same prompts every month so you can see the work land. If you would rather we ran the whole thing, including a custom AI build around it, we begin with a short call and no obligation to go further.
Questions people actually ask
What is LLM SEO, and how does it differ from classic SEO?
LLM SEO is the work of getting your business named inside AI-generated answers, while classic SEO is about ranking pages in a list of blue links. The two barely overlap. Ahrefs found in 2025 that only 12% of the pages AI assistants cite also sit in Google's top ten, so winning one no longer guarantees the other.
How do I get ChatGPT and Perplexity to name my business?
Publish self-contained pages that answer real buyer questions with facts an assistant can lift, such as prices, specs and dates. Add Organization and Article schema so those facts are machine-readable. Then earn mentions on the third-party lists and community threads these tools already cite, because a model tends to recommend whoever the wider web already recommends.
What is an llms.txt file, and is it worth adding?
An llms.txt file is a plain text file that tells AI crawlers which of your pages matter most, a bit like a sitemap written for language models. It is cheap to add and does no harm. Adoption is real, but the payoff is still unproven, so treat it as housekeeping rather than a lever that moves your visibility on its own.
How can I check how visible my brand is in AI search?
Write down the ten questions a buyer would actually ask, run them each month inside ChatGPT, Perplexity and Gemini, and log which brands the answers name. Count how many of the ten mention you to get a simple share-of-voice score. Paid trackers exist, but this manual log costs nothing and gives you a fair baseline to watch.
Is LLM SEO just another name for GEO?
Mostly, yes. GEO, short for generative engine optimisation, and LLM SEO describe the same goal of being cited and recommended by AI systems, and people use the terms interchangeably. The labels are still settling, with AEO and LLMO circling the same idea, so focus on the work rather than the acronym.
