LLM SEO: how to get your business into AI answers
Short answer
LLM SEO is the work of getting your business named and cited when people ask large language models like ChatGPT, Claude, Perplexity and Gemini for answers. Those answers come from two places: what the model learned in training, and pages it fetches live from the web. You can influence live retrieval fairly quickly by letting the search crawlers in and writing answer-first pages. Training data moves slowly and mostly reflects what other sites say about you. Measure it with a fixed prompt log and AI referral traffic, and be sceptical of anyone selling guaranteed placement.
LLM SEO means optimising for the answers large language models write. When a buyer asks an AI tool who to hire, your name should come up and your page should get cited.
This guide is for owners and marketing leads in the US, the UK and Australia sorting the real from the rebrand. Interest is climbing. Semrush (October 2026) puts “llm seo” at about 1,600 searches a month in the US, 480 in the UK and 140 in Australia.
You’ll also see it called GEO, AEO or LLMO. We explain how those labels relate to classic search in GEO vs SEO. This piece is about the mechanics and the levers.
How LLM answers are actually produced
An LLM answer is built from two sources, and they need different work. One is the model’s training data. The other is live retrieval, where the tool searches the web at the moment you ask.
| Training data | Live retrieval | |
|---|---|---|
| What it is | Text the model learned from before release | Pages fetched and read when the question is asked |
| Example crawler | GPTBot, ClaudeBot, Google-Extended (for Gemini) | OAI-SearchBot, PerplexityBot, Claude-SearchBot |
| Cites sources? | No, it answers from memory | Usually, with links |
| How fast your changes show | Only when a new model is trained | As soon as the page is crawled again |
| What shapes it | How often and how consistently the web describes you | Crawler access, indexing, and how well a page answers the question |
The crawler split comes straight from the providers. OpenAI’s crawler documentation says GPTBot crawls content “that may be used in training”, while OAI-SearchBot “is used to surface websites in search results in ChatGPT’s search features”. Anthropic and Perplexity draw the same line between training and search bots.
This split is the most useful idea in LLM SEO. Live retrieval can move in weeks. Training data reflects reputation built over years.
Lever one: let the search crawlers in
If an AI tool’s search crawler can’t reach your site, you won’t be cited in its live answers. That’s the one hard rule every provider publishes.
The bots to check in robots.txt and in your CDN or firewall:
- OAI-SearchBot (ChatGPT search). OpenAI’s publisher FAQ says not to block it if you want your content in ChatGPT summaries and snippets.
- PerplexityBot. Perplexity’s bot documentation says it is “designed to surface and link websites in search results” and isn’t used for training.
- Claude-SearchBot. Anthropic’s crawler help article says blocking it may reduce your visibility in Claude’s search results.
- Bingbot and Googlebot. Plenty of AI tools lean on major search indexes, so normal indexing still matters.
User-triggered fetchers (ChatGPT-User, Claude-User, Perplexity-User) are different. Perplexity says its one “generally ignores robots.txt rules” because a person started the request.
Google is the awkward one. Google-Extended controls whether your content trains Gemini, and Google’s crawler documentation says it also covers grounding in Gemini Apps. It doesn’t affect Google Search or AI Overviews. So blocking it is a real trade-off for Gemini, not a free privacy win.
Our view: you can block training bots and still allow search bots. For most service businesses we’d allow both, because being in the training data is how a model learns you exist. But that’s a business decision, and the robots.txt should reflect it on purpose, not by accident.
Lever two: entity consistency
LLMs recommend businesses they can describe with confidence. That confidence comes from the same facts repeating across the web.
Your business name, what you do, where you operate, who runs it and roughly what you charge should read identically on your website, Google Business Profile, Bing Places, LinkedIn, directories and review sites. A firm with two descriptions and two addresses is harder to summarise, so a model hedges or picks a competitor with a cleaner story.
Start with an About page that states the core facts plainly. Then audit the top 20 places you’re listed and fix every mismatch. It’s dull work, which is why most firms skip it.
Lever three: third-party mentions
What other sites say about you matters more in LLM answers than in classic SEO. A model answering “best payroll provider for UK charities” draws on reviews, comparison articles, industry lists and forum threads, whether or not they link to you.
So the work looks like PR. Get into the roundups your buyers read, collect reviews where your sector looks, and turn up on the podcasts your industry follows. We cover one channel that turns up in AI answers a lot in our piece on Reddit and AI search.
Don’t fake any of it. In the US, the FTC’s rule on fake reviews bans buying reviews and AI-generated fake ones.
Lever four: answer-first content
Pages get cited when they contain a clear, quotable answer to the exact question. Retrieval systems lift passages, so each section should stand on its own.
Open every section with the answer in two or three sentences. Use real specifics: prices, timeframes, locations, named standards. Cover the follow-up questions a buyer asks next, because tools like ChatGPT and Google’s AI features split one question into several searches. We go deeper on Google’s version of this in how to rank in AI Overviews, and on OpenAI’s in how to rank in ChatGPT.
A worked example: an Australian IT support firm
This one is hypothetical, but the pattern is common.
A managed IT provider in Brisbane asks ChatGPT, Perplexity and Claude, “Who are the best IT support companies for small law firms in Brisbane?” Two competitors appear in all three. They appear in none.
The audit finds three things. Their Cloudflare bot settings block AI crawlers, so none of the search bots get through. Their service page talks about “digital transformation” and never says “IT support for law firms” or names the practice software they support. And they have no presence in any Brisbane IT roundup, while both competitors have reviews on the main review platforms.
The fix order follows the table above. Crawler access first, because live retrieval responds once pages are recrawled. Then an answer-first service page rewrite. Then reviews and roundups, which take months and feed future training data.
How to measure LLM SEO
Measure with a prompt log and your analytics. No single dashboard covers every AI tool.
Write 10 to 20 questions a real buyer would ask. Run them monthly in ChatGPT, Perplexity, Claude and Gemini, in fresh sessions. Record if you’re named, if you’re cited with a link, and who appears instead. Answers vary between runs, so trends over months matter more than any single check.
Then use the data that exists. ChatGPT adds utm_source=chatgpt.com to referral links, according to OpenAI’s publisher FAQ. Bing’s AI Performance report, launched in February 2026, shows citations across Microsoft Copilot and Bing’s AI summaries. In GA4, group referrals from the AI tools into one channel so you can see the trend.
On Arché Academy, the tuition centre Glen runs, AI-assistant sessions in GA4 went from 1 in June 2026 to 28 in September 2026. Small numbers, but a clear direction, and they come from the same playbook we use for clients.
What’s hype in LLM SEO
Much of LLM SEO is sound SEO with a new name, plus some products nobody needs.
Hype: guaranteed rankings in ChatGPT or any other model. No provider publishes a way to buy organic placement, and OpenAI’s ChatGPT search help page says “placement is not guaranteed”.
Hype: llms.txt as a ranking factor. It may help some tools read your site, and we cover the details in our llms.txt guide. Google says you don’t need AI text files for its AI features.
Hype: mass-producing AI pages to cover every prompt. Google’s spam policies call that scaled content abuse when it adds no value, and thin pages give a model nothing to quote.
Real: crawler access, consistent entity facts, mentions on trusted sites, and pages that answer questions properly. In our experience, most firms missing from AI answers have a problem in one of those four, and usually it’s the dull one.
Quick answers
What should I check this week?
Your robots.txt and CDN bot settings for OAI-SearchBot, PerplexityBot and Claude-SearchBot. Then run five buyer questions through each tool and note who’s named.
Should I hire an agency for LLM SEO?
If access is fixed and you’re still missing, the remaining work is content and reputation over months. Our GEO service runs alongside SEO from US$1,500 a month for international clients, with a 6-month minimum.
Is there a quick way to see where I stand?
Yes. Our free AI visibility check runs 10 fixed questions across ChatGPT, Perplexity, Gemini and Google AI Overviews.
If you’re outside Singapore, see how we work with international clients, or get in touch for a free consultation and a written proposal.
AI tools change their crawlers and policies often. Details here were checked in October 2026.