The terminology
LLM SEO and LLM optimization
LLM stands for large language model, the kind of system behind ChatGPT, Gemini, Claude and Perplexity. LLM SEO and LLM optimization mean improving how a business appears in the answers those systems write. It is the same aim as answer engine optimization, named after the technology instead of the product.
Not the same as AI SEO
LLM SEO is easy to confuse with AI SEO, which usually means using AI tools to do ordinary SEO: drafting content, researching keywords, checking pages. LLM optimization runs the other way. The model is the audience, not the tool, and the result is measured in its answers rather than in a ranked list.
Where an answer's information comes from
A model can answer from what it learned in training, or from pages retrieved when the question is asked. When pages are retrieved, some engines show them as sources. In our sample of 16 buying questions in Cincinnati, ChatGPT attached sources to all 16 answers, and its 84 citations came from 63 different domains. What that sample contained.
On general how-does-it-work questions, we have recorded answers that arrived with no sources at all. So the first question for any LLM optimization plan is which engine, and which kind of question, it is aimed at.
Letting the crawlers read the site
An engine can only retrieve a page its crawler is allowed to read, and OpenAI, Anthropic, Perplexity and Google each name their own crawlers, which a site's robots.txt can allow or block. Google is a special case. It says its Google-Extended token governs use of content for training Gemini models and for grounding in Gemini Apps, and "does not impact a site's inclusion in Google Search". Google's AI Overviews follow Search's own snippet controls instead. Google's crawler documentation.
LLM optimization techniques: what the evidence supports
The study most often cited for these techniques reported that its methods "can boost visibility by up to 40% in generative engine responses". In the benchmark results its authors released, the best method raised a source's share of the answer from 19.4% to 27.3%, and made it worse on 28% of queries. The top methods were produced by instructing a model to add quotes "even though fake and artificial" and to invent sources. The full breakdown.
What we recommend instead is ordinary and checkable: accurate facts in the places an engine reads, pages that answer a real buyer's question specifically enough to quote, and references from sources already in the category. We measure which of those is missing for a business before recommending any of them.
AEO vs SEO vs GEO · Answer engine optimization services · Should you pay for llms.txt? · How we measure
Questions we get asked about this.
What is LLM SEO?
Improving how a business appears in answers written by large language models such as ChatGPT, Gemini, Claude and Perplexity. It is another name for answer engine optimization.
What is LLM optimization?
The same aim as LLM SEO: being named, described correctly and cited when a large language model answers a question about a business's category.
Is LLM SEO the same as AI SEO?
No. AI SEO usually means using AI tools to do ordinary SEO. LLM SEO treats the model as the audience and measures the result in its answers.
Does blocking AI crawlers remove a business from AI answers?
A crawler that respects robots.txt will not read the pages it is blocked from, so that engine cannot retrieve them. Google says its Google-Extended token does not affect inclusion in Google Search, and its AI Overviews use Search's snippet controls instead.
Which LLM optimization techniques work?
We found no controlled test on a live engine. In the released results of the field's most cited study, the best method raised a source's share of the answer from 19.4% to 27.3% and made it worse on 28% of queries, and it was produced by telling a model to add quotes even though fake and artificial.
We publish what we find.
We're measuring AI answers in Cincinnati and will publish what we find, including our own results, including when they're unflattering. No pitch attached.
One email a month at most, and you can leave with one click. We do not sell your address and we will not add you to anything else.