A client forwarded me a ChatGPT screenshot a while back. Someone had typed “SEO consultant in Bath” into ChatGPT and seoburf.com came up in the response. That was the first time I’d seen it happen with my own site, and it raised an obvious question, why that page, and what put it there?
The honest answer is that optimising for LLMs is not a separate discipline from traditional SEO. It is the same work, done properly, with a sharper focus on a few things that matter more than they used to.
What LLMs actually do with your content
Large language models do not crawl the web in real time the way Google does. The models themselves are trained on data up to a cutoff date. But the products built on top of those models, Perplexity, ChatGPT with browsing, Google AI Overviews, Bing Copilot, do have real-time web access. They use retrieval-augmented generation: they search the web, pull in content from current sources, and use that content to construct their answers.
Like this stuff? Get it weekly.
The Wednesday Roundup. SEO and AI search news, tactics and what is changing this week. No fluff.
Subscribe →When an AI product cites your site, it is because its retrieval system found your page, judged it relevant to the query, and extracted content from it to inform the response. The page that gets cited is not always the one ranking first in Google. It is often the one that most directly answers the specific question being asked.
WORK WITH JAMES
Get a straight answer on your SEO.
No account managers. No generic reports. Direct access to someone who has been doing this for 25 years.
What actually moves the needle for LLM citations
The signals that make a page more likely to be cited by an AI are not dramatically different from the signals that make a page rank well in traditional search, but the weighting is different. Traditional SEO rewards authority and relevance at domain and page level. LLM retrieval rewards specificity and directness at the answer level.
Pages that get cited tend to share a few characteristics. They answer a specific question directly and early in the content, not buried below three paragraphs of context. They are structured so individual answers are easy to extract: clear headings, concise paragraphs, FAQ sections with defined questions and answers. They contain original information not available in identical form elsewhere, first-hand examples, proprietary data, or genuinely distinctive analysis.
The page on my own site that ChatGPT cited was the Local SEO Consultant landing page. It ranks well for local SEO queries and directly mentions Bath in multiple contexts. That geographic specificity is likely why it surfaced for “SEO consultant in Bath” rather than a more generic page. Specificity, topical, geographic, use-case specific, is consistently more valuable for LLM citation than broad content that tries to cover everything.
Technical signals that help with AI Overviews
Google AI Overviews pull primarily from pages ranking in the top five organic results for the query. The clearest path to appearing in AI Overviews is to rank organically for the question being asked. There is no separate optimisation that bypasses ranking, ranking is the mechanism.
That said, content format affects whether Google extracts from it even when you rank. FAQPage schema markup makes your Q&A structure legible to Google’s extraction systems. Clean heading hierarchy helps the model understand your content’s structure. Concise direct answers in the first sentence after a heading increase extraction probability. These are all standard SEO practices that also serve LLM retrieval.
Is LLM traffic worth pursuing right now?
For informational queries, “what is X”, “how does Y work”, “best way to Z”, AI Overviews and chatbot responses are increasingly the first stop for users, particularly on mobile. If your target audience is asking these questions, LLM presence is becoming meaningful.
For transactional and commercial queries, “buy X”, “X near me”, “book Y”, traditional search results still dominate. AI Overviews appear less frequently here and drive less click substitution when they do appear.
Where I am seeing genuine LLM traffic value: specific professional service queries in mid-sized markets. A solicitor in a specific city. An accountant specialising in a specific sector. An SEO consultant in Bath. These are queries where the AI model has limited high-quality sources and where appearing in a cited response creates meaningful brand visibility for people who then search specifically for you.
Frequently asked questions
It means structuring your content so that large language models like ChatGPT, Claude, Perplexity, and Google Gemini are more likely to surface your site as a source when answering relevant questions. This is sometimes called GEO (Generative Engine Optimisation). The signals that matter are similar to traditional SEO, authority, clear structure, cited sources, but with greater emphasis on being definitive and quotable on specific topics.
Traditional SEO aims to rank in a list of ten blue links. LLM optimisation aims to be cited as a source in a generated answer, or to have your content used as training or retrieval context. The key differences are that LLMs reward being the definitive source on a narrow topic more than broad coverage, that exact phrasing matters more because LLMs tend to paraphrase authoritative sources, and that brand recognition across multiple sources amplifies your citation likelihood.
Perplexity AI searches the web in real time and cites sources directly in its answers. Google AI Overviews (Search Generative Experience) pull from web content for most queries. ChatGPT with web browsing enabled can access current content. Claude, by default, does not browse the web but can when tools are enabled. Bing Copilot uses Bing index data. Each has different retrieval logic but consistently rewards authoritative, clearly structured content.
There is no direct submission process. Google AI Overviews draw from content that already ranks well for the query. The most reliable path is to rank in the top five for informational queries in your niche, use clear heading structure, provide direct answers near the top of the page, and use structured data where applicable. Featured snippet optimisation is the closest proxy for AI Overview inclusion.
The signals that help with LLM citation, authority, clear structure, specific expertise, external citations of your content, overlap heavily with traditional ranking signals. Optimising for LLMs does not require a separate strategy. Sites that are genuinely authoritative on specific topics tend to perform better in both environments.
TESTIMONIALS
