A user asks ChatGPT for the best tool in your category. Two things can happen: it names you, or it names a competitor. There is no page two to fall back to. That is the whole game.
AEO, GEO, LLMO, and why the acronyms barely matter
You will see three labels for this work. AEO, answer engine optimisation, is about structuring a page so an engine can lift a clean, correct answer from it. GEO, generative engine optimisation, is about earning visibility and citations inside AI-generated results across platforms. LLMO, large language model optimisation, is about shaping how the models describe your brand and topic.
They are the same job seen from three angles: be the clearest, most trustworthy, most quotable source a model can find on your topic.
How the assistants choose sources
Most consumer AI search does retrieval first. It runs a fairly normal search, pulls a handful of pages, and writes its answer from them. Once you know that, the rules stop feeling mysterious.
You still have to rank, because no presence in the top 10 for the query means no place in the retrieval set and no citation. Classic SEO is the entry fee. The answer then has to be easy to lift, so models favour pages that state the point in the first sentence under a heading over pages that bury it under 400 words of throat-clearing. And when it is a close call, consensus wins: if five trusted sources say the same thing and you are the sixth, you get quoted ahead of a lone contrarian.
Make your page quotable
The pattern is always the same: answer the question in the first sentence under each heading, then expand. Phrase the headings the way people ask, so "how much does X cost" beats a heading that just says "Pricing". Keep the facts in plain sentences and short lists, because anything locked inside an image or a sprawling table is invisible to the model.
On longer pages, add a two or three line summary near the top. And show the date, the sources and the author, so the model has something concrete to verify your claims against.
Become an entity, not just a website
Models build a map of who counts as an authority on each topic from signals right across the web, and you want a clear place on that map. Describe your brand the same way everywhere: same name, same one-line description, same category. Earn mentions on sites the models already trust, even ones that carry no link. Keep any reference-site presence accurate, from Wikidata and Crunchbase to solid industry directories. And publish enough depth on your core topic that the association becomes automatic.
AI visibility is downstream of SEO. The pages getting cited in AI answers are, overwhelmingly, the pages that already rank well and read clearly. Fix the fundamentals first.
Measure it, because rank trackers will not
Once a month, prompt ChatGPT, Gemini and Perplexity with your ten most important buyer questions and log whether you are named. Track branded and category prompts for the trend rather than any single score. Watch referral traffic from chatgpt.com, perplexity.ai and Google's AI surfaces in analytics. Then add a dedicated AI-visibility tool only once the channel is big enough to be worth a line item.
What are AI Overviews, and how do you earn a place in them?
AI Overviews are the AI-generated answer block Google shows above the classic results. Google pulls from a handful of pages, writes a summary, and cites those pages with links. You earn a spot by ranking in the top results for the query and by making one section of your page trivially easy to quote.
Directional, from public citation studies through 2026. The takeaway is blunt: page-one rankings are the entry fee.
What consistently gets a page cited is a direct answer in the first sentence or two under a clear heading, headings phrased as the question a real person would ask, facts kept in plain text and short lists rather than trapped in an image, and a visible author, date and set of linked sources so the claim can be checked.
| Query type | Example | AI Overview behaviour | What to do |
|---|---|---|---|
| Simple factual | "what is a canonical tag" | Answered in place, low click-through | Own it for brand and citation, do not expect the click |
| How-to | "how to submit a sitemap" | Overview plus steps, moderate clicks | Number the steps, answer-first on each one |
| Commercial research | "best crm for small business" | Overview lists options, high click-through | Comparison tables, honest pros and cons, pricing |
| Local or transactional | "plumber near me" | Usually no Overview, map pack instead | Focus on Business Profile and local pages |
Track it: in Search Console, a page that holds its impressions but loses clicks after an update is usually being summarised in an AI Overview. That is your signal to push the page toward commercial and comparison intent, which still earns the visit.
Generative Engine Optimization (GEO): the moves that shift AI citations
GEO is optimising to be named and cited inside AI-generated answers across Google's AI surfaces, ChatGPT, Perplexity, Gemini and Copilot. It builds on classic SEO rather than replacing it: the assistants retrieve from search, so you have to be retrievable first, then quotable, then trusted.
Based on the GEO research literature and our own monthly prompt testing through 2026.
A monthly GEO routine that works:
- Pick ten prompts a buyer would type. Run them in ChatGPT, Gemini and Perplexity. Log whether you are named and who is named instead.
- Read the pages that got cited. Match their clarity, then add what they lack: fresher data, a cleaner table, a real example.
- Chase one or two third-party mentions on sites the models already quote in your space.
- Re-run the prompts next month. Watch the trend, not any single answer.
Should you add an llms.txt file, and how do you handle AI crawlers?
llms.txt is a proposed file at your domain root that points AI tools to your most important content in clean Markdown. It is not an official standard and no major model is confirmed to use it for ranking, so treat it as cheap housekeeping, not a growth lever. The AI-crawler access decisions matter far more.
| Bot | Purpose | Common choice |
|---|---|---|
Googlebot | Search and AI Overviews | Always allow. Blocking it removes you from Google |
Google-Extended | Gemini grounding and training | Allow for AI visibility; block only to opt out of training |
GPTBot | OpenAI training and browsing | Allow for ChatGPT search visibility |
OAI-SearchBot | ChatGPT search index | Allow to appear in ChatGPT answers |
PerplexityBot | Perplexity index | Allow for Perplexity citations |
CCBot | Common Crawl open dataset | Your call; blocking has little SEO cost |
Common mistake: blocking Google-Extended or GPTBot "to protect content", then wondering why the brand never shows up in AI answers. If AI visibility is a goal, those bots need read access to the pages you want cited.
Entity SEO: becoming something Google and LLMs recognise
An entity is a thing the search and language models hold in their knowledge graph: a person, a company, a product, a concept, with attributes and relationships. Rankings and AI citations both lean on entities now, because models retrieve and reason about things, not strings of characters. Topical authority is the same idea from the content side: publish enough depth on one topic that the association becomes automatic.
Strengthening your entity comes down to four moves. Say the same thing everywhere: one brand name, one one-line description, one primary category, across your site, your profiles and every directory. Mark it up with Organization and Person schema, using sameAs links to your verified profiles so Google can connect the dots. Get accurate entries into reference sources like Wikidata, Crunchbase and solid industry directories, with Wikipedia only if you genuinely meet its notability bar. And build the topic cluster: a pillar page plus supporting articles that link to each other and cover the subject end to end.
AI search checklist
- You rank on page one for the queries you want to be cited on
- Every section opens with a direct, self-contained answer
- Headings are phrased as real questions
- A short summary sits near the top of every long page
- Author, date and sources are visible on every article
- Your brand name and description are identical across the web
- AI citations and assistant referral traffic are checked monthly
