field note
first published 2026-11-05
AI Search for Beginners, episode 2: the glossary
The words from episode 2 of AI Search for Beginners (GEO, AEO, LLM SEO, AI SEO, AI visibility): what each one means, which bits are new, which bits are a new invoice, and every number in the episode with where it came from and when.
836 words · 4 min read · ai search
This is the page episode 2 points at. The episode is about the words the industry invented when AI answers arrived, and the one test that matters: does any of them ask you to do something different? The glossary is short on purpose. Every number from the episode is at the bottom, with where it came from and when.
The words
- SEO. Search engine optimisation. Making your pages the ones Google picks when somebody searches. The job for twenty-odd years. Verdict: the job.
- GEO. Generative engine optimisation. Named by a November 2023 paper ("GEO: Generative Engine Optimization", Aggarwal and others, KDD 2024) that changed pages and measured whether AI answers used them more. Adding statistics, quotes and sources, and writing more clearly, got up to 40% more visibility in the AI's answers in their tests. Verdict: the same job, with a paper behind the name.
- AEO. Answer engine optimisation. The older word for the same idea: being the one answer that gets read out rather than one of ten links. Perplexity calls itself an answer engine. Verdict: the same job.
- LLM SEO, LLMO. The same job again, pointed at ChatGPT and the chatbots. A large language model is the thing behind the chatbot. Verdict: the same job.
- AI SEO. Two different meanings: using AI tools to do your SEO, or doing SEO so that AI search finds you. Ask which one is meant. Verdict: ask.
- AI visibility. How often you're cited or mentioned in AI answers. Verdict: the new bit, worth measuring.
One page, three labels
The list a GEO agency, an AEO agency and a 2015 SEO will each give you for the same page:
- The question in the heading, in the words a person would use.
- A plain answer at the top, in a sentence or two, before the detail.
- A date and a name on the page.
- Numbers with their source.
- Proper headings, so the sections are obvious.
- Fast to load, and readable by Google (indexable, no blocked scripts hiding the words).
- Structured data that matches what is on the page.
- Linked from inside your own site.
- Mentioned in places you don't own: forums, videos, review sites.
Google's own page on AI features (updated 10 Dec 2025): the best practices for SEO remain relevant for AI features, and you don't need to create new machine-readable files, AI text files or markup to appear in them.
What's actually new
How you're measured. Count citations (your pages shown as sources inside AI Overviews, AI Mode, ChatGPT and Perplexity) and mentions (your name in an answer, link or no link), as well as rankings and clicks. Search Console's Generative AI performance report (Search) gives the impressions for AI Overviews and AI Mode free, for every site, since 31 Aug 2026; no clicks or queries in it yet.
Where some of the work happens. AI answers lean on forum threads, videos and review sites as well as websites, so part of the job is being talked about in places you don't own.
The new invoice
llms.txt. A text file meant to tell AI companies what is on your site. Ahrefs checked 137,210 sites in May 2026: 28% had one, and 97% of those files got no requests at all. Google has said more than once that it doesn't use them.
"AI schema". Doesn't exist. Structured data is real and useful, and it is the same structured data it has always been.
"We'll rank you in ChatGPT." Nobody can promise a position in a system that re-picks its sources every time the model is updated. Paying for mentions in forums comes under the same heading.
Every number in the episode
| Number | What it is | Where it came from, and when |
|---|---|---|
| Up to 40% | More visibility in AI answers for pages changed the paper's way | "GEO: Generative Engine Optimization", Aggarwal and others, arXiv, 16 Nov 2023 (KDD 2024) |
| 7,800 / 1,400 | Searches a month for "generative engine optimization", US / UK | Ahrefs Keywords Explorer, 31 Oct 2026 |
| 17,048 → 7,168 | US searches a month for it, July 2025 peak and October 2026 | Ahrefs Keywords Explorer, 31 Oct 2026 |
| 7,000 / 900; 461 → 8,288 | Searches a month for "answer engine optimization", US / UK; US in October 2024 and October 2026 | Ahrefs Keywords Explorer, 31 Oct 2026 |
| 3,100 / 800 | Searches a month for "what is geo", US / UK | Ahrefs Keywords Explorer, 31 Oct 2026 |
| 76% | AI Overview citations already in the organic top ten | Semrush, 2026 |
| 37.1% | AI-cited pages ranking top ten for the same query | Ahrefs, March 2026 |
| 31 Aug 2026 | Search Console's Generative AI performance report (Search) available to all websites | Google Search Console Help |
| 137,210; 28%; 97% | Sites checked; share with an llms.txt file; share of those files with no requests in May 2026 | Ahrefs, 15 Jun 2026 |
Next episode
How an AI answer picks its sources: the fan-out, which pages it reads, and why being quoted isn't quite the same thing as ranking.
Jamie runs Sequentis, an SEO and AI-search agency.
written by
Technical SEO consultant and full-stack developer in Hersham, Surrey, in practice since 2007. One person, no handoffs: the audits, the code and the writing come from the same pair of hands. This site is the working proof — it grades itself on the same instrument it runs for clients.