Of the top 100 organic pages across 10 sites, 49 received zero traffic from LLMs. That comes from a May 2026 Search Engine Land analysis of roughly 150,000 pages. Half the pages winning in Google were invisible in AI answers.
That is the honest case for treating this as new work. It is also the reason to be careful about what you stop doing.
You are probably being sold a GEO retainer right now. Some of it renames work you already fund. Some of it is real and new. A small part of it can cost you Google traffic.
What do AEO, GEO and LLMO actually mean?
Three labels for overlapping work, and the differences are mostly marketing.
Answer engine optimization (AEO) aims at being the answer a system returns, rather than a link in a list. It predates ChatGPT and originally covered featured snippets and voice answers.
Generative engine optimization (GEO) aims at being named and cited inside answers generated by models, in ChatGPT, Perplexity, Google AI Mode and Copilot.
LLM optimization (LLMO) is a newer name for the same job as GEO. The label changes. The work does not.
Nobody owns these definitions. Agencies use them interchangeably. Judge the work listed. Ignore the acronym on the cover.
What carries over from SEO?
Most of the technical foundation, and all of the content quality.
Crawlability still decides whether a system can fetch your page. Indexation still matters, because Google AI Overviews retrieve from the live Google index. Rankings still correlate with being cited: Seer Interactive found Google page-one presence correlated around 0.65 with brand mentions in LLM answers.
Site speed, clean URLs, internal links, real content from a named author. None of that stops mattering.
Watch for the second invoice. A proposal that treats AI search as its own discipline, with its own budget, often sells back work you already pay for.
What is genuinely different?
Four things, and each one has a test you can run yourself.
The crawlers do not render JavaScript. A Vercel and MERJ analysis found none of the major AI crawlers execute JavaScript. ChatGPT’s crawler fetches JavaScript files and runs none of them. Googlebot renders. A page that ranks fine in Google can be blank to an answer engine.
The answer gets assembled from other people’s pages. Muck Rack found 84% of citations in AI answers point at earned media rather than the brand’s own site. Ranking is something you own. Citation is something you earn elsewhere. Your blog is not the lever here.
Retrieval and training access are separate decisions. You can block GPTBot from training on your content and still be cited by OAI-SearchBot. Most sites that lose retrieval access never decided to. Cloudflare began blocking AI crawlers by default on new domains in July 2025, and has since split the controls into search, agent and training categories.
The results move on their own. SparkToro found fewer than one in a hundred answer pairs name the same brands. Rank tracking assumes a stable result. Presence tracking assumes nothing, which is why you count across runs instead.
What does this cost you in Google?
Clicks, and some conversion quality. Both are measurable, and most proposals skip them.
AI answers reduce clicks on the page they appear on. Pew Research found users clicked a search result on 8% of visits where a Google AI summary appeared, against 15% without one. Being cited inside the answer helps: Seer measured 2.2 times the organic click-through for brands cited in an AI Overview.
The traffic also converts differently. Kaiser and Schulze found organic search referrals convert about 13% better than ChatGPT referrals.
Then there are direct trades. Rewrite a title tag to match a broader AI query and you can lose the narrower query you rank for today. Name that cost in the same sentence as the recommendation. Skip it and you will explain an organic dip in January.
Do you need both?
Yes, and the useful question is which one you are failing.
Answer three questions in order.
- Can retrieval crawlers reach your pages? Check robots.txt and your CDN. If the answer is no, nothing downstream is measurable.
- Is your primary claim in the initial HTML? View source and search for your headline sentence.
- Would a competitor’s homepage survive your sentences unchanged? If yes, the engines have no basis to pick you.
Fail one or two and you have a technical problem. An SEO contractor fixes it this month. Fail three and you have a positioning problem. More publishing will not move it.
The fix order runs the same way every time: access, rendering, the claim, then presence on the sites the engines already cite.
Which fixes serve both channels?
Four, and they are the ones to get approved first.
- Original media. Your screenshots, diagrams and data in place of stock images.
- A named author with a real bio, a resolvable author page, and Person markup.
- Methodology shown on the page rather than asserted, with dates and sample sizes.
- Answers in the first sentence under each question-shaped heading.
Each one helps Google and helps retrieval. Most of the rest of the list does not. That makes these four the cheapest way to get a content budget approved by someone who only measures organic.
Skip the tactics that have been tested and failed. llms.txt shows no measurable relationship to citations across about 300,000 domains. Adding schema to 1,885 pages produced no citation uplift and a small decline in AI Overviews. Blocking Google-Extended does not remove you from AI Overviews, because those retrieve from the live Googlebot index.
Ask this before you sign the retainer
Put one question to any agency pitching GEO, AEO or LLMO. Which of the three conditions are we failing, and what evidence says so?
A proposal that starts with deliverables rather than a diagnosis is a guess with an invoice attached.
The audit answers the question with evidence. It checks retrieval access and rendering, scores your claims against quoted competitor sentences, and runs 150 observations across ChatGPT, Google AI Mode and Perplexity. You get a fix list in order, with the organic cost named where one exists.
Scott King audits positioning and AI visibility for B2B technology companies. His audit tests whether a competitor could publish your positioning word for word and counts AI presence out of five, never as a rank. He builds AI marketing systems.