See whether the assistants name your brand when buyers ask about your category, which pages they cite, and what to change when the answer belongs to someone else.
LLM SEO is the work of getting a large language model to name and cite your site when someone asks it a question in your category. The difference from classic SEO is not the tactics, which overlap more than the vocabulary suggests. It is the measurement. There is no numbered results page to check, no index you can query, and no console reporting impressions. The only way to know where you stand is to ask the models real questions and read what comes back.
Three acronyms are circling the same problem, and they are used loosely enough that it is worth being specific about which one you mean:
| Term | What it covers | Where you see the result |
|---|---|---|
| LLM SEO | Being named and cited by the assistants themselves | ChatGPT, Gemini, Perplexity, Claude |
| GEO Generative engine optimization |
The broadest of the three: any generative answer surface, including search results that carry one | The above, plus Google AI Overviews |
| AEO Answer engine optimization |
Being the answer to a specific question. Predates the current models and applies to featured snippets too | Snippets, voice answers, AI answers |
In practice the work overlaps almost completely, because all three reward the same things: pages a crawler is allowed to fetch, claims stated plainly near the question they answer, and corroboration on sites the model already reads. What differs is the surface you measure, and that is what decides which tool you need.
Each one answers a different question, and they are meant to be run in roughly this order: find the prompts, measure where you stand, see who is being cited instead, watch it over time, and fix the pages.
The starting point, because you cannot measure a surface until you know what is being asked. It surfaces the real questions buyers put to assistants in your category, and shows whether your brand appears in the answers to each one. The output is a list of prompts worth caring about, split into the ones you already win and the ones you never enter.
Your score out of 100, with the evidence next to it. It sends real category questions, ones that never mention you by name, to every configured engine, then reports how often you were named, how early in the answer you appeared, and what sentiment came with the mention. The bands run from Invisible at the bottom to Dominant at the top, and a zero is reported plainly as a real finding rather than smoothed into a low number.
Mentions and citations are not the same thing, and the gap between them is where the traffic goes. This shows which engines cite your domain as a source, on which answers, and which domains are being cited in your place. That list of rival sources doubles as a placement target list: those are the pages already trusted for your questions.
The closest honest equivalent to a rank tracker for an assistant. It re-runs a fixed set of prompts through the live engines, reports your win rate per prompt and per engine, and shows the trend, so a slip is something you are told about rather than something you notice months later. There is no numbered position to track, so it tracks presence and where in the answer you land.
The page-level half, and it is free and needs no account. It reads a URL the way an answer engine does and scores the signals that decide whether a page can be cited at all: whether the crawler is allowed in, whether the body survives without JavaScript, whether the claims are structured, and whether an llms.txt exists. It returns a ranked fix list with the weight behind each item, so you know which change is worth making first.
Two more free tools sit alongside it and answer narrower questions: the AI Visibility Checker and the AI Citation Audit. All three run without an account.
Thirty days is enough to find out where you stand and remove the blockers. It is not enough to build authority, so the goal here is a baseline and a shortlist, not a transformation.
Run the GEO Audit on your five most commercially important pages. Check that your robots.txt does not block GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot or PerplexityBot, which are separate agents that can be allowed or blocked independently. Confirm the body text is in the HTML the server returns rather than assembled afterwards by JavaScript. Fix anything here first: no amount of content work matters while a crawler is being turned away at the door.
Run Prompt and Question Research to get the real questions, then Visibility Overview to get the score and the evidence behind it. Write the number down. The point is not the score itself, which will be low for most sites, but having something to compare against in a month.
Run the AI Citation Tracker and read the list of domains cited for your questions. Some will be competitors. Many will be forums, roundups and independent reviews, which is the uncomfortable part: those are placements you earn rather than pages you publish.
Work down the audit's fix list by weight rather than by ease. Rewrite the pages that answer a tracked question so the answer sits in the first sentence after the heading, not four paragraphs in. Then put the prompts you care about into Prompt Tracking, so the next month measures itself.
Expect movement to be slow and uneven. You are waiting on a recrawl and then on corroboration elsewhere, so a month is a baseline and a direction, not a verdict.
No card required. The free tools need no account at all.
