AI visibility is how often AI answer engines name your brand, and link to your website, when people ask them the questions your buyers ask. If someone asks ChatGPT "what is the best help desk for a small Shopify store" and the answer lists three of your competitors and not you, you have an AI visibility problem, whatever your Google ranking says. GEO (generative engine optimization) and AEO (answer engine optimization) are two names for the work of fixing it. This guide explains the terms, the metrics worth tracking, and a repeatable way to measure them.
What AI visibility means
Search used to return ten links and let the buyer choose. AI answer engines do the choosing. ChatGPT, Gemini and Google AI Overviews read the web, then write a short answer that names a handful of brands and cites a few sources. The buyer often stops there.
That changes what "being visible" means. There are two separate things to win:
- A mention: the answer names your brand, ideally near the top of a list.
- A citation: the answer links a page as its source. That page can be yours, or a review site, forum or video that talks about you.
You can be mentioned without being cited (the engine names you but links a comparison article), and cited without being mentioned (your blog post is a source, but the answer recommends someone else). Both matter, and they need different fixes.
GEO, AEO, AI SEO: are they different?
Not in any way that matters to a buyer of the service. You will see four labels:
- GEO, generative engine optimization: the term from academic work on how generative engines pick sources, now the most common label.
- AEO, answer engine optimization: older, from the featured-snippet era, now used for AI answers too.
- AI SEO and LLM SEO: agency shorthand for the same goal.
All four describe getting your brand named and cited inside AI-generated answers. The useful distinction is not between the labels but between GEO and classic SEO:
| Classic SEO | GEO / AEO | |
|---|---|---|
| What you compete for | A position in a list of links | A mention inside one written answer |
| Unit of measurement | Rank, clicks | Mention rate, citations, share of voice |
| How stable it is | Fairly stable day to day | Changes between runs of the same question |
| Who decides | One ranking algorithm per engine | Each engine, with its own sources |
The two overlap a lot. AI answers lean on pages that already rank and on review sites that already have authority. But you cannot read your AI visibility off a rank tracker.
The five metrics worth tracking
Most AI visibility tools report some version of these. Knowing what each one means keeps you from being sold a vanity number.
1. Mention rate
The share of answers, for a fixed set of buyer questions, that name your brand. If an engine answered 180 questions and 45 answers named you, your mention rate is 25%. This is the headline metric, and it is the one we publish in the AI Citation Index.
Because answers vary from run to run, a mention rate should come with a range. In our methodology every rate carries a 95% interval, and two brands whose ranges overlap are treated as roughly level.
2. Share of voice
Your mentions divided by all brand mentions in the same answers. It tells you how much of the conversation you hold against competitors, which mention rate alone does not.
3. Average position
In answers that are ranked lists, where you appear. Being named fifth in a list of five is not the same as being named first. Only count it on answers that actually are lists, and only when you appear in enough of them.
4. Citation share
The share of all cited links that point to your own sites. This is where engines differ the most. In our September probe, ChatGPT sent most of its citations to the brands' own websites, while Google AI Overviews and Gemini cited mostly third-party pages. We wrote up which sources AI answers cite in detail.
5. Cited sources around you
Not a single number but a list: which review sites, forums, videos and publishers the engines cite when they answer questions in your category. This list is the to-do list for GEO. If the answers keep citing two comparison sites and you are missing from both, that is the fix.
How to measure AI visibility, step by step
You can do a first measurement by hand in an afternoon. The method matters more than the tool.
Step 1: Write a prompt set that sounds like buyers
Write 20 to 50 questions in the words your buyers use, not your brand name. Mix four kinds:
- Recommendations: "best email marketing tool for a small online store"
- Problems: "how do I reduce support ticket volume"
- Comparisons and alternatives: "Mailchimp alternatives for creators"
- Prices: "how much does a help desk cost for a team of 5"
Our index uses about 200 per industry with roughly that mix (40% recommendations, 25% problems, 20% comparisons, 15% prices). Keep the set fixed, so changes over time come from the engines and not from you rewording the questions.
Step 2: Ask each engine, logged out, more than once
Personal history and settings change answers, so ask logged out or in a clean session, from the market you care about. Ask every question at least twice and ideally three times. One run is an anecdote: a brand can appear in one run and vanish in the next.
Step 3: Record mentions, positions and links
For each answer, write down every brand named, its place if the answer is a list, and every cited link. A spreadsheet with one row per answer is enough to start.
Step 4: Compute the metrics per engine
Calculate mention rate, share of voice and citation share for each engine separately. Averaging engines too early hides the most useful finding, which is usually that one engine names you and another does not.
Step 5: Repeat on a schedule
Weekly is a good cadence. Monthly is the minimum. AI answers move when engines update their models or their search layer, and you want to see the move, not hear about it from a prospect.
If you would rather not do this by hand, our free AI answers check runs 10 buyer questions per engine for your brand and one competitor, and emails you the result.
What the data says so far
At the end of September we ran a probe across seven industries to choose the first verticals for the index: 56 buyer questions in total, asked 3 times each on ChatGPT, Gemini and Google AI Overviews. A few findings are worth knowing before you start measuring:
- Answers name several brands. In help desk software, 83% of answers on Gemini and Google AI Overviews named at least three brands, with a median of four per answer. You are rarely competing for the only slot, but for one of four or five.
- Almost every answer cites something. Citation presence was between 91% and 100% in most industry and engine combinations. Sources matter everywhere.
- Engines agree more on brands than on sources. Top-five brand lists overlapped by 3 to 5 brands between engines, but the cited sites were very different. More on that in why AI engines disagree.
- Answers repeat more than you might fear. The overlap of brand sets between runs of the same question averaged 0.6 to 1.0 per engine (Jaccard). Stable enough to measure, noisy enough that one run is not a measurement.
The probe was small (8 questions per industry), so treat these as directions, not market statistics. The weekly index asks about 200 questions per industry.
How to improve AI visibility
Measuring comes first, but here is where the levers usually are:
- Get onto the pages the engines cite. If Google AI Overviews in your category keep citing two review sites and a Reddit thread, being listed and discussed there is worth more than another blog post on your own site.
- Answer buyer questions on your own site, plainly. Engines that cite brands' own pages (ChatGPT does this a lot) pick pages that answer the question directly: pricing pages, comparison pages, "best for" pages with clear headings.
- Be specific about who you are for. Prompts are sub-category questions: "for Shopify stores", "for teams of 5", "for bad credit". Pages that state the fit clearly are easier for an engine to match.
- Keep facts consistent. Names, prices and plan details that differ between your site, review sites and directories give the engines reasons to hedge or pick someone else.
- Measure again. GEO work without a fixed prompt set and a repeat measurement is guesswork.
FAQ
What is AI visibility in simple terms?
It is how often AI tools like ChatGPT and Google AI Overviews name your brand, and link to your site, when people ask them for a recommendation in your category.
Is GEO the same as SEO?
No, but they overlap. SEO competes for positions in a list of links. GEO competes for a mention inside one written answer, and is measured by mention rate and citations rather than rank. Good SEO helps GEO, because AI answers often lean on pages that already rank.
What is the difference between GEO and AEO?
In practice, nothing. Both describe getting your brand named and cited in AI-generated answers. GEO is the more common term today.
How do I measure my brand's AI visibility?
Ask a fixed set of buyer questions on each AI engine, logged out, several times. Record which brands are named and which sites are cited, then compute mention rate, share of voice and citation share per engine. Repeat weekly or monthly.
Which AI engines should I track?
Track the ones your buyers use. For most categories that means ChatGPT and Google AI Overviews first, then Gemini. Report each engine on its own, because they often disagree.
How often do AI answers change?
Often enough that one answer is not a measurement. Ask each question more than once, and re-measure at least monthly.
Can I get this measured for my brand?
Yes. Start with the free AI answers check, see where your category stands in the AI Citation Index, or browse our data charts. If you collect AI answers yourself, our ChatGPT, Gemini and SERP API endpoints return answers and citations as JSON.