AI search engine optimization
Your buyers stopped scrolling.
They ask, and get one answer.
That answer names three or four companies. There is no page two, no tenth place, and no click for being nearly included. Either an assistant has a reason to say your name or it does not, and the reason lives on pages you do not own.
AI SEO, GEO, AEO, LLM SEO. Same job, different names.
The vocabulary in this category is a mess, mostly because it is new. You will see all of these used for the same work, so here is what each one is reaching for.
- AI SEO
- The broad one. Any work aimed at getting a brand into the answers assistants give.
- AI search engine optimization
- The long form of the same thing, used when people mean search specifically rather than chat.
- Generative engine optimizationGEO
- The term the industry settled on for optimising toward systems that generate an answer instead of returning links.
- Answer engine optimizationAEO
- Emphasises the output: there is one answer, and the question is whether you are in it.
- LLM SEO
- Names the thing doing the reading. Same work, framed around the model rather than the search box.
- AI visibility
- The measurable side: whether, and how often, a brand actually appears.
Underneath the labels the job is identical: be accurately described, in the words buyers use, on the sources these systems already read.
From our own data
What 1,210 real AI answers actually look like
Most advice about AI search cites the same few third-party studies. We run the checks ourselves, so these are our numbers: 1,210 live answers to real buyer questions across 6 models, carrying 19,589 sources between them. Every figure below is read from the same transcripts you can open on this site.
Recalculated 8 October 2026+368 answers in the last 7 days, against 761 the week before
13%
of answers cited nothing at all. The model answered from memory, which is the hardest case to influence and the best argument for being described widely.
6,771
distinct websites were quoted. There is no small club of gatekeeper sites to buy your way onto.
6%
is all the 20 most-cited domains account for. The long tail is where the answers actually come from.
The models do not behave alike
Optimising for one and assuming the rest follow is the most expensive mistake in this category.
| Model | Answers | Sources read | Brands named | Named the subject |
|---|---|---|---|---|
| Grok | 198 | 57.4 | 8.1 | 23% |
| Perplexity | 202 | 22.1 | 5.6 | 20% |
| ChatGPT | 204 | 6.9 | 5.9 | 19% |
| Claude | 202 | 6.7 | 7.8 | 25% |
| DeepSeek | 202 | 3 | 5.1 | 14% |
| Gemini | 202 | 1.9 | 8 | 23% |
Sources read and brands named are per-answer averages. “Named the subject” is how often the model mentioned the company the check was run for.
The sites these models quote most
Counted once per answer. Note how many are communities and independent roundups rather than brand-owned pages.
- forbes.com×67
- reddit.com×65
- zapier.com×56
- youtube.com×54
- apps.apple.com×52
- en.wikipedia.org×48
- vertexaisearch.cloud.google.com×48
- play.google.com×43
- trustpilot.com×43
- pcmag.com×42
- linkedin.com×36
- goodreads.com×33
- nerdwallet.com×33
- amazon.com×32
- clutch.co×30
- techradar.com×30
- textverified.com×30
- melodica.ae×29
- facebook.com×28
- instagram.com×26
Categories measured so far
Every check adds to this. Open one to read what the models answered for it.
How it works
Measure it first. Everything else is guessing.
01
Find the questions your buyers actually ask
Not questions we invent. We pull the terms people really search in your category from Google Keyword Planner, with the monthly volume attached, and turn the highest-volume ones into the questions a buyer would type into an assistant.
02
Ask them live, to every model that matters
ChatGPT, Claude, Gemini, Perplexity, DeepSeek and Grok, with the web open, and we keep every answer word for word along with the sources each model read. No sampling, no paraphrasing, no dashboard between you and the raw answer.
03
Read what the answers are made of
Which companies got named, in what order, and which pages the models pulled from to decide. This is the part that tells you what to fix, because it is the only part you can act on.
04
Change the inputs
Get described, accurately, on the kinds of pages the models are already reading in your category. Then re-run the same questions and compare against the first measurement.
What moves it, and what people waste money on
Moves it
- Being described on third-party pages the models already quote in your category
- Independent roundups and comparison pages that list you next to the competitors you lose to
- A clear, consistent description of what you sell, in the words buyers use, wherever you are mentioned
- Community threads and reviews, which show up in citations far more than most brands expect
- A structured public record of your company, so an assistant asked your name by itself has something to read
Does not
- Buying a position in a model, which is not for sale from anybody at any price
- Keyword-stuffing your own homepage, which is the one source that cannot vouch for you
- Mass low-quality links, which was never what these systems are reading
- A monitoring dashboard on its own, which tells you the score without changing it
- Anything sold with a guaranteed AI citation attached, which is the clearest signal to walk away
Start where it costs nothing
Free
The check
3 real questions from your category, asked live, answers unedited. You see whether your name is in them and who is named instead.
Run it now$97
The full measurement
50 buyer questions across every model, with the list of questions where a competitor is named and you are not, ranked by how many people ask them.
See what is in it$999
The fix
Your story written and published on high-authority outlets, so there is something accurate for the models to read about you.
See the outletsYour story publishes on all 7 outlets, or you get your money back. We do not guarantee a position in ChatGPT or any other model. Nobody can. We measure where you stand, and we fix the layer you control.
Looking for an AI visibility tool?
Two of ours are free and need no account. The check asks live models real buyer questions from your category and shows you the unedited answer. The score reads a single page and tells you whether it is worth citing in the first place.
- The AI visibility check — does AI name you, across 6 models
- The AI visibility score — is this page cite-worthy
Looking for an AI SEO agency?
That is us, with one difference worth knowing before you talk to anyone else. We will not quote you a position in a model, because that is not a thing anyone can sell. We measure where you stand, we change what the models read, and we measure again so you can see whether it moved.
Talk to usFAQ
Frequently asked questions
What is AI search engine optimization?▾
It is the work of getting your company accurately described on the sources AI assistants read, so that when somebody asks about your category the answer has a reason to mention you. It is sometimes called generative engine optimization, GEO, or answer engine optimization. The names differ, the job does not.
How is it different from normal SEO?▾
Classic SEO competes for a position on a page of links, where being tenth still gets a click. An assistant returns one answer naming a handful of companies, so there is no tenth place. The other difference is where the signal lives: a search engine reads your site to rank your site, while an assistant reads everyone else's pages to decide who to mention.
Can you guarantee ChatGPT will mention my company?▾
No, and you should not buy from anyone who says otherwise. We do not guarantee a position in ChatGPT or any other model. Nobody can. We measure where you stand, and we fix the layer you control. What we guarantee is the work itself: Your story publishes on all 7 outlets, or you get your money back.
How do I know whether I have a problem?▾
Run the free check. It takes about a minute, asks real questions from your category to live models, and shows you the unedited answer with your name highlighted if it is there and the competitors named instead if it is not. No account, no card.
How long does it take to see a change?▾
Publication happens in days. The models picking it up is slower and depends on how often they recrawl your category, which is why we re-run the same questions rather than asking you to take our word for it. Anyone quoting you a precise date for a model update is guessing.
Do the models all behave the same way?▾
Not remotely, and this is the single most useful thing in our data. Across the answers we have collected, the average number of sources a model reads before answering ranges from 1.9 to 57.4 depending on which one you ask, and the rate at which a model names the company being checked varies by roughly a factor of two between the most and least generous. Optimising for one model and assuming the rest follow is a mistake.
Is there a small set of websites I need to be on?▾
No, and the data is blunt about it. Across every answer we have collected, the 20 most-cited domains account for 6% of all citations, spread across 6,771 distinct sites. There is no short list to buy your way onto. Breadth of accurate description beats a handful of placements.
What does it cost?▾
The check is free. A full measurement across 50 buyer questions and every model is $97. Getting described on high-authority outlets starts at $999 for a bundle, or per outlet from the catalogue if you would rather start with one.
Keep reading
- Generative engine optimization, explained with our data — what GEO is, and how far the six models disagree.
- Answer engine optimization — citation depth per engine, and how to write the page that gets lifted.
- The LLM SEO playbook — seven steps, each with the measured number behind it.
- Every question we have put to AI and what each model answered.
- What AI recommends, by category.
- Every company we have checked.
- Score whether a single page is cite-worthy, free.
Comparing us against the rest of the market?
We publish the whole category: 9 AI visibility tools ranked in one table, each with the price it publishes, three things the money buys, three things it does not, and the case for choosing it over us. Prices checked 4 October 2026 on each vendor's own site.