What we check
The four checkable reasons AI assistants leave a site out
Each of these is something we can point at on your own site, not a ranking theory nobody can verify. A scan checks all four every time.
01
Things you do that have no page of their own
Your copy names a service, a location, a capability. Nothing on the site has a URL, a title and a heading about it. There's nothing to cite, so you don't get cited.
This shows up most on multi-location and multi-service businesses, where a branch or a treatment gets a line in the footer and nothing else. A page that exists only as a mention inside another page's paragraph gives a crawler no title, no heading and no URL to point at, so there is nothing for a model to retrieve even though the fact itself is true and stated somewhere on the site. The fix is not always a brand new page: sometimes a thin page just needs enough real content to stop reading as boilerplate.
02
The same fact, stated two different ways
Two sets of opening hours. Two turnaround figures. Two phone numbers. Faced with a conflict it can't settle, an assistant usually drops the fact rather than pick one.
We see this constantly on turnaround times and opening hours, where a hero section says one thing and an FAQ two scrolls down says another. A model extracting facts from your site has no way to know which of two contradictory statements is current, so the safer move for it is to drop the fact rather than risk repeating the wrong one. We normalise units before comparing, so "1 to 3 working days" and "24 to 48 hours" get caught as the same kind of fact stated two incompatible ways, not treated as unrelated text.
03
Proof buried where nothing reads it
Accreditations, certifications and registrations sit in a paragraph halfway down a page, absent from every title, heading and meta description you have.
A genuine advantage, a faster turnaround, a rarer accreditation, a lower price, does nothing for you sitting only in a price table or a structured data block nobody reads in prose. If it never appears in a title, a heading or the actual sentence a visitor reads, a language model has no text to lift when it is deciding who to mention. This is usually the cheapest fix in the whole report: the fact is already true and already on the site, it just needs saying somewhere a model will actually look.
04
Answers that aren't marked as answers
Content that answers a real question with no schema saying so, and no structured line anywhere declaring what kind of organisation you actually are.
Some of this is structural: no FAQPage schema, no llms.txt, AI crawlers blocked outright in robots.txt. Some of it is quieter, an answer that only makes sense once you have already read the paragraph above it, which is fine for a human scrolling the page and useless for a model that lifts one self-contained chunk at a time. We check the mechanical blockers first, since a blocked crawler makes every other finding moot for that site.
Frequently asked questions
What is generative engine optimisation?
Generative engine optimisation is the practice of making a website straightforward for AI systems such as ChatGPT, Claude and Perplexity to read, understand and cite correctly. It covers the same underlying facts search engines have always needed, structured so a language model can extract them without guessing. When it works, an AI assistant answering a question about your business can find you and quote you accurately.
Is GEO the same thing as SEO?
No, though the two overlap. SEO is mostly about ranking in a list of links, while GEO is about whether an AI assistant can read your site at all and extract facts it trusts enough to repeat. A site can rank well in a search engine and still be invisible to an AI assistant if a bot blocker turns crawlers away or its facts contradict each other across pages.
Will adding schema markup get my site into AI answers?
No. Structured data removes a specific obstacle, it does not buy you a mention. It gives an AI system a clean way to read a fact it would otherwise have to guess at from prose. Anyone who promises you will appear in AI answers because of markup alone is guessing, not reporting a result they can evidence.
Do you change anything on my site?
No. We read your site publicly, the way a search engine or an AI crawler does, and we never write to it or ask for credentials. Every fix we generate is shown to you first, and you decide whether and where to paste it in.
How many pages do you read?
A free scan reads up to 25 pages, sampled from your sitemap where one exists. We state plainly how many pages that is out of your site's real total, so a small sample on a large site reads as a sample, not as a complete picture.
Do you need access to my analytics or my CMS?
No. We never ask for a login, an API key or write access to your content management system. Everything we check comes from the public pages of your site, the same as any visitor or crawler would see.