Getting cited by the AI, not just ranked by the search engine.
People increasingly ask an assistant instead of searching. Being invisible to ChatGPT, Claude and Perplexity is a new category of problem, and it has a different fix from classical SEO.
A ranked list of ten links is not what your buyer is looking at any more.
Classical SEO optimises for position in a list. Generative engines don't produce a list, they produce an answer, synthesised from a handful of sources, with one or two citations. Being eleventh has always been bad. Being the fourth-best source for an AI answer is worse, because there is no fourth slot.
What determines whether you're the source a model reaches for is different from what determines ranking. Machine-readable structure matters enormously. Unambiguous entity data, who you are, where you are, what you do, who runs you, matters more than keyword placement. Specific, checkable claims get quoted; marketing adjectives don't.
And there's a prerequisite most sites fail without knowing: the AI crawlers have to be allowed in. We check this first on every audit, and it is wrong more often than anyone expects, usually because a CDN's managed bot rules block them by default and nobody looked.
This is not hypothetical. We recently checked a UAE agency that sells AI services and found their robots.txt blocking ClaudeBot, GPTBot, Google-Extended and Applebot-Extended outright. They cannot be cited by any of the assistants their own clients are asking about, and as far as we can tell nobody there knows.
How it works, step by step.
Crawler access audit
robots.txt, CDN managed rules, firewall bot rules, and anything silently overriding what you think you've configured. This is first because everything else is wasted if the crawlers can't read the page.
Entity resolution
Making it unambiguous who you are. Organization, Person, address, identifiers, and consistent references across the web that a model can reconcile into one entity.
Structured data
Organization, Person, Service, FAQPage, BreadcrumbList, HowTo where it genuinely applies. Implemented properly, validated, and not contradicting the visible page.
Content built to be quoted
Specific, checkable claims replacing marketing language. Honest limitations included, because 'what can't X do' is a question models get asked constantly and almost nobody answers.
Classical SEO alongside
Titles, metas, internal linking, performance, sitemap hygiene. GEO doesn't replace this; it runs next to it.
Measurement
A repeatable score, tracked over time, from a baseline taken before anything changes.
Four things people assume this is.
SEO with a new name
They overlap but the optimisation targets differ. Ranking optimises for position in a list; GEO optimises for being quoted in a synthesised answer where there's no list.
Keyword stuffing for robots
Repeating 'AI automation Dubai' unnaturally is the weakest pattern in this market and correlates with every other bad signal. Models are unusually good at recognising it.
A guarantee of citation
Nobody controls what a model quotes. What's controllable is whether you're eligible, structured, specific and accessible, which most competitors aren't.
A one-off project
Assistants re-crawl, models get retrained, competitors improve. This is maintained, or it decays.
Concretely, what's delivered.
- AI-crawler access audit, robots.txt, CDN and firewall rules, including what's silently overriding your config
- Entity clarity work so assistants can resolve who you are without guessing
- Full structured data implementation, validated, consistent with the visible page
- Content rewritten to be quotable, specific claims, honest limits, real numbers
- Classical SEO fundamentals alongside, titles, metas, internal links, performance, sitemap
- Baseline and tracked score, measured repeatably rather than asserted
- Competitor comparison on the same measures, so you know where you actually stand
- Ongoing maintenance, because this decays if left
Why we can say this.
We run Ranqr on this site before we sell it to anyone. Anqor's own GEO score moved from 42 to 75, and the audit log is published on the site rather than described in a case study. That's the evidence we'd rather point at than a testimonial.
This page is itself an example of the approach: named entities, explicit structure, specific claims, stated limitations, and a robots.txt that allows every major AI crawler by name.
What this won't do.
- We cannot guarantee a model cites you. Anyone promising that is selling something they don't control.
- It's slower than paid acquisition. If you need leads this month, this is the wrong line item, run ads and do this in parallel.
- It can't fix a site with nothing to say. Structure makes real substance findable; it doesn't manufacture substance.
- Results are hard to attribute precisely. Assistant referrals often arrive with no referrer at all, so measurement is directional rather than exact, we'd rather tell you that than present a confident number we can't stand behind.
Straight answers.
Is GEO just SEO with a new name?
No, though they overlap. Classical SEO optimises for a ranked list of links. GEO optimises for being the source a model quotes in a synthesised answer, where there's no list and often only one or two citations. Structure and specificity matter far more; keyword density matters far less.
Can you prove it works?
On our own site, yes, the GEO score went from 42 to 75 with the audit log published. For client work we measure the same way from a baseline taken before anything changes. What we won't do is claim credit for a citation we can't attribute.
What's the most common mistake you find?
Blocking the AI crawlers by accident, almost always through a CDN's managed bot rules rather than a deliberate choice. It's the first thing we check and it's wrong surprisingly often, including on sites belonging to agencies selling AI services.
How long before we see anything?
Structural changes are picked up over weeks, not days, and citation behaviour shifts more slowly still. Expect a meaningful read at three months, not three weeks.
Do we still need normal SEO?
Yes. Most search volume is still classical search, and the two share fundamentals, crawlability, structure, real content. We do both rather than treating GEO as a replacement.
Will you tell us if our site is already fine?
Yes. The audit either finds problems worth fixing or it doesn't, and manufacturing work isn't a business model we're interested in.
Anqor is a small studio, so you deal with the person who builds it. If this isn't the right fit for your business, I'd rather say so on the first call than sell you something you don't need.
LinkedIn →We'll audit your crawler access for free.
It's the single check that most often comes back broken, and it takes minutes. If nothing's wrong we'll tell you that.