Diagnostics
A fixed-scope diagnostic that shows you exactly what AI assistants say about your brand, who they name instead of you, and what is stopping them from citing you.
An AI visibility audit is a fixed-scope diagnostic that measures whether AI assistants can find, understand and cite your brand when a buyer asks them for a recommendation. Orphea Society runs these audits for companies in Dubai and across the UAE. We agree a prompt set with you, run it through ChatGPT, Perplexity, Gemini and Google AI Overviews, record which competitors and sources get cited in your place, then examine your own site for the technical and editorial reasons behind the gap. It ends with a report and a prioritised fix list. It is a one-off engagement, not a retainer.
We agree the questions before anything runs. Not brand-name lookups, which almost always work, but the buyer-intent prompts someone uses when they do not yet know you exist. The set is written down and reused every time the audit repeats, because a moving question set cannot be compared.
The same prompts go through ChatGPT, Perplexity, Gemini and Google AI Overviews. Each answer is captured as a screenshot rather than summarised, so you are reading what the engine actually said on the date it said it, not our paraphrase of it.
Who is named instead of you, and which sources those answers pull from. This is usually the part that changes the conversation, because the cited source is rarely a competitor website. It is a directory, a listicle or a forum thread you had never considered a channel.
Then we look at your own site for the reasons: entity clarity, structured data, whether your pages answer questions in a form a model can lift, and whether anything credible links to you at all. The output is a prioritised list, ordered by cost to fix against likely effect.
The common assumption is that AI assistants rank websites the way a search engine does, and that a brand missing from an answer must be ranking badly. That is not what happens. An assistant assembles an answer from sources it can read, trust and quote, and a great many well-built websites fail at least one of those three tests without failing anything a traditional SEO report would flag.
The first failure is readability. A page written as a brochure, where the value proposition is implied across a hero image and three headline fragments, gives a model nothing to lift. A page that states plainly what the company is, where it operates and who it serves gives it a sentence it can quote directly.
The second is entity confusion. If your company name, address and category appear in three slightly different forms across your site, your schema and your external profiles, a model has no confident way to decide that these are all the same organisation. It resolves the ambiguity by citing someone less ambiguous.
The third is absence of corroboration. Assistants lean heavily on sources that are not the brand itself: directories, comparison pages, industry roundups, forum threads. A company with an excellent website and no presence anywhere else has told the model exactly one story, from an interested party, with nothing to confirm it.
The audit measures a state on a date. It tells you what four engines said in response to a fixed set of prompts at the time they were asked, which sources those answers drew on, and which properties of your site plausibly explain the result. All of that is observable and reproducible, which is why the prompt set is agreed in advance and the answers are captured rather than summarised.
What it cannot do is guarantee a position. Nobody controls what a language model says, the answers vary between sessions and shift as the underlying models are updated, and any agency that promises you a fixed place in an AI answer is describing something that does not exist. The honest version of the goal is different: make your brand the easiest thing in your category for a model to read, verify and quote, then measure whether citation frequency moves over quarters.
This is also why we treat a single audit as a baseline rather than a verdict. One run tells you where you stand. Two runs, months apart with the same prompts, tell you whether anything you changed made a difference. The second number is the one worth having, and you cannot get it without the first.
The deliverable is deliberately short and structured the same way every time, so that two audits taken six months apart sit side by side without translation.
An audit is a diagnostic, and diagnostics are only worth buying when the result would change what you do. There are three situations where it would not.
If your site cannot be crawled at all, the audit will tell you that in one line and the rest of the report will be empty. Fix the indexation problem first, then measure. If you already know your technical foundation is broken, you are buying a confirmation, not information.
If nobody in your category is being cited yet, which still happens in narrow B2B niches, then the visibility map will be blank for everyone including your competitors, and the useful work is content and corroboration rather than measurement. We will say so at the scoping call rather than after the invoice.
And if you have no capacity to act on a fix list for the next two quarters, the audit will age. The engines change, the sources change, and a list of recommendations nobody executes is worth less than the meeting it took to explain it. It will still be there when you do have capacity, and it will cost the same to run then.
It is a diagnostic that answers one question: when someone asks an AI assistant for a recommendation in your category, does your brand appear, and if not, who does. We run an agreed set of prompts through ChatGPT, Perplexity, Gemini and Google AI Overviews, capture the answers, identify the sources those answers rely on, and then examine your site for the reasons you are absent from them.
An SEO audit asks whether Google can crawl, index and rank your pages. An AI visibility audit asks whether a language model can understand and quote them. The two overlap in the technical layer, because a page no crawler can reach is invisible to both, but they diverge after that. SEO rewards a page that earns the click. AI visibility rewards a page that can be lifted from, often without any click at all.
ChatGPT, Perplexity, Gemini and Google AI Overviews as standard. These four cover the assistants most commonly used by buyers in the UAE and they source their answers differently enough that a brand can be visible in one and absent from the others. If your buyers concentrate on a different tool, we add it to the set before the audit starts rather than after.
A report in three parts: where you currently appear across the prompt set with the captured answers as evidence, who is cited in your place and which sources those citations come from, and a prioritised fix list ordered by effort against likely effect. The prompt set is documented alongside it so the same audit can be re-run later and compared against the same baseline.
No. The audit is scoped and priced as a one-off piece of work and it ends with the report. Some clients take the fix list to their own team, some ask us to run the ongoing generative engine optimization work that follows from it, and some do nothing for a quarter and re-run the audit to see whether the picture moved on its own. All three are normal.
Let's talk about your goals and see if this is the right fit.