Getting your brand cited when ChatGPT, Claude, Perplexity or Google answer on your behalf. It can be measured, which is why it can be improved.
More and more searches end without a click
A growing share of the questions that used to land on your site are now answered inside the search engine itself, or inside ChatGPT, Claude, Perplexity or Gemini. The user reads the answer and doesn’t come in. If your brand isn’t cited in that answer, for that person you don’t exist — even if you’re still on page one of Google.
You can be on Google's first page and not exist for someone asking an assistant.
This isn’t the end of SEO, but it does change what you optimise for. Ranking a page is no longer enough: you have to be the source the model chooses to cite when it summarises the topic.
What we do
Measure where you appear today. We put the queries that matter in your category to the main assistants and record which brands they cite, from which sources, and in what order. Without that baseline, any later improvement is an opinion.
Content that can be cited. Models cite what is specific, attributable and easy to extract: a figure with its source and date, a clear definition, an honest comparison. Generic agency prose never gets cited, because it adds nothing that isn’t already on twenty other sites.
Structure machines can read. Structured data, headings that match real questions, answers that start with the answer, and a site that can be crawled without running JavaScript to see the content.
Presence beyond your own site. Much of what models cite isn’t your domain: it’s trade press, forums, directories and third-party documentation. We work that presence with the same rigour as link building, but looking at which sources the assistants actually consult.
Control over what is said about you. Sometimes the problem isn’t absence but error: a model repeating an outdated figure or confusing your product with someone else’s. It can be corrected, and it starts with fixing the source it came from.
What we don’t promise
Nobody controls what a model answers. There is no console where you push a position up, and no magic format that guarantees a citation; anyone telling you otherwise is selling smoke. What can be done is systematically raise the probability of appearing, measure it, and check whether it goes up.
We also don’t recommend abandoning classic SEO. The two feed each other: assistants lean heavily on what is already well ranked and well linked.
What we measure
Share of sector queries where you are cited, position within the answer, which of your pages get used as a source, and how all of that moves month to month. Alongside it we track referral traffic from the assistants, which is now an identifiable source in analytics.
Why us
We work with these models daily, not just as an object of study. We know how they retrieve information, what makes them prefer one source, and why they repeat errors, because we build things on top of them.
And we have been doing SEO since before any of this existed, which avoids the commonest mistake in this area: treating it as a new, separate discipline instead of what it is — one more layer on work that is already well understood.
