Interesting discussion with a client today.
“We want to know how many times we’re cited in AI-generated answers. We want to know our share of voice (SOV).
They’d paid for a new SEO/AI-visibility tool to uncover hundreds of prompt variations, hoping to understand how users might ask questions about their services.
But this approach creates a critical 'Authority Gap.'
To bridge it, you must move beyond winning the click and focus on winning a high 'Citable Authority Score' from the models themselves.
You see, understanding how the prompt is asked and then designing a strategy to address that won’t actually get you cited by the LLMs.
Think about it, there are almost 350 million people in the United States. That’s 350 million ways to ask the same question.
To be cited by the LLMs, you must understand several other things, too:
Who is your ideal customer profile (ICP), and what is their intent?
Which LLMs are you focusing on, and are your brand signals up to date?
How an LLM will search for and verify its sources before generating an answer.
That last one is important, and I think SEOs are misusing these new tools.
Let’s dig into our situation a bit more.
My client wanted to know which prompts they should create content for.
They’d given the tool a topic (one of their services) and asked for a list of ways a user may prompt an LLM when they needed that service.
The tool did its job and predicted hundreds of prompt variations. They were all top-of-funnel, informational intent questions.
What is the best…
Who is the top…
The problem here is this: LLMs answer these questions without you having to produce this type of content.
To cite you, an LLM will ‘fan out’ and verify your entity as a 'safety lock' for its logic.
For agents performing tasks, this verification prevents operational failures and ensures they are grounding their actions in a high-trust 'Truth Set' rather than guesswork.
So, if you’re producing top-of-funnel content on your site, hoping that LLMs will use you as a source, you’re misunderstanding how this works.
You want to prove that you’re the trusted expert when it comes to your product or service, and you want to distribute that information across channels.
That means:
Writing detailed, proprietary content about your product or service, and ensuring that it’s available and accessible to LLMs on your website.
Sharing the information on social media.
Having your subject matter experts appear in videos and podcasts, providing expert insights on the topic, with transcripts.
Appearing in third-party, industry publications, and other trusted publications.
Showing up in review aggregator sites, hopefully with great (not just good) reviews.
Appearing in industry databases.
Most importantly, you must ensure your site moves beyond visual design to include machine-readable protocols like schema markup, so agents can actually 'read' your data as a source of truth.
Your website should be the source of truth.
Your truths should be spread across the internet intentionally so LLMs can verify you…
…so they can trust you…
…so that they’ll cite you.
—
If you want to understand where your brand actually stands in AI Search, my audits break this down with complete clarity: structure, signals, and opportunities.
More on that in future issues.