Your ICP profiles and internal data (e.g. forms, sales call transcripts, client interviews) are the foundation for structuring prompts that reflect how your target persona uses LLM platforms in their user journey.
You can use transcripts and written data to reveal the actual language your buyers use. That language should shape your prompts, not assumptions.
You should have a clear topical map of your website with assigned key and supporting pages. Identify which pages you expect to see cited for specific prompts.
Competitor benchmarking and basic technical health checks (status codes, page speed) come before AI search work, not after.
What this video covers
Search has changed a lot in the last year - probably one of the biggest shifts we've seen in SEO in the last decade. Buyers journey is spread across multiple platforms and using more modalities to find the vendors and products they need. That shift is exactly why the prep work below matters - before you can measure your brand's visibility in an AI search visibility audit, there are several inputs worth getting right.
Start with your ideal customer profile (ICP)
The first input isn't a prompt. It's your ICP profiles - specifically their pain points and how they map to your services/product features. Having that context is what lets you structure prompts that actually reflect how your buyers search, since most people base their AI search queries around the problem they have.
Pull real language from sales, not assumptions
Sales calls, notes, recordings, and form submissions are the next input - this is how you validate whether you're heading in the right direction.
This is where you find the actual wording your buyers use, and the terminology specific to your industry. That real language is what your prompts should be built around, so your monitoring reflects user intent and how people actually search, not how your team assumes they write.
Do a topical map of your website
If you don't already have a topical map of your site, start by selecting your key pages: homepage, about page, solution and product pages, strong evergreen guides, and your best-performing funnel content.
Use Search Console, SEMrush, Ahrefs, or GA4 to confirm which pages actually drive volume, then assign each one a specific topic and stage of the funnel. This step matters as it will enable you to identify the set of core pages that you expect to see in your audit results and verify if this is the case.
Competitor analysis: know exactly who you're benchmarking against
Pick three to six real competitors and analyse what works best for them and your industry: how their content performs, how their pages are structured, and what asset format they use most frequently - case studies, guides, and so on.
Start with a manual review of the site itself, the way a first-time user would - how does it feel to navigate, and where's the friction?
Then move to the SEO platform of choice (e.g. SEMrush/Ahrefs) to check their top-performing pages and ranking keywords.
A solid competitor analysis SEO process helps you understand whether other brands have stronger topic coverage, better UX, or more comprehensive content than your own site. If they do, it's often a sign that you need to strengthen your SEO and AEO basics, and website foundations, before going deeper into AI search visibility.
Crawl your website before your AI search audit
Crawl your site and get a clear picture of how your key pages are actually performing technically, confirm:
- Pages return a 200 status code
- They load fast
- Pages are accessible to AI bots
These may seem "too basic", but they matter more than people think and we see them being ignored more often than not. Technical issues on your key pages can hold back your results, not just in AI search and AI Overviews, but in traditional Google search too.
Fixing these SEO fundamentals before the audit means your results reflect your site's real potential, not technical blockers getting in the way.
Get these fundamentals in place, and your next AI search audit actually tells you something you can act on.