Citation baseline
Ask the major assistants the questions your buyers ask, record which sources they cite, and establish where you currently appear. Without this the rest is guesswork.

Norml works on whether an AI answer names you. That is a different problem from ranking: it rewards checkable claims, clean structure, and sources a model can attribute. Also called AI optimization or AIO, and the work is the same under any of the three labels.
A few clients we’ve worked with










Prospects arrive already informed, and the information did not come from your site.
Asking an assistant about your category returns your competitors and not you.
Your best evidence is in a PDF, a deck, or a video, where nothing can read it.
The pages make claims that cannot be attributed, so nothing on them is safe to quote.
Nobody is measuring whether any of this is changing.
We start by asking the assistants about your category and recording what they say. Then the work splits into what the pages need to become citable, and what needs to exist elsewhere for a model to corroborate it.
Ask the major assistants the questions your buyers ask, record which sources they cite, and establish where you currently appear. Without this the rest is guesswork.
Turn what you know into something quotable: a measured figure, a stated method, a dated comparison. Models cite specifics and skip adjectives.
Clear headings, direct answers near the question, structured data, and content out of images and PDFs and into text.
A claim a model can verify in more than one place is one it will repeat. This overlaps with authority work, which is why it is scoped alongside link building.
Deciding which AI crawlers may read the site and confirming they can, rather than discovering a blanket block after months of invisibility.
Recurring prompt sets, logged over time, so the change is visible rather than asserted.
The agreed scope identifies which of these deliverables your project needs.
A citation baseline across the major assistants, with the prompts used.
A prioritized list of pages to restructure, with the reason for each.
Original, dated evidence assets a model can attribute to you.
Structured data and answer formatting implemented on the priority pages.
A crawler access decision, documented rather than left to a default.
A recurring prompt-tracking report.
Our process
Record what the assistants currently say about your category and who they cite. You see the starting position before anything is spent.
[ 01 ]Decide what you can say that is specific, true, and checkable, and what needs measuring first.
[ 02 ]Make the priority pages readable and quotable, with answers where a model will find them.
[ 03 ]Get the claims referenced elsewhere, so a model has more than one source.
[ 04 ]Re-run the prompt set on a schedule and report the movement.
[ 05 ]