The thesis
Your best content is already in your experts' heads.
The content that earns AI citations is first-hand operator expertise, and most never gets written down. Recording the people who do the work, on video or voice, turns what they know into transcripts, articles, and clips a model can read. The expert talks for ten minutes; the system does the rest. The bottleneck is capture, and capture is solvable.
The scarce input is the expertise, and it already exists inside your company.
01
The expertise is already in the building
I trained as an engineer before I moved into marketing, and the habit that came with me is a bias toward the source. When something breaks, you do not ask the person who wrote the ticket. You ask the person whose hands were on the system. The same instinct applies to content. The most quotable thing your company could publish is what your senior operator would tell a client across a table, unscripted, when they explain why the obvious approach fails and what they do instead. A blog post assembled from ten other blog posts cannot match it. It is the same thread I picked up in my last column: before you can win the citation, you need something worth being cited for.
That knowledge rarely reaches the web. Not because anyone is hiding it, but because the path from "knows it cold" to "wrote it down" is longer than most experts will ever walk. They are busy doing the work. Writing is a separate skill they did not sign up for, and a blank document asks them to perform it alone. So the expertise stays where it started, in conversation, in the review call, in the answer they give for the hundredth time and never record. The bottleneck was never a shortage of things to say. It is that saying and publishing are two different acts, and we keep asking the expert to do both.
Your most quotable content is a conversation your expert has already had a hundred times.
Try thisThink of the one person on your team whose answer you would trust over any article. Now count how many words of theirs are actually published on your site. For most teams the number is close to zero, and that gap is the whole opportunity.
02
Why does first-hand expertise win in AI search?
First-hand expertise wins because both search engines and AI models are built to reward it, and they say so plainly. Google's own guidance on helpful content asks creators to check one thing directly: "Does your content clearly demonstrate first-hand expertise and a depth of knowledge (for example, expertise that comes from having actually used a product or service, or visiting a place)?" (Google Search Central). The extra E that Google added to E-A-T stands for Experience, and it points at exactly the knowledge your operators hold and your competitors' content usually fakes.
This matters more for AI answers than it did for a list of links. A model is trying to assemble a trustworthy answer from text, and text that reads as lived, specific, and sourced is easier to trust and easier to quote. Generic content that could have been written by anyone about anything gives a model nothing to hold onto. A sentence that only a practitioner could have written, the tradeoff they learned the hard way, the number they actually see in the field, is the kind of thing an answer engine reaches for. If you want to be quoted, publish the parts no one without the experience could have said. That is also the standard Innovative Group holds its own AEO work to, and the reason senior, cross-channel content is a program a fractional marketing leader is built to run rather than a task you delegate to whoever has spare time.
Publish the sentence only your expert could have written. That is the one a model quotes.
Try thisTake one page you are proud of and highlight every sentence that could only have come from real experience. If most of the page survives, it is quotable. If the highlighter stays capped, you have written something a model has already read a thousand times.
03
Why is video the fastest way to capture expert knowledge?
Video is the fastest capture method because talking is the format experts already fluently use, and audiences prefer to receive it that way too. Wyzowl's 2026 survey found that when people are asked how they would most like to learn about a product or service, 63% choose a short video, far ahead of the 12% who pick a text article (Wyzowl, 2026). The demand side already leans toward the spoken explanation. The supply side does too, because an expert who will never write a thousand words will happily talk for ten minutes about the thing they know best.
The move is to stop treating video as the finished deliverable and start treating it as the capture device. Sit the expert down, ask the questions a client actually asks, and record the answers. What you get is raw material: a transcript you can edit into a first-hand article, quotes you can lift, clips you can cut, and answers you can shape into the questions buyers really type. You have inverted the usual order. Instead of an expert struggling to author text, you have a system turning their natural speech into every text format you need. The published pieces then feed the Insights library that answer engines read.
Record the expert once. Harvest the article, the clips, and the answers from the transcript.
Try thisBook one thirty-minute call with your best operator and record it. Do not brief them to prepare a talk. Ask them the ten questions your buyers ask most and let them answer as if you were a client. That recording is more usable raw content than a month of asking them to "write something."
04
How do you turn one recording into content an AI can read?
You turn a recording into AI-readable content through a short pipeline: record, transcribe, edit, structure, publish. An answer engine cannot watch a video. It reads the text around and beneath it, so the transcript is the asset, and the video is what earns attention and trust once the text has done the indexing work. Skip the transcript and you have published something a model cannot parse. Include it, cleaned and structured, and you have handed the engine first-hand expertise in the one form it can actually quote.
Here the engineering discipline earns its place. I treat the raw transcript the way I once treated a raw test log: full of signal, not yet fit to ship. Editing is triage. You cut the throat-clearing, keep the sentences only the expert could have said, add the headings a reader and a crawler both need, and mark up the questions and answers so the structure is legible to a machine. Publishing is the fix that ships, and like any fix it is not done until it has been checked. The teams at Innovative Group's digital marketing and technology practice run this as a repeatable flow, so one recording reliably becomes a clean page, a set of clips, and schema that parses, rather than a file that sits on a drive until everyone forgets it exists.
The transcript is the asset. The video earns the click; the text earns the citation.
Try thisTake your last recorded call or webinar and pull the transcript. Read it looking for the three sentences you would be proud to have quoted back to you. Build a page around those three, add the question each one answers as a heading, and you have a citable asset from content you already owned.
05
Whose job is it to get the expertise out?
It is the job of whoever decides to own it, and the work compounds only when it stops depending on a single person's goodwill. Extracting expertise cannot be a favor an expert does when they feel inspired, because inspiration does not keep a schedule. It has to become a system: a standing cadence of short recordings, a defined path from transcript to published page, and a named owner who runs it whether or not anyone is watching. I care about this because I have watched good knowledge die on a drive, captured once and never shaped, and that waste bothers the engineer in me more than almost anything else.
There is a leadership version of the same point. The expert's time is expensive and their patience is finite, so the system has to protect both. Ask for ten minutes of talking, not ten hours of writing. Do the shaping downstream, where a content operator and increasingly a model can carry the load. Give the expert a finished piece to approve instead of a blank page to fill. Done well, this turns your scarcest people into your most-published voices without stealing the hours they need for the actual work. That is the kind of leverage a Next Best Action program is designed to create, and if no one on your team owns the flow yet, it is the sort of engagement we take on at IG.
Expertise that lives in one head is a single point of failure. Expertise on the record is leverage.
Try thisPut one recurring thirty-minute recording on the calendar with your top operator, monthly, and assign one person to turn each session into a published page. A cadence you can run beats a burst of inspiration you cannot repeat.
Ashlesha Khond, Innovative Group