Alex Yarosh Get Free Snapshot

Source record

@tjrobertson52 TikTok profile avatar

@tjrobertson52

2026-06-05

High-quality AI content needs a brand knowledge base, reusable instructions, revision memory, and fact checking.

Share source record
Tiktokexcerpt onlyen1

Source Text

i

The most effective way of getting recommended in AI search is to optimize and create new content on your website. This makes up about 70% of the work we do at our agency, and I like to think we've gotten pretty good at it.

However, sometimes when I explain this to prospective clients, they have one of two responses. They'll either say, oh, AI could never write our content. We need to write it ourselves.

Or they'll say, oh, you're just having AI write the content, right? We can do that with ChatGPT.

I'd like to tell you what I tell them and then give you the exact process we use to create content in our agency. We create content for some of the most complex industries, including immigration, attorneys, cyber security, finance, education. We'd yet to find an example where we couldn't get AI creating content that's of the same quality as they can create themselves.

But it's not easy. Take time. And if you have ChatGPT just spin up a bunch of content for you, it's very unlikely to rank, and if you're doing it at scale, it could get you penalized.

The first thing I understand is the topic matters a lot. If you're not writing on a topic that AI is actually searching for, or you're writing on topics that are way too competitive for your brand, you're wasting your time.

I have other videos going through how we pick topics, but in short, we're using Peak that AI to see which pages AI is citing, and then finding patterns in those topics and identifying which ones would be relevant for our client.

The main reason companies often feel like AI can't write their content, it has to be someone on their team. Is it? AI doesn't have the complete understanding of your company, and it doesn't have that tacit knowledge you get from experience.

And so, of course, you need to give that knowledge to the AI.

You need a knowledge base. This is the first thing we do for every client. We call it their brand ambassador, and we create this for them.

It's typically about a dozen markdown files, but it might also include Pdfs or spreadsheets.

You want to provide comprehensive information on your company, on each of your products and your services, your unique value proposition, your ideal customer profile. You want to understand every important page on your website and when to reference those page. Share information on your brand, voice what to say, what not to say.

As many examples as possible of things you've written. The canonical version of any guides or processes you've detailed.

If you're having it. Create images for your content. You want guidance on how to create images in your brand style.

Don't worry about collecting this all at once. Just collect as much as possible and then assume that you're gonna be iterating on this forever. I think two years from now, this is just gonna be a standing operating procedure.

Imitation like this is no longer for humans. It's for AI. And if humans need information, they can go to the AI.

Ideally, all this information only lives in one place, so you don't have to be updating multiple sources. As of right now, when we create this for our client, we store everything in Google Drive, so it easily connects to cloud. But pretty soon, I think we'll start integrating it directly with their internal systems.

That knowledge base is half of the puzzle. The other half, to get an AI to write high quality content for you, are the instructions. And nowadays, those instructions take the form of Claude Skills.

First skill we use generates an article brief. For this, we give it the topic and any other information we have. So perhaps the client has given us some notes or a voice memo, or they've done a video on this topic, and we can use the transcript.

Other times, we just have the topic. Whatever we have, we dump it in there, we give it the skill, and we let Claude go.

Claude then searches in a search engine for the term we're trying to target. It starts doing some research. It's trying to identify a few things.

One, what's the intent behind this term? And what type of information does the searcher expect to find? We want to make sure we're matching that intent.

Two, what information do we need to collect to make sure this article is robust. That will ultimately define the research we need to do.

And 3, what insights from the client need to be incorporated into the article. That will anchor the point of view of the article.

Next up is the actual research, and we currently use ChatGPT's deep research for this. It's still better than clouds. That takes 10 or 15 minutes.

We give the research and the article brief to another cloud skill. This skill creates the first draft.

This is one of our largest skills. It contains a detailed description of what success looks like. It has SEO guidelines, guidelines for maximizing fact density and for structuring the article to maximize the chance to get cited by a I.

Search. And of course, throughout this whole process, the A I. Has access to that knowledge base.

That knowledge base is also helping it match the tone of voice for this specific brand. And then once that draft is complete, we're giving it to our first revision skill.

First revision compares the article to all content guidelines we created for this brand. And those guidelines come from all previous content we reviewed and provided feedback on. Anytime we or the client had feedback on an article, all of that is getting fed back into this document.

Part of their knowledge base.

And this entire step is devoted to making sure any feedback that has been received in the past is applied to every article going forward.

Then the second revision skill is a fact checker. We make sure that every factual claim in the article is backed up by another source. We make sure we're never linking to competitors, and we make sure that every link is working and actually backs up the fact it claims to.

Finally, we have a steel that extracts descriptions for any infographics we might want to put throughout the article. We'll usually add five or six, and then those descriptions are given to ChatGPT to generate the images.

Source Intelligence

i

High-quality AI content needs a brand knowledge base, reusable instructions, revision memory, and fact checking.

AI Content Workflow · asserts

  • Before scaling AI content, build a knowledge base and revision workflow that captures brand voice, client feedback, source requirements, and factual checks.

Questions this source answers

i

What is this source mainly about?

High-quality AI content needs a brand knowledge base, reusable instructions, revision memory, and fact checking.

What should an operator take from it?

Before scaling AI content, build a knowledge base and revision workflow that captures brand voice, client feedback, source requirements, and factual checks.

Which topics does it connect to?

This source is connected to AI Content Workflow.

What public evidence supports the record?

First revision compares the article to all content guidelines we created for this brand. And those guidelines come from all previous content we reviewed and provided feedback on. Anytime we or the client had feedback on an article, all of that is getting fed back into this document. Part of their knowledge base...