AI is so good now, specifically cloud, that I think for certain types of work, it's just unethical to do it without AI. Let me explain what I mean using an example from our SEO agency.
A common task that any SEO will be familiar with is internal link optimization. An internal link is just any link from your website pointing to another page of your website, and it can have a huge difference on how you rank in Google. Throughout the years, there have been many tools that attempted to automate this process, but they've all effectively been garbage.
Doing this well requires deep knowledge not just of SEO but of this specific client's website and their current rankings.
Every time you link from one page to another page of your website, you pass some authority to that page. But if you add too many links, you dilute the value of each individual link. You only really want to link to a page if it's going to make a meaningful difference.
The words you link from, or the anchor text, make a big difference in how that authority is passed through. It is not just about whether a link will make a difference to the page, but for what term it will make a difference.
Not every link is treated equally. The more authority a page has, the more authority it can pass through with the link. Some links are global.
If you put a link in the main navigation or footer, it is going to be on every page. Except Google doesn't treat links in the main navigation the same way it treats links in the footer. Mapping all of this out and doing it well for a large site is easily a week's worth of work.
We've been using AI to help with parts of this process for a while now. Yesterday I decided to finally sit down and redo the whole process because I knew Fable the exact process I wanted it to use. I spelled out in great detail the goal.
In fact, I spent 30 to 45 minutes just rambling about everything that matters for internal linking.
I knew it was going to need a lot of data about all the pages on the website, the content on those pages, and how they already link to each other. I had to make sure it could get all that data using the Ahrefs MCP connector, and sure enough, it could. I also knew it would need data from Google Search Console about what terms the pages were getting impressions for and what their average position was.
I know we can get that data using the Ahrefs connector, but sometimes Google Search Console data doesn't correctly pass through to Ahrefs. To make sure we had a backup, I just wanted a cloud code and had it spin up a Google Search Console MCP we could use. With all that context in the connectors, I just let Fable go.
I said, figure out the best process for doing this, and then we're going to turn it into a reusable skill. It got to work. It spun up a bunch of Opus subagents.
It performed six separate runs of the process to test against each other. A few hours later, it had created a skill it was proud of.
I ran a test of the skill on one of our clients that has about 10,000 pages on their website. The results were nothing short of amazing. Honestly, it's better than what I could have done in a week's worth of work.
The net result is that we only have to charge our client a couple hundred dollars for work that's better than what we would have otherwise charged several thousand dollars for.
It's not that the work isn't worth several thousand dollars. That's what I've charged in the past, and it has been well worth it. It's just that there's no need to charge that anymore.
This is just one example, but I think there are plenty of times where doing this kind of work without AI just isn't ethical anymore.
