The Old Affiliate Website Back to Life Using AI Project - Week 10

Hey Wa'ers and yes, we are up to week 10. It seems a lot longer than that - and certainly my impatience thinks it is longer. Anyway, onwards and upwards.

But first, for those of you new to my run of posts here, in my 'real life' my wife and I recently moved from Australia to the US to open Pilates studios. In that time we have found the differences between the two countries wider than I ever imagined. Trust me when I say that Australian English and American English are totally different dialects of the same language. e.g. Last time my wife asked for butter, she was given Siracha sauce.

But today, a milestone. I made it through a drive through without having to repeat myself. Two coffees ordered without "sorry, what are you after?", and they were the correct order! Winning!!

Now, in terms of my rebuild, this week led me to two rather large learnings. One is something that I am a little embarrassed about and the second is a massive learning. So before I get into the AI generated part of the report, here they are.

Spelling mistakes on my websites

For our studios, and for the site I am rebuilding, I have AI tools that do everything from checking SEO, GEO, internal links, page speed and so on. However this week I was doing some wording updates on the Studio site and there they were... typos and blatant grammatical errors through out the pages. So I asked AI to do a run and the fix link was embarrassingly long! The MMO was just as bad. They are all fixed now, but the learning: don't forget the basics first.

Failure to get a baseline

This one is a bit more complicated but refers back to the impatience I mentioned above. If you read below you will see that the site has more traffic this week but still ZERO referrals. This is frustrating as before I started the AI driven relife, I was getting around one every 2 weeks. Now I have had 1 in 10 weeks.

So, I asked Codex to go back and see where they had come from before we started. And it couldn't as I only had generic linking set up to WA. So, for those of you who have left comments that you are thinking of doing this yourselves, don't forget to build a starting baseline so you can see where things have improved or not.

Anyway, here is the base report fir the week - (note: again, this is all AI written)

Tool Feedback

This week, a reader's suggestion led to a useful change in the Make That Sea Change rebuild: give people something they can actually try before asking for their email address.

I had a page explaining the beta tools, but visitors had to sign up before getting access to the suite. Describing what something does only gets you so far. Being able to use something is a different experience.

Tightening CTA and First teps

For anyone following along, this is my attempt to use AI to help bring a neglected affiliate website back to life. The goal remains 100 Wealthy Affiliate Starter-to-Premium conversions by the end of 2026. There is still a considerable gap between the work going into the site and the business results coming out of it.

This week's work focused on giving readers a useful first step, fixing a misleading signup signal and cleaning up some inherited copy

I added the existing Niche Idea Checker to the beta landing page, above the signup form. Visitors can now open a clearly labelled example or answer questions about their own idea, get a result and download it without handing over their email.

It is an extra planning tool, separate from the five website-audit tools in the beta. Its score helps highlight things to investigate; it cannot tell you whether an idea will make money.

Proofreading

There was a less glamorous job this week too: proofreading.

We screened 279 published pages and made spelling, grammar and punctuation corrections on 85 of them. That included old articles, About Me and Contact Me copy, and text generated by the planning tool.

This was a broad screening pass with reviewed corrections, rather than a full factual rewrite of the archive. My conversational wording and Australian spelling stayed. The changed pages were checked to confirm the corrections appeared publicly and their links remained intact.

A fresher design can still contain years-old mistakes. Working through the copy was a useful reminder of that.

Traffic

The traffic figures moved in a more encouraging direction this week, although there are a couple of important qualifications.

Here is the fresh Google Analytics comparison for two complete seven-day windows, using the property's Sydney timezone:

Recorded sessions rose by about 16%, and organic search sessions increased from 49 to 61. That is welcome movement, but these are analytics counts, not verified human prospects. Owner checks and automated activity may still be included. Most sessions were classified as Direct, so I cannot treat the overall increase as proof that the SEO work caused it.

Google Search Console gives a separate view of Google search visibility. Its latest available final data ends a day earlier, so this comparison uses 18–24 September against 11–17 September:

Google clicks increased by six, while impressions fell by about 30%. The click-through rate improved, but average position worsened; a lower position number is better. So this is a mixed search result, with a small increase in actual clicks rather than growth across every measure.

Search Console clicks and Analytics sessions are different measures with different reporting windows here. They should not be added together, and neither comparison can tell me which particular edit made a difference.

And the business results?

For the latest verified Wealthy Affiliate reporting window, 19–25 September, the figures were:

- New referrals: **0**

- Premium upgrades: **0**

- Revenue recorded in that window: **$0**

WA also recorded 99 raw clicks. Those are not 99 verified people, so I am not treating them as a count of potential customers.

The new routes from selected reviews into the beginner guide have not produced recorded guide-click events in the latest check. There are no opted-in responses from the visitor chooser to analyse yet either. Checks confirmed that the tracking code was still present on all six pilot pages, but that does not establish that real visitors have used it.

There is too little evidence to judge those changes. Missing optional events also cannot tell me that nobody used the chooser.

AI has helped make practical repairs this week. The harder question remains whether the site can attract enough relevant readers and help them take a useful next step. A cleaner page and a working tool are progress, but the commercial outcome is still unproven.

This week

My next priority is to bring more relevant people into the beginner lessons and give the recent changes time to produce evidence. There is already a first-article video script prepared; recording and publishing it is still work to do.

If you try the checker, let me know where you get stuck or what you wish it explained. A reader's suggestion started this week's most useful change, and that kind of specific feedback helps me decide what to improve next.

Anyway - see you all next week.

Paul