The Old Affiliate Website Back to Life Using AI Project - Week 8

Good morning WAers and welcome back you my series where I try and relife an old MMO website using AI.
This has been an interesting week in terms of AI and its influences on our daily lives. Here in the US congress is discussing things like AI kill switches and ex ChatGPT employees are resigning over safety concerns. Here in WA land, there has also been debate on the use of AI for writing posts and comments etc.

For those interested, this project is about me using AI as a 'consultant' and error checker as I rebuild my site. Am I letting it write posts? A limited Yes. Am I proofreading and making corrections? Also yes. I am also pushing back when I want something done and listening when an argument is pushed back. (I am also running an AI only Pinterest experiment in the background that I have not discussed as yet - I want to give it a couple of months before I report findings so that will be end of September).
In short, I am using AI to check all the coding, Schemas, SEO influences and so on against current best practice and Google algorithms etc. to see if we can generate some more WA affiliate income. 8 weeks in, we are making small progress but that first referral income still alludes me.

And in light of some of those conversations in WA this week, These weekly reports consist of my own writing in the intro (i,e, no AI was harmed in this first section), then the main report is 100% AI reporting. I will be clearer in this each time if for no other reason you can compare my voice and the supposed "Paul's voice" that AI is using.

I also have a story to share in regards to a very interesting AI interaction I had in our other business this week. It's response was actually recalcitrant. I will write about that in a separate post.

Anyway, here is the report for this week.

Note: the rest of this post is 100% AI written! The command: Can we do up this weeks WA report?

The plan

For anyone following along, this is Week 8 of my AI Affiliate Challenge: can AI help bring Make That Sea Change, a neglected affiliate website, back to life? The goal remains 100 Starter-to-Premium conversions by the end of the year.

My last confirmed WA outcome check, on 5 September, was still no new Starters or Premium upgrades. This report does not claim a newer confirmed result.

Search took a step back

Last week brought a small improvement in Google search clicks. This week's complete seven-day comparison moved the other way.

Eight fewer clicks is disappointing, but these are still small numbers. I can't declare the recent content changes a success after one good week, then a failure after one weaker week.

The Taking Action Online review, GeneratePress comparison, Juice Plus review and supplement MLM guide each received two clicks in the latest window. Traffic is still spread across several different reader interests.

GA4 recorded 52 Organic Search sessions versus 44 in last week's snapshot. That includes search engines beyond Google and uses a different reporting window, so it isn't the same measure as Search Console clicks.

Overall sessions rose from 144 to 264, but Direct traffic accounts for most of the increase: 92 to 204. Both snapshots include their reporting day and are provisional. Given the site's previous traffic problems, I am not treating that jump as proof of audience growth.

The zero that needed investigating

On 6 September, we found broken tracking code on the WA review. The page's own click listener could not run, while the sitewide tracker was stepping aside because it thought the page already handled those clicks.

I had been looking at a WA review page with no recorded outbound clicks. It would have been easy to decide the page wasn't convincing anyone and start rewriting it again.

Then we tested the buttons properly. The links worked. The click tracker didn't.

That left a gap: a reader could follow the affiliate link without the event being recorded.

We removed the broken duplicate and tested both affiliate buttons. Each registered one event once the page had initialized. An ordinary link to WA's pricing information registered none. Both affiliate buttons still reached WA with my referral details intact.

We also placed a short affiliate disclosure above the top button and checked that both fitted within the first mobile screen.

There is a limit to that verification: a click immediately after a reload did not register during testing. I can say the initialized-page tests passed; I can't claim every possible click will be captured.

This week's processed analytics now shows two WA outbound events. We checked their actual page URLs, and both came from our tagged audit visit. They are test clicks, not two prospective customers.

There were also three recorded transitions into the WA review. Those counts include the audit period and are events rather than verified individual readers. A fourth transition event led to the niche tool, so it doesn't belong in the WA-review total.

The practical lesson is fairly simple: before changing a page because nobody appears to click, check whether the measurement works. The earlier zero cannot establish reader drop-off, and the exact date the fault began is unknown.

What the recordings showed

After restoring access to the correct Clarity account, I could inspect some actual journeys through the review.

Three selected search-origin visits stood out. Readers arrived through MLM content, spent 7, 9 and 49 seconds on the WA review, then returned to MLM material.

That raises a useful question. Is someone investigating a particular MLM company ready to consider building an affiliate business, or have I introduced that alternative before explaining why it might suit them?

Three visits cannot answer that. The wider recording set also contained our audits and suspicious rapid browsing, so I am not presenting it as a clean sample of potential customers.

For now, I am leaving the review stable. If the same pattern appears in more relevant visits, a clearer explanation in the link from the MLM article may deserve a test before another full review rewrite.

The toolkit still needs an audience

The authenticated MailerLite check on 6 September showed five recorded visitors to the Build Something Real form, up from one on 31 August.

Collected subscribers: zero. Completed delivery runs: zero. The existing account subscribers were still my test records.

The delivery workflow is active, but five recorded visits are nowhere near enough to judge the offer. This week's GA4 email-success events are not being counted as leads without confirmation in MailerLite.

The two internal routes added from the homepage and rebuild article reach their first checkpoint on 14 September. Those dated MailerLite figures are the latest verified email results used here.

Turning the rebuild into useful tools

I also completed the first round of four tools in the private control center: an Affiliate Link Checker, Internal Linking Audit, Content Refresh Finder and Combined Self-Audit Report.

They turn public page evidence into repair lists, CSV downloads and a combined 30-day plan. The idea comes directly from the repetitive work involved in this rebuild.

One boundary matters: external affiliate destinations still need manual verification. A network blocking an automated request is not enough to label someone's link broken. Last week's overly confident audit report was a good reminder to make those limits explicit.

These are tools being developed and tested. I am not claiming paying customers or revenue from them.

What happens next

The late-August search tests reach their first 14-day checkpoints this weekend. The fresh Search Console data currently ends on 9 September, so it doesn't yet cover their full observation period. I need the complete data before calling individual tests winners or losers.

The next useful evidence is more qualified readers moving from the content to the WA review, non-test clicks onward to WA, confirmed email subscribers and actual Starter and Premium outcomes.

This week made the reporting more honest. The harder job is still turning the site's traffic into a business result.

Ok, so that is it for another week.

Off to see what we can get moving this week

Paul