The Old Affiliate Website Back to Life Using AI Project - Week 3

Good afternoon all,

I am back from my son's basketball tournament and my Aussie in US learning this week is that DFW is a VERY large city... hahaha.
So, as usual, I will leave the basis of the report to my good mate Chat Codex below - but in human talk, this week has been about managing my AI credits for the website refresh and getting some tools build for my intended salable product.

As I worked through the website refresh - which to be honest was a lot of me going "That's good - please continue" - I realised something. You see we finished refreshing the pages that still attract traffic which led to Codex to start cleaning up behind the scenes.

It kept saying things like "cleaning up residue obsolete code", or "found and removed out of date embeds" etc. I actually made me think that those of us who setup and managed websites pre-AI deserve some recognition. The fact that we kept sites ranking and earning income without AI cleanup seems to be some sort of miracle in itself.

Anyway, here are the formal bits to do with the website refresh.

The Stats

This week produced two numbers that look impressive and one that matters more than either of them. The Make That Sea Change rebuild recorded 52 confirmed implementations or experiments between 2 and 8 August.

Cloudflare then congratulated me for passing 10,000 pageviews last month and showed 53,204 pageviews since I connected the site on 11 July.

The number that matters most?

Zero new Wealthy Affiliate referrals and zero Premium upgrades in the authenticated seven-day WA report. That is the honest Week 3 headline. A great deal was built, cleaned and measured. The commercial result has not arrived yet.

The authenticated WA dashboard for the inclusive 31 July–7 August window showed:

- 183 raw clicks;

- 49 unique clicks;

- 21 content clicks;

- 0 new referrals;

- 0 Premium upgrades;

- $0 revenue in the selected period.

The 1 August row is a useful warning. It showed 133 raw clicks but only three unique clicks and three content clicks.

That was not 133 interested people. It was a small number of unique clicks accompanied by repeated activity.

The latest processed GA4 window was similarly modest: three content-bridge events, two Niche Checker email submissions and one email success. The new sitewide outbound listener had only just been deployed, so its genuine visitor baseline begins after the validation clicks.

Google

Clicks 9 (9 last week 0%)

Impressions 3,391 (4,952 last week−1,561 −31.5%)

CTR 0.27% (0.18% last week +0.09 percentage points

Approx. average position 42.2 (45.0 last week improved by 2.8)

Bing

Clicks 6 (8 last week −2 −25.0%)

Impressions 578 (699 last week −121 −17.3%)

CTR 1.04% (1.14% last week −0.10 percentage points)

The Website Rebuild

As above, the active rebuild queue was effectively complete. Once the obvious high-value pages had been rebuilt, we stopped treating “the queue is empty” as permission to invent more work. Instead, we looked for unfinished conversion dependencies: broken navigation, duplicate beginner articles, obsolete lead pages, missing sitewide click tracking and old archive routes that sent visitors around the useful parts of the site (again, hbow we ever built websites in the past was beyond me - haha).

The completed Week 3 total is now 52 confirmed implementations, with three experiments still genuinely pending rather than being declared successful early.

Earlier in the week, we rebuilt some pages that whilst not high in rankings, can potentially provide us with a good bridge for potential WA referrals. These were pages covering topics such as:

- choosing and testing a niche;

- finding questions and keywords;

- building an affiliate website;

- writing useful posts;

- adding images and affiliate links;

- completing basic SEO;

- understanding affiliate marketing honestly;

- comparing training and tools;

- reaching the Wealthy Affiliate review when that is the relevant next decision.

Website tool and Email capture

The Free Niche Idea Checker also went live. It asks five practical questions, calculates a result in the browser and can continue through a MailerLite email journey. The visitor receives something useful before being asked for an email address.

That replaced the old approach of promising a generic “ultimate toolkit” through forms that did not work.

Navigation updates

Next we looked again at the website structure in light of our page updates. In short, the main navigation was cleaned up as follows:

• Two menu choices—Buying Guides and SmartPhone Apps—led to empty, noindex archives. They were removed.
• The Affiliate Program Managers menu branch served merchants trying to recruit affiliates rather than beginners trying to become affiliates. That menu item was removed while the four articles themselves stayed published and internally discoverable.
• The MLM menu was different. Its ten current status guides still represent a warm audience and produced more than 9,000 impressions in the original 90-day Search Console export. Instead of removing that route, we changed it from a raw category list to the rebuilt “MLM vs Affiliate Marketing” decision hub.

The menu now does more than expose categories. It helps people make the next decision.

A Real Google SERP Error

When I searched for Make That Sea Change in google, an old page called “Enter Your Details Below” appeared prominently. It contained a static AWeber form even though I no longer have an AWeber account. Its success URL was also malformed.

That is not the page I want representing the brand. We confirmed that no published MTS article still linked to it, backed it up, created a permanent redirect to the homepage and moved the source page to Draft. The redirect is live and the obsolete URL has left the XML sitemap.

Google will decide when to replace the old result, but the live destination is now correct.

Every real WA affiliate link is finally measurable

The earlier tracking work covered the rebuilt pages, but a full audit found 106 direct Paul-affiliate WA links across 50 published posts.

Many of those links were not covered by the general `wa_outbound_click` event.

I extended the sitewide journey listener to recognise both `www.wealthyaffiliate.com` and `my.wealthyaffiliate.com`, while requiring my affiliate identity so ordinary links to official WA sources are not reported as commercial departures.

The first browser test actually found a duplicate-handler problem and counted one click twice. That was fixed before deployment. The final validation produced exactly one Realtime event for each test click, including a page that had never had its own tracking code.

GA4 now has six MTS reporting dimensions:

- source page;

- CTA location;

- CTA wording;

- WA tracking tag;

- journey origin;

- internal destination path.

The new journey dimensions collect forward only. They do not rewrite old events.

The point is not to celebrate another event name. It is to connect a future WA outcome to the route that produced it without counting one click twice.

The Cloudflare milestone is real—but it is not 53,204 readers

Cloudflare emailed to say the site passed 10,000 pageviews last month and displayed 53,204 pageviews since 11 July.

Some of those pageviews definitely came from this rebuild.

Every time we published a page, checked its public source, validated schema, inspected a redirect, regenerated a sitemap or tested internal links, we requested live HTML. Yesterday alone, several full-page audits checked 50–70 unique internal destinations each, and the public HTML sitemap audit checked 287 destinations.

Cloudflare defines a pageview as a successful HTTP response with an HTML content type. That means this verification work can count.

However, it cannot reasonably explain the whole figure.

Cloudflare was also seeing roughly 30,000 requests per day. One identified crawler produced more than 8,000 requests in a day. We also found a hostile WordPress/PHP scanner and an image-probe burst. The image requests themselves are not HTML pageviews, but crawlers and scanners that fetch successful HTML pages can contribute to Cloudflare totals even when GA4 does not treat them as normal browser sessions.

So I am treating the email as evidence that Cloudflare has processed substantial site activity—not as evidence of 53,204 human readers or a traffic breakthrough.

This is exactly why the challenge needs Search Console, GA4, Cloudflare and WA to answer different questions.

The narrow security rule remains narrow

The Cloudflare investigation identified two exact hostile signatures: one source probing fake WordPress and PHP paths, and an `ImageResProbe/1.0` burst.

I blocked those two patterns with one narrow rule. Cloudflare estimated that it would have matched 2,190 requests, or 7.3% of the preceding 24 hours.

I did not block Singapore. I did not block the large verified crawler. Those require separate evidence and trade-offs.

The short follow-up window showed no new matches for the exact rule. That does not mean the rule failed; it means those exact sources had not returned during the observable period. The proper comparison still needs a full post-rule window.

What else AI contributed

Over the week, AI did far more than rewrite old pages.

It classified overlapping intent, read live WordPress source, checked current Google and WA information, built guarded redirect scripts, backed up every source, verified first-response headers, regenerated sitemap storage and audited each destination for canonical tags, indexability, H1s, image alt text, internal links, schema, WA bridges and old video/review output.

It also stopped and reversed an otherwise sensible-looking action. One consolidation initially updated 11 old inbound posts to point directly at a new destination. That would have republished 11 legacy pages that still contained unfinished video and Hustle debt. The changes were restored from backup and the single redirect was left to preserve those journeys until each source receives a proper rebuild.

That is an important distinction. Working faster is not useful if it quietly expands the unverified change surface.

What's next?

The site is materially cleaner. The beginner journey is coherent. The direct WA links are measurable. Navigation points toward useful decisions. Broken lead pages and duplicate articles are being consolidated safely.

The business outcome remains zero. The next stage is not another frantic content sprint. It is a cleaner evidence window:

- Which refreshed pages earn relevant impressions and clicks?

- Which journey origins reach the WA review?

- Which review visits become affiliate departures?

- Which departures become Starter referrals?

- Does any Starter finally become Premium?

The 53,204 Cloudflare pageviews are a useful reminder: a large number can describe real activity while still answering the wrong business question.

I would rather report one verified Premium upgrade than another 50,000 unexplained pageviews.

For now, I have neither the upgrade nor a reason to pretend the pageview email is one.

Still there?

So that's it, and if you made it through all that without falling asleep - then another congratulations to you!

See you next week

Paul