Three months in, average position has improved from 41.4 to 25.6, the SEO page reached page three for “Cumbria SEO” for the first time, and a chunk of this month’s work went into finding and fixing a problem the site had caused for itself. The AEO score also dipped for the first time, from 88 to 81, and that is worth being just as honest about as the wins.
| # | Post | Date | Key metric |
|---|---|---|---|
| 1 | Building a Website That Ranks — SEO Diary 1 | 6 May 2026 | Baseline established |
| 2 | Building a Website That Ranks — SEO Diary 2 | 5 June 2026 | 2,260 impressions, AEO 38 → 70 |
| 3 | Building a Website That Ranks — SEO Diary 3 | 4 July 2026 | AEO 70 → 88, SEO page converting clicks |
| 4 | Building a Website That Ranks — SEO Diary 4 | 7 August 2026 | Average position 41.4 → 25.6, AEO 88 → 81 |
For reference, the launch post explains the plan in full.
The cannibalisation problem
The most useful finding this month came from a tool built specifically to catch it: a keyword relevance checker that scans every page on the site and shows what search engines actually see when they read it, then flags where pages compete against each other for the same term.
Running it in early July turned up something worth admitting plainly: two of this site’s own blog posts were outranking the SEO service page for “Cumbria SEO,” its primary keyword. The blog had been doing its job so well it was undermining the page it was supposed to support.
The fix had two parts. The SEO and AEO service page was rewritten to focus more explicitly on “Cumbria SEO” as its core term, strengthening the exact signal it was losing. The competing blog posts were retitled and demoted, dropped from targeting the head term directly, while keeping their internal links to the service page intact so they still send authority to it rather than compete with it.
It did not resolve instantly. On July 9, the service page still was not reflecting the changes in search, and Search Console was showing it underperforming immediately after the edit, the kind of dip that comes with any significant on-page change before Google recrawls and settles on a new assessment. By July 10 it had moved to page three for “Cumbria SEO,” the clearest sign yet that the fix was working rather than just shuffling the problem sideways.
The same relevance tool also flagged softer versions of the same issue on other pages, which led to content updates on the homepage, design and build, and hosting and email pages in early July, all resubmitted through Search Console once live.
The rest of July’s work log
- July 1 — all blog posts briefly dropped from search. Traced to console errors on blog images following a restructure that added categories and breadcrumbs to every post, a false positive rather than a real problem, but worth checking rather than assuming. Sitemap resubmitted with the newly added pages.
- July 3 — month two diary published and interlinked with the rest of the series.
- July 4 — the keyword relevance tool built and run for the first time, surfacing the cannibalisation issue above. A redundant contact page link causing an unnecessary redirect was also found and removed while working through the results.
- July 5 — homepage, design & build, and hosting & email content updated based on the relevance tool’s findings, resubmitted in Search Console.
- July 9 — SEO page not yet reflecting changes in search; resubmitted again, Search Console showing a temporary dip.
- July 10 — SEO page reaches page three for “Cumbria SEO.”
- July 11 — website maintenance blog post published, with category mentions interlinked to their relevant service pages. Site diary titles across the series retitled away from the “month one / month two” naming, which was targeting redundant keywords rather than anything useful. Reindexing requested.
- July 24 — hosting bill blog post published.
- July 25 — resubmitted in Search Console.
Then a deliberate pause. After a fortnight of structural changes, letting things settle and recrawl rather than continuing to intervene.
Search Console — month three
4 clicks, 1.68K impressions, average position 25.6. Down slightly on impressions from June’s 1.93K, but average position improved sharply again, from 41.4 to 25.6. Three months running, the same pattern holds: the raw numbers move less than the underlying position, which is the harder-to-see but more meaningful metric at this stage.
The SEO page holds its place as the one to watch: 1,122 impressions and 1 click this month. Still only a single click against four figures of impressions, but that gap is exactly what the cannibalisation fix and the page-three ranking are aimed at closing over the coming months.
Hosting & Email picked up 301 impressions and its own click, the homepage logged 103 impressions and a click, and Design & Build appeared with 52 impressions and a click of its own. Every core service page is now converting at least one click a month, which was not true even last month.
Google Analytics — a false positive worth explaining
Google Analytics initially showed 56 active users for the month, which would have been real growth on June’s 42. It was not real, and figuring out why matters more than the number itself.
Ten of those 56 users were landing directly on the site’s raw hosting-platform domain rather than 1418.co.uk, the fallback address every deploy gets automatically, including the live production build. Filtering the report down to that hostname specifically showed ten sessions, all landing on the homepage, all with zero engagement time, spread in small spikes across the month. That is the pattern of automated requests or crawlers hitting the raw domain, not real visitors.
With that traffic set aside, the real figure for July is closer to 46 active users — a modest but genuine increase on June, not the jump the raw number suggested.
The fix goes beyond just filtering the report. The real issue is that the fallback domain, and the individual preview-deploy subdomains that come with every branch and pull request, are visible and crawlable at all. The next step is adding a noindex header scoped to anything that is not the production deploy, so preview and fallback URLs stop attracting this kind of traffic in the first place rather than needing to be filtered out after the fact.
AEO score — the dip
| Baseline (May 4) | Month 1 (Jun 3) | Month 2 (Jul 3) | Month 3 (Aug 3) | |
|---|---|---|---|---|
| Site Health | 86 (B) | 100 (A) | 100 (A) | 100 (A) |
| AEO Score | 38 (F) | 70 (C) | 88 (B) | 81 (B) |
Site Health held at a clean 100 with zero critical issues, but the AEO score dropped from 88 to 81, the first fall in the series. Eight warnings are showing across 19 crawled pages, up from a cleaner month two. The most likely cause is the volume of content change this month: retitled diary posts, demoted blog posts, and rewritten service page copy all landed inside the same few weeks, and AEO scoring is sensitive to exactly that kind of structural churn while pages resettle.
It is not being treated as a real regression, but it is being written up as one until proven otherwise. The plan for month four is to work back through the warning list page by page, the same methodical approach that pushed the score from 70 to 88 in month two, rather than assuming the drop will resolve on its own.
What month four looks like
Priorities for August: closing out the noindex fix for preview and fallback domains, working through the eight AEO warnings, and watching whether the SEO page’s page-three position for “Cumbria SEO” continues to climb now that the cannibalisation issue is resolved rather than compounding. Content continues at the usual pace.
The next update will be at the start of September with August’s data.
1418 provides web design, SEO and AEO, hosting and email, and website maintenance for businesses in Cumbria. This blog documents the process of building search visibility from the ground up.