Case study

I Tested 7 Google Core Update Assumptions on My Own Website: What Actually Moved the Needle (and What Didn’t)

By Ravinder Cheema, Clicks Dynasty · September 29, 2026 · Data collected September 26–29, 2026

Short answer: word count, exact-match pages and a higher PageSpeed score did not decide where our pages ranked. What lined up with rankings was matching search intent, consolidating pages that competed with each other, and publishing original data. Our two page-one results both came from that second group.

I run a small SEO agency in Bakersfield, so when a client loses traffic after a core update, the first question is always “what do we change?” I got tired of answering with other people’s theories. So I tested the common ones on the one site where I can see every file and every number: ours.

This is a one-site case study, not a study of hundreds of domains. That’s a real limit, and I’ll point out every place it matters. The upside is that nothing here is borrowed: every number comes from our own pages, our own tests, or two studies we ran ourselves.

The experiment: what we tracked, with which tools, and when

2026 has already had two broad core updates (March 27 to April 8, and May 21 to June 2) plus spam updates in August and September. The September spam update started on September 24 and was still rolling out while we measured, so treat every ranking here as a snapshot, not a verdict.

• Site: clicksdynasty.com, 86 HTML pages, including 29 service and guide pages and 27 blog posts.
• Rankings: 29 manual Google checks on September 26, 2026, covering 11 of our URLs. We recorded which results page each URL appeared on.
• Page speed: Google PageSpeed Insights (Lighthouse 13.5, mobile, simulated slow 4G), one run before our fixes and one after, both on September 29.
• Content: a script that counted the words in each page’s main content and compared every pair of pages for shared five-word phrases.
• Search Console: the AMP report.
• Our own research: a robots.txt check of 100 Bakersfield business websites and a test of 40 ChatGPT and Perplexity answers across five California cities.

Then we took seven things SEO advice commonly says about surviving core updates and checked each one against that data.

What failed: the common myths our data didn’t support

Myth 1: Longer pages rank better

They didn’t for us. Our two page-one results were a 1,237-word service page and a 1,009-word salary guide. Our longest tracked page, a 1,813-word pricing guide, reached page two. A 437-word About page reached the same page (page four) as a 1,067-word agency page for a similar search.

Google says the same thing directly. Its helpful content guidance asks whether you’re writing to a word count because you heard Google prefers one, and answers: “No, we don’t.”

Myth 2: One exact-match page per keyword wins

This one surprised me the most. We have a page built for “SEO company in Bakersfield,” with that phrase in the URL, the title and the H1. For “seo company Bakersfield,” Google showed our local SEO page instead, on page four. The exact-match page didn’t appear in any of the results pages we checked.

We’d already seen this once. A separate “Bakersfield digital marketing agency” page was competing with our homepage for the same searches, so we merged it into the homepage with a permanent redirect. Two pages chasing one phrase split the signals Google uses to pick a winner. One stronger page is usually better than two similar ones.

Myth 3: A higher PageSpeed score is a quick ranking lever

We fixed the performance issues PageSpeed flagged in our own homepage scripts and markup, then retested. Here’s what actually changed:

The fixes were real, but the score went slightly down. One fix even backfired in an instructive way. The homepage headline used to fade in from invisible, so Lighthouse measured our small logo as the “largest” element instead. After we removed the fade, it correctly measures the headline, which is honest but not faster.

That matches Google’s page experience documentation. It says there is no single page experience signal, and that Google shows the most relevant content even when the page experience is sub-par. Speed is a tiebreaker between equally relevant pages, not a way to outrank a better answer. We’re still fixing it for visitors, just not expecting rankings from it.

Myth 4: You need AMP pages to compete on mobile

Search Console told us: “We did not find any Accelerated Mobile Pages in your site.” Some site owners read that as a warning. It isn’t one. Google stopped requiring AMP for Top Stories in 2021, and our homepage scores 100 for SEO in Lighthouse’s mobile test without it. The only AMP work we did was a cleanup: we redirected leftover /amp/ addresses from our old WordPress site to the real pages.

What actually worked

With one site, I can’t prove causation. These are the patterns that lined up with our best results, in order of how strong the evidence is.

How do you match search intent after a core update?

Our salary guide ranks on page one for “digital marketing salary.” It answers the question people are actually asking: what these jobs pay in our region. When we audited it, though, its headings read like a service page (“What You Get With Our Digital Marketing salary Approach”), because it had been built from the same template as our sales pages.

That template problem ran across 29 pages. Their text was mostly original: on average they shared only 4% of their five-word phrases, and 9% at most. But they all used the same headings with the keyword dropped in. We rewrote 85 of those headings so each page reads like what it is: a jobs guide reads like a jobs guide, and a service page reads like a service page.

That kind of pattern is what Google’s new SAFE spam system, described in September 2026, is reported to look for: templated content with small swaps across many pages.

Does publishing original research help rankings?

Our other page-one result is our AI SEO services page. It ranks on page one for “ai seo bakersfield ca” and “ai seo company in Bakersfield.” It’s also the page that links to our California AI Visibility Index, built from 40 AI answers naming 198 local businesses.

That research is information nobody else had. It showed that ChatGPT and Perplexity agreed on only 24% of the businesses they recommended. Our crawler check found that 7 of 100 Bakersfield business websites turn away at least one AI crawler. Google’s own self-assessment asks whether content “provide[s] original information, reporting, research, or analysis.” In our small sample, the page backed by original data outranked pages that had only a strong keyword.

How do you fix internal links and pages that compete with each other?

Our audit turned up links pointing at an address that redirected. Clicks worked, but each one went through an extra hop, and the link’s value was passed through a redirect instead of going straight to the page. We also found that anchor text was sending “SEO company” signals to a page Google wasn’t choosing to rank.

We pointed those links straight at the pages Google already preferred, and added contextual links from our blog posts to the most relevant service page for each topic. The rule we followed: support the page Google has already picked instead of fighting it.

Should you delete low-quality pages after a core update?

Not as a first move. Google’s core update guidance warns against quick-fix changes and treats removing content as a last resort. We didn’t delete anything. We merged the one page that was competing with our homepage, and rewrote the pages whose structure didn’t match their purpose.

Conclusion and a 3-step checklist

The assumptions that failed were the mechanical ones: more words, exact-match pages, a higher score. What lined up with rankings was editorial: a clear purpose for each page, one strong page per topic, and information only you have. That’s the same thing Google means by experience in E-E-A-T. Google also says E-E-A-T “isn’t a specific ranking factor,” so you can’t add it with a plugin.

1. Find pages that compete with each other. Search your main keywords and note which of your URLs Google shows. If it picks a different page than you intended, strengthen that page and merge or redirect the other.
2. Read your headings out loud. If a heading only makes sense with a keyword pasted into it, rewrite it for a person searching that topic.
3. Publish one thing only you know. A small test, a local dataset, or a real before-and-after. Our 100-website AI crawler study took a few days of work.

If you want to see how your own site shows up in Google and in AI answers, our local SEO team starts every engagement with exactly this kind of audit.

Frequently asked questions

How long does it take to recover from a Google core update?

Google says some changes take effect in a few days, but it can take several months for its systems to recognize improvements. Google also notes that some improvements may not show until later updates. There were two core updates in the first half of 2026.

How do I know if a core update affected my site?

Check the Google Search Status Dashboard for the update’s start and end dates. Wait at least a week after it finishes, then compare Search Console clicks and positions for the same length of time before and after. A small drop, like position 2 to 4, usually doesn’t need action.

Does word count matter for SEO?

No. Google says it has no preferred word count. In our data, a 1,009-word guide ranked on page one while an 1,813-word guide reached page two.

Is E-E-A-T a ranking factor?

Not directly. Google says E-E-A-T itself isn’t a specific ranking factor, but its systems use a mix of signals that tend to identify content with strong experience, expertise, authority and trust. Trust matters most.

What is the difference between a core update and a spam update?

A core update is a broad change to how Google assesses content overall, and it doesn’t target specific sites. A spam update targets pages and sites that break Google’s spam policies, such as scaled content or link schemes.

Limits of this case study

• One site. These patterns come from our own website. They’re a starting point for your own testing, not proof of how Google works.
• Page-level rankings. We recorded which results page each URL was on, not exact positions, and we checked by hand, from one location.
• Timing. The September 2026 spam update was rolling out during our checks.
• Single PageSpeed runs. Lab scores vary a few points between runs.

Update plan: the heading rewrites and internal link changes went live on September 29, 2026. We’ll add Search Console clicks and rankings for the eight weeks that follow to this post in late November 2026, including anything that didn’t work.