His joining Yahoo! was part of a swap, where Google got Jeremy Zawodny:

Zawodny was part of Yahoo!’s platform engineering group. He was traded in exchange for Google giving up Cutts plus just over 2.5k computers.

Jeremy and Matt also swapped blogs out of the deal. (Maybe that confirms search engines don’t care who your author is …)

Meanwhile, Robert Scoble, a technology evangelist at Microsoft, also joined Google. He was offered a job from Larry Page directly.

And just like the blog readers of today, all the posts have comments from well-wishers.

Here’s John Mueller commenting on Matt Cutts’ post:

Of course, this news coming on the 1st of April in 2006 likely had no bearing on its veracity.

Most likely …

It’s a satirical article making fun of actual best printer lists. (You can find it below in the Articles section, h/t: AJ Kohn.)

However, by mocking that type of content (and the historical SEO-first thinking behind it), the author actually created what is a great piece of helpful and people-first content.

The page is succeeding for the reasons we’d expect in today’s Search environment: it offers something relevant and useful that no one else (and no generative AI summary) can quite capture.

Plus, Nilay (the author) has got a brand and following behind him, which resulted in a ton of comments that no doubt helped the page’s quality and relevance.

Then again, I viewed this post before, so personalization of my search results likely plays a role here, too.

And who’s to say that lime green SGE answer won’t pop up with a huge shopping graph-filled extravaganza of information on the next search?

As for the context of the March core update …

Google Search Central’s documentation says this core update is “a more complex update than our usual core updates, involving changes to multiple core systems.”

Therefore, “the rollout may take up to a month.”

Since it began rolling out on March 5th — 33 days ago — in all likelihood, it’ll be completed this week.

The documentation says something else, though …

It notes how the update “also marks an evolution in how we identify the helpfulness of content.”

What that means is “There’s no longer one signal or system used to do this.”

Perhaps we can extrapolate that the inputs and individual models that aggregated together in an unhelpful content classifier previously (the old helpful content system) have now been redistributed within the larger core ranking systems, thus refining that logic to page-level attributes, as well.

To explain this point further, Google’s documentation uses the analogy of its Reliable information systems: “Just as we use multiple systems to identify reliable information, we have enhanced our core ranking systems to show more helpful results using a variety of innovative signals and approaches.”

I examined those Reliable information systems in the context of Discussions and forums earlier this week.

One takeaway I had is that the systems likely pertain to specific details for particular circumstances, such as suppressing freshness signals for less authoritative forum or social media content during breaking news or using MUM to confirm consensus of YMYL information in featured snippets.

However, they also apply to the holistic impacts of ranking systems in their totality, including core systems, such as what quality raters might judge to be the highest quality results.

In a core update, Google could introduce new signals (or entire systems) or re-weight existing signals. Consequently, what was deemed “helpful content” previously may no longer be seen that way.

One way Google Gemini explained a core update to me was, “It’s like changing the underlying ‘blueprint’ the model uses, not just fine-tuning what it’s already learned.”

This gets at a popular line from John Mueller this week about “a re-evaluation of the whole site in the new / current online world & user expectations.”

However, earlier this week, I looked historically at Google’s updates in 2011 to 2012, where we can see not only parallels to today but also how these ranking system adjustments are part of a continuous chain of events, rather than isolated episodes.

More updates (and opportunities) will come.

For example, if we listen to interviews from Google representatives like Alan Kent from even a few years ago, they acknowledge how neural networks in ranking systems adjust weights automatically.

This means that while an engineer can likely speak to system inputs (such as attributes derived from examples of previously helpful content), how the hidden layers of neural networks get weighted to arrive at the outputs (the improved search results, likely based on offline testing) is less interpretable.

That word “interpretable” is key in the realm of AI.

In Hamsterdam Research, we’ve looked at two recent Google Research papers so far, including for ELM and AutoBNN. Although they cover different areas of AI, both papers speak to the interpretability factor — understanding why data or results imply what they do.

That’s likely the reason the advice “optimize for users” is vague, because what users want (per weighted signals in complex AI systems) isn’t clear on the outside looking in.

Other than giving users what they want — and sometimes they don’t want website content, but rather Reddit or social media, and maybe even AI overviews (SGE) or summaries (Perplexity or Copilot) — we’re not always sure why users want the content they do …

… other than that it satisfies their search intent, of course. 😉

Maybe we can look at top-ranking competitors and glean insights … or maybe that approach just leads us to fall victim to a copy-cat strategy, which in turn leads to mishaps like site-wide unhelpful content classifiers.

Maybe we can dissect ranking system signals given trends, but there’s a lot of obscurity and dynamism at play in terms of what Google’s signals and systems are evaluating circumstantially, and by the time it gets interpreted for us, the systems behind it will likely have changed.

In other words, we’re dealing with a world of SEO where the interpretability of how machine learning-based ranking systems operate isn’t always clear.

If we’re going to hedge our bets, it’s probably smartest to bet on a sure(er) thing — Google progressing toward satisfying users’ demands, as expressed in their oodles of data.

Two things we know about users are they’re generally tired of “SEO spam” but also a little hesitant about “AI summaries.”

There’s fertile middle ground in there …

Meanwhile, as alternatives like Perplexity, ChatGPT, and other generative AI-based surfaces with organic web links or brand mentions also grow their user bases or points of visibility (like across our phones), we may need to further evaluate if we’re optimizing for users or Google, literally.

In the meantime, the next few weeks and months will no doubt be filled with analysis of Google’s March 2024 core update and pending AI overview (SGE) launch.

Keep a discerning ear and eye and consider the data in the context of your website.

But knowing that Google’s ways of identifying (and presenting) helpful content have evolved (and will always do so), it’s probably a good idea for us to evaluate if what makes our content helpful in an SEO context should constantly (and considerably) evolve, as well.

Buckle up for a full week’s recap, and enjoy the vibes (and bonus tracks below):

Bonus tracks to celebrate the eclipse! Choose your own musical “eclipse” journey:

• Cryptorchid – Marilyn Manson (for the brave)
• Total Eclipse of the Heart (Turn Around) – Bonnie Tyler (for those who like hairspray)
• Eclipse – Pink Floyd (for the introspective)

Thank you for supporting Hamsterdam and the cause of SEO & AI learning.

Missed last week? Don’t worry, I got you! Read Part 51 to catch up.

Other weekly SEO news sources:

• The SEO Weekly – Garret Sussman, iPullRank
• SEOFOMO – Aleyda Solis
• Weekly Video Recaps – Barry Schwartz, SER

---

Weekly review of SEO-related social posts & more …