The March 2024 Core Update. What to know and how to adapt.

The March 2024 Core Update. What to know and how to adapt.

The March 2024 Update has finally finished rolling out. We have a lot to study and learn about before we can understand what happened with this update.

The March Core Update

What’s most important to know:

  • Google called the March Core and spam updates “an evolution in how [they] identify the helpfulness of content.” (See below for my thoughts on Gemini 1.5’s potential role here.)
  • The update was designed to show “less content that feels like it was made to attract clicks, and more content that people find useful.”
  • The purpose of the update was to reduce unoriginal content in search results. This effort started in 2022 with the launch of the helpful content system.
  • There is no longer a helpful content system. The core ranking systems have been refined to help Google better understand if webpages are helpful or if they were created for search engines instead of people.
  • Google predicted the changes in understanding helpfulness that started in 2022, and were improved with this update would reduce low-quality, unoriginal content in search by 40%. After the update finished they changed that number to 45%. (I share below where this number likely comes from.)

Added May 7, 2023:

Ranking turbulence following the update

There’s significant turbulence in rankings for several sites that I monitor. Some that improved with March core are declining.

And some sites are seeing wild ranking swings.

March core update improvements

Perhaps it will take time for the systems to learn and adapt and settle down. Or, perhaps this is the new norm? I do hope to study these cases more thoroughly in the weeks to come.

There is also a May 5 update which is going to confuse people

Google pre-announced the release of a new spam policy against people using the reputation of third party sites for SEO purposes. The Site Reputation Abuse policy went live on May 5. Google started handing out manual actions to many large sites including large news publishers with coupon subdomains or folders on the evening of May 6. Danny Sullivan said that algorithmic actions are to follow.

Given that we’re seeing fluctuations in rankings for sites impacted by March Core, I expect that many sites are going to be confused as to what impacted them. If your rankings are down significantly in early May, unless you are a reputable domain that is renting out space on your site to people who are using you for your site authority, assume that the changes are related to the new updates to Google’s ranking systems introduced with the March 2024 core update.

Important Google Documentation

I’m listing these here because I can never find them! Here are the blog posts from Google that relate to the March core and spam updates.

What web creators should know about our March 2024 core update and new spam policies 

Helpful content and Google Search results FAQ

New ways we’re tackling spammy, low-quality content on Search – by Elizabeth Tucker. 

Google Search Ranking Update Status Dashboard

Help me study the March core update

In order to study and share about sites that are impacted, I need to be able to spend a good amount of time studying and observing. I would love it if your site could be a part of this. I’ll be making videos as a part of my course, and also to be available publicly. 

Here’s more information on how you can share your site to help me analyze this update.

First, let me share my frustrations about Google

Well this section turned into its own blog post: 

Dear Google. Please do more to help the website owners impacted by HCU and March Core.

Google updates have devastated many sites and there is very little clear guidance from Google on what to do.

Here is my friend’s recipe site. She worked so hard on this.

It frustrates me that Google has given so little guidance to help site owners. I understand they don’t want to help those who work to manipulate rankings. But so many good people who have done good honest work are suffering right now with no clue of whether recovery is possible. I have many ideas. I will share some in this post, and more in the weeks to come. 

I believe for most sites significantly impacted by the March core update, or the September HCU recovery may not be possible.

For others, this may be your opportunity to step outside of your comfort zone and experiment with creating original content that truly adds to the body of knowledge on the topics that you write on, in ways that people truly want to learn about and engage with. I believe we will see more and more individual posts of helpful content rewarded in SGE-like helpful content carousels.

Creating truly helpful content is no easy task. There is no direction from Google. And you will be guessing on what to do. You’ll likely need to thoroughly understand your audience and produce information that gives them an advantage, teaches them something they didn’t know and does it in ways that no one else on the internet does.

Some sites did quite well!

Not all sites did poorly though! Some sites impacted by previous core updates (not September HCU) have seen wonderful improvements.

improvements in march 2024 core update

And some sites are seeing incredible improvement .

doing well in March core 2024 update

It will take a lot of study to understand why. This is an entirely different type of update than our industry has ever seen before.

More details to know about the March Core update

  • The update ran from March 5, 2024 until April 19, 2024. For some strange reason, Google did not announce the end of the update until a week later. I suspect we’ll see turbulence in rankings in the weeks to come as the machine learning systems continue to learn and adjust their weights.
  • Multiple core systems were updated. Google doesn’t say this, but I suspect these likely are all machine learning (AI) systems.
  • The helpful content system is now a part of the core ranking systems. There’s a lot for us to understand here. Stay tuned as I hope to publish more on what this means. There are questions about whether or not there still is a sitewide signal of unhelpfulness, and more. Google says there is no longer a single signal or system, yet clearly entire sites are suppressed as we have not seen HCU recoveries. 
  • Google also started a spam update that ran from March 5-20, 2024. (A few more thoughts below on this.)

What should you focus on if you were impacted?

Google updated their advice on diagnosing traffic drops with some helpful information. If most of your drops are small – like from position 2 to 4, then this can be normal ranking changes. No changes may be necessary. I do think that some who believe they’ve been impacted by recent updates are simply seeing normal ranking changes. 

If there is a large drop in position (like from the top 10 to position 29), Google says to self-assess your website to make sure it’s helpful, reliable and people first

What does it mean to be helpful, reliable and people first? 

I used to think that Google’s documentation on helpful content was a list of the types of things their systems tried to reward. Today, I understand this differently. I see that the questions Google says to ask are not criteria that Google’s systems look for, but rather, these are the types of things that people tend to like and find helpful. The systems use user engagement signals to determine how to predict what other users are likely to find helpful.

After intensively studying Google’s AI system for some time now, it’s incredibly obvious to me that search is primarily driven by preferences of users. I know that statement brings up a lot of questions. I will be sharing much, much more on how I came to this understanding. It takes a whole book to explain it, but I’m getting better at distilling my thoughts down. 

How I think search works

  • Every search sends Google signals they can use in machine learning systems. The systems use signals like clicks, scrolls, mouse hovers and others to help determine if the algorithm did a good job at satisfying a search.  
  • The machine learning systems learn which signals are best to use and how much weight to give each of them so that the algorithm continually gets better at predicting which results are likely to satisfy a search.
  • Quality raters rate the rankings that this system produces. They look at each site in a SERP given to them and rate them as likely to be helpful or not, according to the guidance in the rater’s guidelines. If raters are rating results as helpful, then this means the machine learning systems are doing a good job at ranking results. The feedback of the raters helps fine-tune the machine learning systems.
  • The Freshness System helps Google identify new, original and insightful content that searchers are likely to find helpful. It understands what is relevant to a query by learning from what searchers click on and find helpful. There’s a lot for us to learn here as well. I’ll be sharing more on what I’ve learned about this system in the weeks to come. Till then, here’s an interesting patent from Google on Freshness-based ranking to read.

What this means is that if your website has suffered following the March Core Update (or Sep HCU), then it is because the algorithm created by the complex mixture of machine learning systems has decided that your content is unlikely to be what searchers choose and engage with. 

These decisions are primarily based on signals from actual user searches and journeys, as mentioned above. It’s not hard to imagine that a system could use signals like clicks, scrolls, return to the search results, dwell time, shares and bookmarks and hundreds of other possibilities to help indicate if a search has been satisfied.

Every time I bookmark or share something I found from search I think of how my actions told Google’s systems they did a good job in satisfying my search.

If you were impacted by recent updates, your top priority is to look at what Google is ranking now, put yourself in the shoes of a searcher and determine how you can be the choice that people would consistently choose to click on and engage with.

You need to convince people…not Google…that your content is the answer they are looking for.

This is no easy task!

Put yourself in the shoes of a searcher

I will soon have several resources out to help us do this. In the next few weeks I’ll publish a book, some workbooks and also a course for those who want to really dig in and learn. And there will be videos too as I work through analyzing the impact of this update.

I have been working like a maniac for over a year now to produce these resources because I could see that search was changing and we are going to need to learn a lot as an industry.

Is recovery possible?

Historically, if you have been impacted by a Google core update, recovery is often possible provided you’re able to put in some good, hard work. However, it is not always. I think it is going to be significantly more difficult to recover following a March core update hit than any core update before.


Core updates often reflect changes related to E-E-A-T. What I mean by this is that the systems shift to recognize which queries should be answered by authoritative sites or sites with real world experience that you can’t replicate. If you look at your SERPs and see that most of your rankings are held by known authorities, nationally recognized brands, or perhaps government or academic institutions, then recovering these search terms is likely going to be difficult. If people are preferring authority, then you need to somehow convince people to choose you over those sites. No easy task.

I do think there is opportunity here. I’d suggest focusing on aspects of your topics that you have unique and helpful knowledge on. I’ll be talking a lot about information gain and also topical authority (we can learn a lot by studying the topic layer of the knowledge graph) in the weeks to come.


Core updates also reflect changes in Google’s understanding of what people are finding relevant and helpful. Remember those signals of user satisfaction we have been discussing? They help Google understand what people find relevant to a search. This is important! Relevancy is not just about having language that is semantically relevant to a query.

As you look at your search results, do you see sites similar to yours ranking? Ask yourself why. How do they better meet user need? Focus on understanding what needs they are meeting and determining what information gain they offer. Determine how you can offer more value. This usually comes from drawing from your experience to produce helpful information that truly is original and insightful. 

I realize this sounds vague. And that’s why I’ll be creating a bunch of content to help you as I review sites impacted by this update. Here’s my YouTube channel and podcast where most of it will be published. 

If you were very strongly impacted by either the March core update or the previous September helpful content update, then I am concerned. I believe that the main reason for this is that your website previously thrived because you previously did well at creating content that search engines like. If I’m right and Google has shifted more towards using signals of user satisfaction in AI systems to determine the weights of the ranking systems, then it means that it’s less and less helpful to organic search performance to have well structured content, good keyword use, and links. I mean, of course those things are important! But they are not usually the things that make your content actually more helpful than others.

What matters is to have content that people consistently choose when offered in the search results and consistently find helpful. And, it’s important to say this too – your content needs to be consistently found helpful by people who are searching. It’s not enough to have an engaged audience. You need to be the choice of people’s searches and consistently be the content they were looking to find when they did their search.

Search is a library. Google is the librarian. Searchers decide what resources are helpful.

This analogy came to me as the result of a Bard conversation months ago. 

Google’s mission is to organize the world’s information and make it accessible to all. Our website pages are like books in a giant library that a searcher might choose to check out. Historically, our job as SEOs has been to make our books look good to the librarian (Google’s systems.) The librarian does not know if our books are helpful. But they can predict it. 

They have historically predicted what will be helpful based on what they know about the resource – the title, the headings, the topics and entities within and the recommendations of others, links.

Links are signals that were at the core of the PageRank algorithm. This algorithm used links as its main signal to predict which results are likely to be helpful. However, today’s machine learning systems can use hundreds or even thousands of signals. Those signals include clicks, hovers, scrolls, return to the search results and more – signals indicating user satisfaction. 

Machine learning systems have made it so that it’s much less important to please the librarian than the people who have checked the resource out of the library

I will have much more to come on this topic in my book, workbooks and course. Writing to please the searcher rather than the librarian is not easy. And not something we have been taught to do.

More questions on the helpful content system

The helpful content system started in 2022. After almost two years of running, Google has now incorporated the helpful content system into the core ranking systems.

We have a lot to learn about what this means.

I hope to write more soon on:

  • Is there still a sitewide classifier for HCU? 
  • Should you still focus on removing content if impacted? Google says, “having relatively high amounts of unhelpful content might cause other content on the site to perform less well in Search to a varying degree. Removing unhelpful content might contribute to your other pages performing better.”

Is it even possible to see an HCU recovery? I’ve got one case that looks promising. Although it’s possible that these declines are more related to the October core update than the September HCU. This site worked on noindexing loads of pages that had little or no content. This site has given permission for me to share (more on this below if you’d like to do the same), so I hope to dig in further. I want to make a video to show how I’d work now on making the current content original and insightful.

poss hcu recovery starting?

Thoughts on Gemini 1.5 and Search

I don’t think this is a coincidence that the March Core update happened shortly following Google’s announcement of new AI technology, in Gemini 1.5. This is not to be confused with the Gemini Chatbot. Gemini 1.5 is a whole new architecture. Most of the language models and AI systems we use today are built on the transformer architecture. Gemini 1.5 uses a different type of technology called a mixture of experts model. MOE has been around for a while, but Gemini 1.5 is a whole new type of MOE. 

Google says:

“Gemini 1.5 delivers dramatically enhanced performance. It represents a step change in our approach, building upon research and engineering innovations across nearly every part of our foundation model development and infrastructure. This includes making Gemini 1.5 more efficient to train and serve, with a new Mixture-of-Experts (MoE) architecture.”

This technology makes all of Google’s AI systems that use it more efficient, and able to handle larger amounts of data.

Search is essentially a bunch of AI systems. There is no doubt in my mind that Gemini 1.5 has drastically improved Google’s ability to use machine learning for rankings. 

My thought: We are no longer in the days where we optimize for an algorithm that focuses on a small handful of signals, the most important of which being PageRank. Today’s search algorithms are run by a complex mixture of machine learning systems. These systems can use hundreds, or perhaps even thousands of signals.

How to analyze your data in regards to the update

I’ll be publishing more soon on ideas to help you understand whether you were impacted, whether particular parts of your site were impacted more and help you decide with data, what to work on.

For now, if you have ChatGPT Plus, have fun with this Ranking Changes Scatter Plot GPT. You can prompt “how do I use this” and it will guide you through exporting data from GSC and creating scatter plots that help you visualize how keywords and pages were impacted.

scatter plot GPT for GSC keyword rankings

I have found it to be helpful, but know, as with any LLM tool, accuracy can be an issue!

What if you’re not sure if you are impacted?

Try my algo update checker GPT.

I’ll soon be publishing a free guide to help site owners diagnose whether Google updates have impacted you. You can try this GPT I’ve been working on. It’s not perfect, but I’ve been pleased with most of the answers it gives. The knowledge base of this GPT has a bunch of my thoughts and processes on analyzing traffic. It will point you to resources that can help you understand more.

Marie's algo update checker GPT

How can Google say we’ll see 45% less low-quality results when search results still clearly have issues

Earlier I mentioned that the quality raters rate search results. The raters label results as helpful or unhelpful.

raters provide Google with examples of helpful and unhelpful results

The DOJ vs Google documents tell us that these ratings of the raters produce something called an IS score. The document does not say what IS stands for, but I believe it is “Information Satisfaction.”

DOJ vs. Google - IS scores copmuter by rater rankings.

If search is producing more results that the raters rate as helpful, then IS scores go up.

I suspect that this is where the prediction of a 45% reduction in low-quality, unoriginal content comes from. 

Google says that the improvements are compared to results before the helpful content system was created in 2022. 

If you are still seeing unhelpful results in your search results, it’s because only 45% have been reduced. 

There is still a lot more change to come!

What about the spam update?

The March Core update started on the same day as Google announced a spam update. Many sites were given manual actions for pure spam or thin content. From what I’ve seen these were all sites that were really overdoing it. They weren’t all blackhat spam. Some decent sites that have been working on content for a long time. Yet, their content was generally mediocre and rarely original or insightful. They previously thrived because they knew very well how to please the librarian.

If you were impacted between March 5 and March 20, it will be impossible to tell whether it was the spam update or core update that impacted you.

As I do site reviews (see below), I may have more information to help here.

In most cases, if you are not sure, assume it was the core update.

Where to find Marie

Join my community, the Search Bar. We’re a group of SEOs and website owners who are excited about the future and learning about AI and Search. There are free and paid areas.

Read my SEO and AI newsletter.

Listen to my podcast (resuming soon, I promise!) or Subscribe to me on YouTube to hear more about my investigations into the March Core Update Changes.

Stay tuned for upcoming resources:

  • SEO in the Gemini Era: The story of how AI changed Google search. (book)
  • Put yourself in the shoes of a searcher (paid workbook)
  • What it means if you have been impacted by a Google update (free workbook to help you diagnose whether a site has been impacted by an update or not.)
  • Marie’s method for SEO in the Gemini Era (a thorough course and soon workbook as well to help you learn as much as you can about Google’s machine learning ranking systems. You’ll have a huge advantage over those who do not know this knowledge. 

Work with me.

Discover more from BloggerSEO

Subscribe to get the latest posts sent to your email.


No comments yet. Why don’t you start the discussion?

    Leave a Reply

    Your email address will not be published. Required fields are marked *