Hello everyone,

Oh boy a lot of things happened this week! In this episode, we’ll talk about the potential demise of SGE, Google’s cancellation of their contract with quality rater provider Appen Labs, and the introduction of new AI features like Circle to Search and Multisearch’s AI-generated answers. Plus, we'll discuss Bard’s expected integration with Assistant, a bug causing sites to fluctuate in search rankings, and more.

​

Read this episode in the Search Bar This is Search News You Can Use, free for everyone to read and comment on. This is a long episode, so it may get cut off in your email provider.

I have been writing furiously this week. I added a chapter in my book for schema and also topical authority and am currently writing the section on intent. I also spent time working on improving the prompts in my GPTs I'll soon be releasing.

Five people reached out to me this week to ask me to be on their podcast. Boy do I love to talk and I cannot wait until I finally get this book and course done so I can start saying yes to these. We are soon going to have some great conversations about what is happening to search because of AI.

Let's get into the latest news!

Google canceled their quality raters contract with Appen Labs

​Google canceled their contract with Appen Labs, one of the companies that supplies them with quality raters. Some, including me, wondered if perhaps this was related to the recent research that Dawn Anderson wrote about in an article on Search Engine Land that shows that AI systems can likely eventually replace human raters.

​Google told Barry Schwartz that this was not the case. They've been working on this move away from Appen and towards a new provider for a year. It is worth noting that a year ago, Appen workers protested and got a gigantic raise from $10 to $14.50 per hour. If my math is right, assuming a quality rater works 15 hours a week, and assuming Forbes estimates that 3000-5000 quality raters were contracted from Appen, that’s potentially over $17 million in savings for Google.

Quality raters are still an important part of Google’s systems. We'll talk about this more in a moment.

Circle or scribble to search

Google introduced a new AI feature. You can circle or squiggle over anything on your phone screen and then search. Like if you see a bag you like, you can squiggle (that is such a fun word) over it and then search for that bag:

How on earth will we track these visits?

Circle to search goes live January 31 on Pixel 8’s and also the new Samsung Galaxy S24.

Circle to Search makes use of Multisearch, which allows you to search using images and text. If you have not tried searching with images via Lens, I’d encourage you to give it a go. It seems like a gimmicky thing at first, but I find myself using it often and finding it helpful. Actually, this brings up a really good point. I am betting that there are quite a few AI type features that you've tried once or twice, found unhelpful and now have moved on from. These tools continue to learn, so it's worthwhile going back every now and then to see if they are better.

For example, if you have plants, try "searching" with a picture of your plant combined with a text question like, "how often should I water this?"

Multisearch has AI generated answers now

Here we go with an SGE-like feature that is now in production. Now when you use multisearch you can get an AI generated answer alongside search results. I can’t get this to pop up, so let’s look at the example in the video Google provides for this.

Say you’re watching a YouTube video and they mention something you don’t understand. You can pause the video, scribble over the words, ask a question about them, and up pops an AI generated answer.

Or you can use the camera on your phone and ask a question and get an AI generated answer.

It’s interesting to see there are dropdowns, I’m assuming with links, in the example they give:

I can’t get this to work just yet on my phone. It likely is not live yet. I am hoping this is not another issue with Canada blocking AI.

By the way, I just got a response from the government from the email I sent imploring them to allow Canada more access to AI. Our tech sector is at a disadvantage by not using it. For example, if you are a Canadian SEO, you cannot access SGE and many of the AI features Google is pushing out to search. This makes a Canadian SEO much less likely to be hired to consult with businesses around the world who want to talk about the future.

We should be innovating with AI rather than falling behind. They gave me a person to contact who wants to hear more. Let's see where this leads.

Is it possible SGE will not launch?

In Google’s announcement about upcoming AI features like circle to search, they spoke of SGE. This certainly does make it sound like SGE will not replace search, but rather, will act as a testbed for new ideas that then individually can be introduced into search.

Some are interpreting this to mean that the feedback on SGE was poor, and as such not good enough for Google to push it live. Here are some of the conversations I've seen on this wording in Google's announcement.

​Nicole’s LI post​

​Ross’s X post​

​Joe’s article​

Here’s what I think.

It makes sense to me that the entire SGE as we see today was never meant to be something that fully went into search. My speculation has always been that SGE as we see it in labs today was not the same as what we would eventually see when it goes live.

But will SGE go live?

The announcement says that SGE is a testbed for new AI ideas that will be introduced into Search more broadly over time.

I went back and read the pdf Google published when they first told us about SGE. They call it "A new way to search with generative AI," and say, "These models will enhance Google Search and revolutionize how people engage with information." This doesn't sound to me like an experiment that is not intended to go live.

As I shared in last week’s newsletter, I think there’s a good chance that over the next few years we see a transition where many people use Assistant more than they use what we’ll eventually refer to as “classic search.”

I also mentioned how Assistant seems to pick perhaps one SGE feature to show me when I ask it a question. I think Google is using SGE as a testbed for new features that not only could be added to search, but also those that could be used in Assistant.

I just asked my Assistant on my Pixel 8 phone some questions and each brought up a different kind of response, each of which could be a component we saw in SGE. As a reminder, on my phone's home screen, if I chose a voice search over typing, I always get an Assistant answer rather than classic search.

Sometimes Assistant gives me a single website as an answer. Note how the website featured here is one that we’d happily call helpful content - it is content from a site that has EEAT on the topic of drums. This answer is often all I need. I rarely find myself tapping the search button.

Sometimes we’ll get a carousel of sites. This is similar to the SGE feature that would show us a carousel of sites all with helpful content on a topic. There are 6 sites in this carousel. Again, each with a variety of experience on the topic: a home reno tv show website, Youtube (real life experience), Home Depot, and articles from sites with a reputation for gardening. This certainly looks SGE-like:

I do believe there are plenty of opportunities for smaller bloggers who create helpful content to be featured in these carousels, especially those who know what fresh and new questions their audience has.

If I search for something local, I get a maps pack. And only a maps pack. I have to hit the G in the bottom left if I want to see the rest of the search results.

Currently I cannot get SGE-like responses for eCommerce or other types of searches. I expect that over time, we’ll see Assistant evolve to show more AI related features in its answers. And of course, Bard is expected to be added to assistant any time now, which will make things even more interesting!

There are two main reasons why I feel we still need to pay attention to SGE.

First, Sundar Pichai told us in a recent earnings call that over time SGE will “just be how search works.” You may not use Assistant today, but if I'm right, when Bard, powered by Gemini Ultra (which we haven't seen yet) is added to Assistant, it will be useful enough to convert many to use it more and more often. I'm saying this, because I find this happening myself. When I want a quick answer, I'm thinking, "Oh, Assistant will get me exactly the page I need," and usually it does, without me needing to visit a full page of search results. I can't imagine how much more helpful this will be when I can converse with it. Well, actually I can, which is why I'm so excited about the future of AI use in search.

Second, Google’s goals that they have laid out for 2024 are all about delivering and improving upon AI. In the q3 earnings call Pichai said that AI was a foundational platform shift for Google, saying, "It starts with Search."

It is quite possible that we never do see the big full SGE monstrosity that we see in labs.google.com go into production. This means that those studies we have been sharing that talk about how much traffic is likely to be lost may not matter much.

This doesn't mean that SGE is something we can ignore though. Getting into the parts of SGE that recommend websites is something we are going to want to aim for.

I do expect that we will continue to see SGE components added to search and we will see more people using Assistant as it becomes more helpful. Assistant will likely use the components of SGE that Google's testing has shown as helpful.

Google confirmed a bug causing sites to drop in and out of search

Some sites were reporting dropping in and out of search rankings. Google said, We're aware of a very narrow issue that caused temporary fluctuations in search results for a small number of websites. The issue has since been resolved, and the sites should no longer be seeing its effects. This was an interesting issue because Barry had noted patterns consistent with an unannounced update happening over the weekends. These dates line up with the times when those sites dropped out of rankings.

​
​

SPONSORED by North Star Inbound We earned 154 links, including from CBS and Fox News local stations, for our report on the states with the most cheaters 1. We surveyed Americans on cheating behaviors. 2. We scored each state to determine a score out of 100. 3. From there, we analyzed the numbers by state and ranked them accordingly. Finding good ideas for a gaming client can be easy if you're willing to take a broader view. Here, we viewed cheating as a broader topic than one that relates only to game play. Check out our post to see how we finished this amazing campaign here. ​ Check out our work here​

Discussion on X re redirecting an HCU impacted site

This was an interesting discussion.

The site Ian is talking about was impacted by the helpful content system. They redirected the old site to the new and the new site is doing quite well.

He did not share the details of the redirect publicly, but in his private community. What is interesting to me is that this new site was able to rank. Google’s documentation says the classifier for the helpful content system “runs continuously,” allowing it to “monitor newly-launched sites and existing ones.”

It seems that monitoring newly-launched sites is not something that happens instantaneously. My bet is that this site will soon receive its classification, and given that the content is the same as the old site, it should drop out of meaningful rankings.

Rathnesh Kuman shared how he did something similar with redirecting an impacted site to a new one, saw nice improvements and then suddenly all was lost.

My assumption is that the redirects helped Google pass along the signals that were helping these sites rank. Then, the helpful content classifier, which runs continuously, eventually did its work, classifying the content once again as unhelpful.

More SEO news and tips

Shameem Adhikarath noticed a filter on mobile search that allows you to filter by Publisher. He searched for “SEO”. The dropdown suggested he could filter by Entrepreneur, SEL and SEJ. It then appears to do a site: search with that information.

Search Console now reports clicks and impressions differently for job listing pages. You might see more of them now.

Miriam Ellis noticed products ranking organically in a product block despite the fact she has not signed up for Merchant Center.

Speaking of Merchant Center, we are starting to have some good conversations on this topic in the Search Bar. Desmond gave me permission to share his strategy where he is aiming for consistency between the shopping feed data, schema data and website data. I love this idea:

​

Screaming Frog has a tip for preparing for the third party cookie phaseout in Chrome.

Ever since local SEOs teased out of Google that openness is a ranking factor, a lot of lawyers seem to be open 24 hours. I have not seen anyone report that simply changing your GBP hours helps with rankings. My suspicion is that Google can tell when businesses are open and when a searcher is looking for a business that is open.

Here’s a case where someone scraped a site and then started to outrank them. Danny Sullivan reached out in private and then boom, the spamming site stopped ranking. Except, Danny says Google did not take action here. Google’s automated systems did what they were built to do.

Google is making changes related to the Digital Markets Act (DMA) in Europe. Google says, “Over the coming weeks in Europe, we will be expanding our testing of a number of changes to the search results page. We will introduce dedicated units that include a group of links to comparison sites from across the web, and query shortcuts at the top of the search page to help people refine their search, including by focusing results just on comparison sites. For categories like hotels, we will also start testing a dedicated space for comparison sites and direct suppliers to show more detailed individual results including images, star ratings and more. These changes will result in the removal of some features from the search page, such as the Google Flights unit.”

AI changes for Google ads

I don’t generally write about ads, but I feel I should start paying more attention to what is happening here. Gemini is now powering something called the conversational experience. I got excited about this at first but it turned out to not be what I thought it was.

I was picturing a conversational experience as an ad that would encourage a searcher to have a conversation with your chatbot. For example, say you are doing a search, trying to figure out why your Google traffic is down. You might see an ad that says, “Talk to Marie Haynes’ AI Assistant about your traffic drop”.

Alas, this is not what conversational experience is. Rather, you are having a conversation with an AI assistant about creating your ads. This is still pretty cool though.

When my book is out, I’m going to experiment with using this tool to run some ads for it. It sounds like Google is trying to make it easier to create great ads even if you aren’t a PPC expert.

Interesting AI News

​
​Chrome got some more AI features. Tab groups let you organize tabs in groups that Chrome suggests. You can now customize your browser theme. And most interestingly, in the coming months you can use “help me write” to get AI’s help when filling out any form input on a website.

DeepMind made an LLM that trained itself by synthesizing millions of theorems and their proofs. Then they combined that with a system that can make decisions for challenging problems. Alpha Geometry learns & solves complex problems with no human input.

​Bard Advanced is expected to be live in early 2024. It will use Gemini Ultra, and give more complex and better responses. (Bard currently runs Gemini Pro, which is on par with GPT3.5). It does sound like this will be a paid product. I have more LLM tool subscriptions than video streaming ones. Does that make me a nerd?

You may soon be able to text with Bard in Android Messages. ngl this looks like something I will use.

Mark Zuckerberg says Meta’s long term goal is building general intelligence. They are working on building the best AI assistants. They’re currently training Llama 3. They have a LOT of AI computers (GPUs). He predicts people will soon commonly interact with the metaverse using glasses.

​Grok 1.5 is coming out next month. I don’t use Grok often, but do find myself occasionally turning to it. It hallucinates a lot. This seems to be par for the course for a new LLM. But they always improve. I am really interested in seeing how Grok improves as once you combine a good LLM with real time Twitter knowledge there is a lot of potential. Right now, it's quite good if you ignore the AI answer and just look at the tweets it surfaces.

Elon Musk says we have a silicon shortage now and we will have a voltage transformer shortage in a year and electricity shortages in general in 2 years.

A new paper claims that a novel method called “Binoculars” can accurately detect text detected by LLMs.

Recommended learning

Articles

​Google Organic Product Grids: How Ranking Works. Brodie Clark.

When a product grid appears they can take up almost half of the organic real estate of page one. Brodie says what’s important is ensuring your items are valid in Merchant Center alongside high quality information and images. High numbers of reviews help, especially for unique products sold on your site. This is a must read if you are in eCommerce.

​How Entities are Shaping the SERPs and What That Means for SEOs. Genie Jones. This looks really good as well.

Here’s a good read from Ethan Lazuk on Google’s spam problem. Ethan says that to him, SEO spam refers to “search engine-first content.”

And this article by Gianluca Fiorelli looks good too:Dear Warhammer, let's talk about SEO, JavaScrip, and website migrations.

Videos

I set up a computer in my kitchen and have started a new habit of watching AI videos while I cook.

Here’s a playlist of the most interesting videos I watched recently. I’ll continue to add to it.

Here's Mirella Lapata, a professor of NLP talking about AI. This was so enjoyable:

​

​

​

Research

​Introducing ASPIRE for selective prediction in LLMs. Google researchers are working on a framework that will allow LLMs to give answers with confidence scores. This will improve their reliability. This is a big deal!

​Search result ranking and presentation. This looks like an important paper by Google. It describes semantic search. Olaf Kopp has done a great job summarizing this paper. I expect I will be reading this quite a bit over the next few weeks.

If you are interested, here is a chat I had with ChatGPT on this paper. We talked about how a knowledge graph can help search return better results because it helps the system better recognize the entities and concepts in a query. The part of this paper that intrigued me the most was when they talked about how search engines modify our query to better match it with contextually relevant content. The conversation ends with some suggestions of other papers to read so I can learn more on this.

Marie’s Thoughts

Last week was a popular episode with several new folks signing up and lots of interesting discussion as well.

Here is another resource with a schema case study. They added author markup and built out more clear information on authors and much more. The results are impressive.

There's no vote for this week's Marie's Thought's topic because I found myself writing it while writing this newsletter. We are going to dig in deeper to learn more about information satisfaction signals and the role that the quality raters play.

The paid newsletter is $18/m. Each week I do a deep dive into a topic with a goal of learning something important to SEO and giving you something actionable.

​Read past episodes.

For paid subscribers that haven't yet logged in to my new system on the Search Bar.

Tools

​Seer’s Screaming Frog & Technical SEO companion GPT​

You can now advertise your tool in this section of newsletter to go alongside the tools I am recommending.

Jobs

​SEO Consultant ~ SearchLab Digital ~ $45k-$55k ~ Remote (US)​
​
​Generative Search Specialist ~ Golden Hippo ~ $100-$150k ~ Hybrid – Woodland Hills, CA (US)​
​
​SEO Account Manager ~ Clearscope ~$80k ~ Remote US​
​
​SEO Account Lead ~ Digital Position ~ $50k-$70k ~ Remote (US)​
​
​Director of SEO Editorial ~ Reach Plc ~ Remote (UK)​

Hope you're having a great week! I'll be frantically writing and working on GPTs. Please do leave a comment if you've got anything interesting to add re this episode.

Marie