The Internet Marketing Driver

Archives for June 2023

People Also Search For, Or Do They Always? How Google Might Use A Trained Generative Model To Generate Query Variants For Search Features Like PASF, PAA and more [Patent]

By Leave a Comment

Google using a generative model to generate query variants for people also search for and people also ask

I was checking out some patents from Google the other day and surfaced an interesting one that was granted and published on May 30, 2023. It was titled “Generating Query Variants Using A Trained Generative Model” and it definitely piqued my curiosity. It was originally filed in 2018, but was granted in late May. And since I’m always interested in important SERP features like People Also Ask (PAA) and People Also Search For (PASF), I had to dig in.

Also, this is something I would have pinged the brilliant Bill Slawski about in the past. Unfortunately, Bill is not with us anymore. Going through the patent made me realize how much I miss Bill’s posts about patents and being able to DM him questions about his analysis. Losing Bill was definitely a massive loss for our industry. Anyway, without Bill to dig in the way he always would, I decided to start digging in myself. And I’m glad I did. It was super-interesting.

Leveraging Generative Models Using Neural Networks For SERP Features
Below I’ll cover how the patent describes using a trained generative model to generate query variants for SERP features like “People Also Search For”, “People Also Ask”, and maybe more. The patent mentions “People Also Search For”, but it’s not a stretch to believe the process could be used for PAA as well. I cover that in my analysis below.

It was fascinating to learn more about what Google is doing on this front (at least based on the patent). Like with any patent, we don’t know if Google has implemented this yet, or if they will, but it sure made sense based on what I was reading.

In addition, and I found this fascinating, the patent explained how Google could even generate query variants for novel queries (brand new), and long-tail queries where there isn’t much data available yet. And with 15% of all queries never seen by Google before, it would make sense to use an approach like for generating query variants. I’ll cover more about this soon.

Key points from the patent:
I think the best way to cover the patent is to bullet out some of the highlights. Below, I’ll cover several key points from the patent, which I hope you find interesting as well.

Generating Query Variants Using A Trained Generative Model
US 11663201 B2
Date Granted: May 30, 2023
Date Filed: April 27, 2018
Assignee Name: Google LLC

Diagram from a Google patent about using a generative model to generate query variants for PASF and PAA

1. Query variants can be generated at run-time utilizing a trained generative model based on tokens from the original queries and additional input features. I’ll cover more about the additional input features soon.

2. The system can generate query variants even when the model is not trained on that query. So it can generate variants for novel queries (brand new) or what Google calls “tail” queries where there isn’t a lot of data yet. I found that very interesting, especially since Google says 15% of queries have never been seen before. So the generative model can predict which query variants to generate even for low-threshold queries by using a neural network (with memory layers).

Google's generative model working for novel queries and long-tail queries.

3. The generative model can be trained based on submissions of previous queries by users. But the patent also explains that the query variant training data can also be based on query pairs that have clicks on the same documents. That makes sense and shows how user engagement can play a factor in what is generated by the model.

Google's generative model trained on query pairs that have clicks on the same document.

4. The patent also explains that the model can be trained as a multitask model to enable the generation of multiple types of query variants. So it’s a sophisticated system that can generate different types of query variants, including follow-up queries, generalization queries, canonicalization queries, language translation queries, entailment queries, and more.

Google's generative model can be trained as a multitask model to generate multiple types of query variants.

5. After the query variants are generated, they are scored by the model. The system provides response scores for each variant. And the system can grade those variants by checking for answers to those query variants. That can help the system detect “potentially fake” query variants. Very interesting…

Google's generative model scoring query variants to determine quality.

6. The patent goes on to explain that the system can return answers in addition to just query variants. For example, the system can return a search result (PAA anyone?), a knowledge graph entity, a null response (no answer), or even a prompt for clarification (with clarifying user interface input). That could be in the form of disambiguation chips we see when Google is looking for help from users when trying to understand what the user is looking for. Again, interesting.

Google's patent explains that the system can return answers in addition to just query variants.

7. The patent goes on to explain that the model can take more than just tokens from the query, including “additional input features”. Those input features could include location, a task the user is interested in or performing (like cooking, repairing a car, travel planning, etc.). It can also take into account weather and more. And the task could be based on stored calendar entries for the user, chat messages or other communications, past queries submitted by the user, etc. So the query variants could be based on personalization or current context.

Google's patent explains the model can take more than just tokens from the query, including “additional input features”.

8. The model can also generate variants of a query and advertisements or other content. So the model can not only generate query variants, but it can generate (or maybe retrieve) ads or other content that can be displayed in the SERPs. I think I have to go back through that section again, but that was interesting… :)

Google's generative model can generate variants of a query and advertisements or other content.

9. The patent also explains that there can be a number of generative models based on different attributes or tasks. So there can be specific models for various tasks like shopping, traveling to a location, etc.

Google's patent explains that there can be a number of generative models based on different attributes or tasks.

Summary: Generating variants for PASF and PAA can be more complicated and nuanced than some think.
I hope breaking down this patent a bit helped you understand how Google could use a trained generative model to generate query variants, or other content, that can be displayed in various SERP features. And this can happen for novel queries (new) and long-tail queries where there isn’t much data yet. In addition, there could be multiple models being used that focus on a specific discipline. And the results can be personalized as well (based on additional input features).

So, the next time you view “People Also Search For” or “People Also Ask” in the SERPs, know that a generative model might have been used to provide those query variants. And if personalized, then maybe those queries are specific to your case. Again, Google’s systems are much more sophisticated than some people think.

GG

Disavowing The Disavow Tool [Case Study] – How a site owner finally removed a disavow file with 15K+ domains, stopped continually disavowing links, and then surged back from the dead

By Leave a Comment

Google Disavow Case Study

There aren’t many topics in SEO as controversial as disavowing links. Ever since Google introduced the ability to disavow links, there has been a ton of confusion about how to use the disavow tool, what types of links should be disavowed, when to ignore the disavow tool altogether, and more. Then add third-party tools that evaluate links for site owners and flag some as “toxic”, and you have a dangerous recipe of confusion and fear. And that fear makes it easy for some site owners to spend a lot of time continually disavowing links, spinning their wheels, and without any way to know if it was actually helping them.

I cannot tell you how many companies have contacted me over the years explaining they take time every week or month to review their latest links via third-party tools and determine what to disavow. They are deathly afraid of some type of negative algorithmic action Google will take based on random, spammy links showing up in their link profiles. The fact of the matter is that every site builds random, junky, spammy links over time. It’s not unusual to see those random links show up in a link profile. Google has explained this as well.

Here is just one of John Mueller’s tweets where he explains this. I’ll cover more about Googler comments about the disavow tool soon.

It’s also important to note that many of those companies reaching out to me have never bought links, participated in any link schemes, etc. After I explain more about the disavow tool to those site owners, why it was created, when Google actually recommends to use it (hint, not often), my calls with those site owners often go eerily silent.

Note, I am NOT referring to sites that have set up unnatural links in the past, had manual actions for unnatural links, participated in link schemes, etc. If you have, then you should take care of that situation, which could involve using the disavow tool. For example, you should have those links removed, nofollowed, and if you can’t do that, then you can use the disavow tool. But the reason you would be doing this is to avoid a manual action based on setting up unnatural links and not because of some boogeyman algorithm that’s going to downgrade your site.

I have also been extremely vocal that I believe Google could remove the disavow tool from Search Console altogether, and maybe even this year. With SpamBrain now neutralizing unnatural links (as of the December 2022 Link Spam Update), I can’t imagine Google will need to provide a disavow tool for long (at least one that’s available to use any time you want). That’s just my opinion, but it does make a lot of sense. Here’s a video I recorded with Barry Schwartz where we cover the disavow tool and how it could eventually go away (at 11:46 in the video):

Nuking A (Large) Disavow File: A Case Study
I am writing this post because I have a great case study to share. And this case follows a number of other companies I have helped that have decided to nuke their disavow files after understanding how the tool is supposed to be used. Those companies haven’t seen any negative impact from nuking their disavow files. And for the case study I’m going to share today, you’ll see how a site completely removed a large disavow file with 15K+ domains in it and actually surged back from the dead after being down based on a terrible migration.

And I’m not saying the site surged back due to removing the disavow file. I’m simply saying the file didn’t matter at all. They surged back AFTER removing that giant disavow file filled with random, junky, spammy domains.

Note, this is a blinded case study, since everyone and their mother would over-analyze the situation if I revealed the site. But it’s too good of an example to sit in the G-Squared Interactive archives, locked away for only me to see. Everyone in the SEO community should read the case, speak with their clients about disavowing, and determine the best way forward. And for most sites (99.99%), that’s probably removing their disavow file and stopping the continual disavowing of links that don’t need to be disavowed. Again, that’s unless the site actively built unnatural links, participated in link schemes, had a manual action in the past, or think they could get one based on their own unnatural link building.

The Disavow Tool Is Buried In GSC:
I have been saying for a long time that most site owners never need to touch the disavow tool. From Google’s move to devalue unnatural links with Penguin 4 to SpamBrain now neutralizing unnatural links, Google is not penalizing random, junky, spammy links. Instead, it’s just ignoring them. Manual actions for unnatural links have dropped off a cliff over the past several years, which makes complete sense based on Google neutralizing random spammy links like I just mentioned.

Unless you actively set up unnatural links to try and game Google’s algorithms, then you should never, ever have to touch the disavow tool. That is one reason it’s literally buried in the GSC user interface. Seriously, try and go find it. It will probably take you a few minutes since you can’t access the disavow tool from any of the menus in GSC.

And think about it, if Google really wanted you to use the disavow tool all the time, why in the world would they bury it in Search Console? You can’t even search the help documentation from GSC anymore to quickly find it. You have to go the help docs, search for the disavow tool, and then scroll way down the page to find the text link to the tool. Again, it’s buried.

Disavow Tool in Google's help documentation

Googler Comments About The Disavow Tool:
In addition, there have been some great quotes from Googlers about the disavow tool over the years. From Gary Illyes explaining many sites end up hurting their efforts by using the disavow tool to John Mueller’s continuous communications that most site owners should never use the tool, it’s a surprise that some site owners still disavow links on a regular basis.

Here are some of Gary’s comments about the disavow tool (@methode is Gary’s Twitter handle). These are people quoting Gary based on Pubcon presentations, Q&A sessions Gary has done, etc.

And here is a great quote that always stood out to me. Gary even said that if it was up to him, he would remove the tool. Yes, a Googler saying he would remove the disavow tool…

And here is a quote from Gary about not being afraid of sites you don’t know and how he trusts the Google filters:

Gary Illyes on using the disavow tool.

Moving on, here are several tweets from Google’s John Mueller (there are more, but I can’t provide all of them here):

Even the disavow tool itself explains you should not use the tool unless you have a manual action, or you think you could get one. Google’s messaging there is pretty aggressive and scary. Let’s face it, they don’t want site owners using the tool for random, spammy links that show up in a link profile.

Disavow Tool messaging

But not everybody thinks that messaging is clear, including the site owner I helped. So, I asked the site owner if he wanted to provide a quote about the confusion he initially had about disavowing based on Google’s messaging in the tool and in the documentation. Like many site owners, he would like to see clearer messaging from Google about when to use the tool and when it’s necessary.

So here we go, in the site owner’s own words:

“Google documents about how, and when, to use the disavow tool are too open to interpretation, especially in situations after an (algorithm) update or when your rankings rapidly decline out of the blue. I hope others that read this case study will not interpret Google’s messaging like I did, and instead look for all other possible outlets before deciding to use the disavow tool. I very much hope this tool gets nuked!”

With that out of the way, I’ll cover the case study.

The Case: Down And Out, And Disavowing In Circles:
In the fall of 2022 I received a dire email from a site owner. The site changed domain names and it went very, very wrong. It really wasn’t their fault in my opinion… they were an edge case (which can happen with any migration). I’ve seen this a number of times over the years unfortunately, which is why I tell most site owners to NEVER change domain names unless it’s absolutely necessary. They were down about 70-80% and not coming back.

Here is trending when blending GSC click data from both domains via Looker Studio:

Google Search Console clicks trending after a botched domain name change.

After digging in a bit, I noticed they were disavowing links. Many of them. So I asked if they ever set up unnatural links, if they ever had a manual action in the past, if they participated in any link schemes, etc. The site owner emphatically explained they NEVER set up any unnatural links. They just noticed many random, junky links in their link profile and they were scared those links would negatively impact rankings.

So, I explained that most sites end up with random and junky links like that and there was really nothing to worry about on that front. But the site owner had read many posts explaining how dangerous those links were, how it could drag your site down over time, and that actively disavowing links was the way to go. When they contacted me, they had over 15K+ domains in their disavow file.

A huge disavow file with 15K+ domains.

After sending the site owner quote after quote from Googlers about the disavow file, they started to come around. They believed me, but were deathly afraid to nuke their disavow file. So they removed it for a bit. Nothing changed at all, but again, they were down about 70-80% from the domain name change that went sideways. So when nothing changed, they added the disavow back out of fear it would keep them down.

See how that works? The fear of some “boogeyman” algorithmic action led them to continually include a disavow file filled with random, junky domains. And they continued to spend time analyzing their link profile on a regular basis, and adding more and more domains to the disavow file over time. It was a maddening spiral of disavowing links. And I was determined to get the site owner out of the death spiral.

By the way, checking a number of the domains and links revealed some weren’t even indexed. And if the pages aren’t indexed, the links can’t hurt you anyway. I sent this information to the site owner as well.

Unnatural link not even indexed.

Business-wise, the site owner had to lay off most of his employees based on the domain name situation. It was sad to see… especially since if they would have contacted me before the domain name change, I would have told them to NOT change domain names unless absolutely needed. For their situation, it wasn’t absolutely needed. It was a nice-to-have thing they wanted to do. Bad move and they were paying a heavy price.

After analyzing the situation, and having seen edge case migrations like this before, I truly believed they needed a major algorithm update to roll out, which could bring a site quality re-evaluation. For example, a broad core update or reviews update could possibly help them surge back from the dead.

The December 2022 helpful content update (HCU) and product reviews update (PRU) rolled out and the site didn’t surge back, but I still had hope. I told them to sit tight and let’s see if another update could help them out… Then the March 2023 broad core update rolled out, and they still didn’t recover… But again, as of that time, the disavow file with 15K+ domains was still being used…

Pulling The Band-Aid Off. FINALLY Nuking The Disavow File:
The site owner was strong throughout this entire situation. He listened to my guidance, continued to improve the site the best he could with reduced staff, and had faith things could come back at some point. And as a last-ditch effort, they decided to completely remove the disavow file in late March. I was thrilled they made this decision. It was a long time coming…

So the disavow file was removed that contained 15K+ domains. Poof, it was gone. Now the site owner needed to continue to drive forward with the site, forget about disavowing, and just focus on improving the site as much as possible quality-wise.

Disavow file removed from Search Console.

And Along Came The April 2023 Reviews Update (With A BIG Tremor):
On April 12, 2023 Google rolled out the reviews update (RU), a major algorithm update that could impact any site with reviews or recommendations content. Google’s product reviews update (PRU) evolved to just the reviews update and it now evaluated more than just product reviews. It made the earth shake for many sites… and it came packed with a powerful tremor on 4/19. With that tremor, it looked to me like Google refreshed a site-level quality algorithm (or several).

Well, a funny thing happened with the site I’m covering in this case study. It began to surge with that tremor. And I mean REALLY surge. Rankings started coming back big-time for the site. And it was their most powerful queries returning from the dead. The site jumped 5, 10, and 20+ rankings for key terms. It was amazing to see.

The site is now up 140% based on the April reviews update, and that’s without a single disavowed link. And that’s down from 15K+ disavowed domains in the past. It’s a great example of why I believe a disavow file is NOT necessary for 99.99% of sites. They clearly didn’t need to be disavowing those links…

Surge in clicks based on the April reviews update tremor.

Here is search visibility surging for the site based on that tremor (Sistrix data):

Surge in visibility based on the April reviews update tremor.

And here is a snapshot of the site’s rankings surging with the April reviews update tremor. Over 31K keywords increased in position that now rank in the top 10. Some of those keywords weren’t even ranking in the top 100 before the April reviews update tremor:

Surge in rankings based on the April reviews update tremor.

And again, I’m not saying they surged back due to removing the disavow file. I’m just making the point that the disavow file wasn’t doing anything (in my opinion). They surged back from the dead without the file in place.

The site owner has been through so much with the drop based on the domain name change debacle, weird volatility over time that never turned out well in the long-term, etc., that they are still fearful this won’t last. We are now almost two full months out from the surge, and they still don’t feel comfortable. And I get it. When you’re an edge case, the battle scars remain. That said, it’s great to see the site back doing so well. Let’s hope things continue that way.

Key Points About Disavowing Links For Site Owners:
I’ll end this post with some key points for site owners that are actively disavowing links or thinking about disavowing. This is based on my experience helping many companies over time:

  • In my opinion most sites, and I mean 99.99%, don’t ever need to use the disavow tool.
  • The disavow tool is buried in the GSC UI for a reason. As Google would explain, that’s by design.
  • Several Googlers have explained that most site owners never need to use the disavow file. And that it should only be used if you have a manual action, or think you could get one based on buying links, participating in link schemes, etc. It’s not for random, spammy links that show up in your link profile.
  • Do not just start disavowing random, junky, spammy links that show up. Most sites have those types of links and Google is very good at ignoring them.
  • Do not simply look at third-party tools that flag links as “toxic” and think you need to move quickly to disavow those links. Google has addressed that as well, and said it’s a terrible idea to disavow links purely based on what some third-party tool claims is “toxic”. Sure, you can go analyze those links, but if you haven’t set up unnatural links in the past, then there’s no reason to worry about that.
  • When to use the disavow file: Now, if you did buy links, or participate in some type of link scheme, then you should look to remove those links or nofollow them. And if you can’t for some reason, then it’s fine to disavow them. Again, do not just start disavowing random, junky, spammy links you see showing up in a link profile. If you truly weren’t involved in setting those up, just ignore them and move on.

Summary – For this site owner, nuking a giant disavow file and stopping the disavow madness was the way forward.
I hope this case study helped you learn more about the disavow tool, Google’s advice about using it, and why most site owners never need to use it. Google has gotten very, very good at simply ignoring random, spammy links on the web and there’s no reason to start disavowing those links. So, if you haven’t set up unnatural links, paid for links, or participated in some type of link scheme, then you should step away from the disavow tool. Just continue to improve your site the best you can and avoid over-analyzing links. For the site owner I covered in this case study, that was the path forward.

GG