SEO News & Updates

What happened in SEO during Q2 2023

David Broderick
Last Updated: Jul. 24, 2023

If you’re out of the loop, it’s safe to say there’s plenty to catch up on from the last quarter…

Google has dragged us all kicking and screaming onto GA4, whether we liked it or not.

AI-generated search results have hit the SERPs… while OpenAI is facing a $3bn lawsuit for stealing “vast amounts” of personal information to train its artificial intelligence models.

And Mark Zuckerberg and Elon Musk have agreed to fight each other in a cage match (no, seriously).

We’ve been busy here at TTT too. We published 44 lessons in Q2, including:

To help you catch up on anything you missed, we’ve pulled together a refresher on the hottest takes and smartest content from the world of SEO from the last three months…

The most popular stories from our Rich Snippets newsletter

Get up to speed with the ten most popular links we shared in our weekly Rich Snippets newsletter during Q2 2023:

1. In times of E-E-A-T: Why simple SEO is no longer enough – oncrawl

Technical SEO can seem like it’s on the other end of the SEO-ocean from Google’s push for Experience, Expertise, Authority, and Trust.

E-E-A-T is an abstract concept. Like a mood board for “hey, let’s not ruin the internet.”

But fresh Google SEO Fundamentals have provided us with some interesting context: E-E-A-T isn’t a ranking factor. It’s a mix of them.

This is what makes Olaf Kopp‘s beautiful madness of E-E-A-T-meets-technical-diagrams so delightful.

How do all the red strings connect through the pipelines and services? What are the possible ranking factors? Where are the data sources?!

The diagrams, y’all. Delicious 🍽️

2. Entity SEO: The definitive guide – Search Engine Land

If you’ve pondered the implications of generative AI without a fact-based schema, then you’ll be very excited for this guide to entities from Timothy Warren and Andrew Ansley.

Entities are the secret sauce behind the infamous “things, not strings” announcement of Google’s Knowledge Graph. They’re the key to knowledge-based answers.

This expansive guide covers critical areas, including:

  • What an entity is and why it’s important
  • The history of semantic search
  • How to identify and use entities in the SERP
  • How to use entities to rank web content

It makes sense right? ChatGPT is spouting off ads in results like a pay-to-play Furby and Google is pressing SEOs to live, breath, and sweat E-E-A-T.

This article is densely-packed with information retrieval models, diagrams, and supporting links. If you want to conquer the world of E-E-A-T, start with this slice of entities.

3. Supercharging Search with generative AI – The Keyword

Google I/O has been an annual developer’s conference/nerd festival since 2012. The event was back in person this year, but with a limited audience and single day of sessions. Search was a notable casualty of the truncated agenda. 

Since 2018, the event has held at least two sessions designed to facilitate developer engagement. This year? 0. (Search did get a shoutout during the keynote but more on that in the next bit.) 

What was the show stealer this year? Yeah, you already knew. 

Here’s a recap of key takeaways from AI’s I/O takeover: Generative AI SERPs are coming.

Google gave us a first peak at Search Generative Experience (SGE). This above-the-fold behemoth block generates an AI-powered snapshot of key information to consider, with links to dig deeper. It’ll even provide you with follow up questions to tap, continuing the interaction in a conversational matter.

SGE functionality will of course cover shopping queries as well. These results will be powered by Google’s Shopping Graph. One area that GSE won’t touch? Health and medical content

No word yet on if GSC has data for clicks and impressions for links used in AI-generated answers.

Those interested in testing GSE can sign up for Search Labs now. Currently, that’s open in English for the US only, with more experiments rolling out in the coming weeks.

If the rather spartan details provided by Google aren’t enough for you, head over to Search Engine Land for Barry Schwartz’s in-depth breakdown of what we know about SGE so far.

4. A new way to search with generative AI: An overview of SGE – Google

We all knew generative AI was bounding itself way to search results. We weren’t expecting such a long pause between Google’s announcement of Search Generative Experiences and follow through with documentation. A new PDF entitled ‘Overview of Search Generative Experience’ was released weeks after Google I/O.

The doc has five key sections:

  1. What is Search Generative Experience
  2. How Search Generative Experience works
  3. Applying generative AI responsibly
  4. Known limitations
  5. Building the future of Search together

The ‘how’ section is the clear winner for most enticing. It gives us details on four established modes, what the UX will be like, and reveals not even GSE will be safe from ads.

The modes include:

  1. AI-powered snapshot appears to be the default result (“When appropriate, SGE will show an AI-powered snapshot”).This snapshot will include links that corroborate the information shown. There does not appear to be a limitation to which sites can appear as citations, but YMYL queries appear to have higher standards and require “reliable” sources.
  2. Conversational mode retains the context of the original query and reiterates the results based on interaction with the suggested follow up questions.
  3. Vertical experiences include shopping and local queries. Shopping results are built on Google’s Shopping Graph.
  4. Creativity mode allows users to generate AI content. Google is keeping the training wheels on for this rollout, saying: “users will notice constraints on creative uses to start, as we’ve intentionally placed a greater emphasis on safety and quality.”

User experience cues will let us know how the user got to the results and if they’re still in Kansas. We can expect “crafted call-outs” and highlighted states when the modes shift. The color of the blocks will change to “better reflect specific journey types” and the query’s search intent itself.

Other key takeaways from the doc include:

  • SGE is currently powered by a variety of LLMs, including an advanced version of MUM and PaLM2.
  • SGE’s LLM was trained on search-specific tasks. One boon of this specialized training is – fingers crossed – reduced hallucinations and inaccuracies.
  • SGE goes through adversarial testing in which the red team pretends to be the baddies and break it. Honestly, I’m delighted to learn this.
  • The system is designed to not generate a response when the query falls into a data void, involves health/medical information, or contains explicit or dangerous topics.
  • GSE won’t respond in the first person. In fact, it doesn’t have a persona at all. Instead, Google “fine-tuned the model to provide objective, neutral responses that are corroborated with web results.”
  • Being without a persona won’t stop slip ups. It was trained on human content and may present human-like emotions in the response.
  • Because they’re different systems, we’ll likely spot contradictions between the information in GSE and that presented in the rest of the SERP.

5. How to optimize content for Google Perspectives – Search Engine Land

Perspectives, which was announced at Google I/O, is a search functionality designed to lean into the “Experience” portion of E-E-A-T. In theory, valuing unique, fresh viewpoints diminishes the incentive for copying reviews from other sites or buying a few thousand genAI reviews on Fiverr.

The execution is still up in the air. Currently, Perspectives is only available for Top Stories.

But as soon as your C-Level gets bored waiting to hear about AI-content strategy, they’re probably going to ask you how to optimize for Perspectives.

Dan Taylor saw an opportunity to help you prepare with useful context, important factors, and example strategies for publishing and ecommerce sites.

6. Study: 31% of international websites contain hreflang errors – Search Engine Land

International SEO is a whole different beast – specifically, a hydra. It’s a multi-headed monster with distinct personas connected to one main body.

What’s stopping your international kaiju from dominating SERPs across the globe? Its Achilles’ heel – which happens to be holding each head to the body…

A teensy tiny meta tag called hreflang.

Dan Taylor teamed up with NerdyData to study 19,000 websites reliant on the bone feather. I won’t steal Dan’s thunder here, but dang – 3/10 hydras have at least one head eating another.

Sometimes there’s just a neck with some random letters on top.

Since conquering hreflang tags can feel like a Herculean task, Dan gives the errors context with simple explanations, examples, and links.

7. Misadventures in SGE with Lily Ray

Frankly we were shocked (shocked we say!) to hear SGE is a hot mess. After having her hands on it for just a week, the peerless Lily Ray had already managed to get it to:

Obviously this is far from SGE’s final form. And since Google is explicitly calling it an “experiment”, you can’t be mad when it plagiarizes you (it’s just an experiment, bro!).

I guess let’s see if Google irons these wrinkles out before it unleashes SGE on the public? 🤷

8. RIP metaverse – Business Insider

During all this AI buzz, no one seemed to notice that former tech darling Metaverse is dead. It was three years old…

9. 100 things we announced at I/O 2023 – The Keyword

In an effort to show that Google can cram just as many announcements into a downsized developer conference, a Keyword post listicles 100 things shared at the event. 

These include:

  • 1 new “About this image” tool that will tell users when it was first indexed, where it first appeared, where else it’s been seen online and if it shows up on fact-checking sites. (A similar AI-audio detection tool is in the works)
  • 1 new stat that people now use Lens for 12 billion visual searches a month – a 4x increase in just two years. (Still no access for SEOs to this data though).
  • 1 ranking change to prioritize Search results with “unique expertise and experience”.
  • 1 upcoming expertise showcase in SERPs with a Perspectives filter that will display content from social media, forum, and “video-sharing communities” (so… TikTok? Or a new YouTube Tok knockoff?)
  • 5 generative search experience hints of things to come.
  • 13 talking points about PaLM 2 and the revamped PaLM 2 powered BARD
  • 8 ways AI will invade Google Workspace. This will make walking out of an exhausting meeting where no one took notes or action items even more exciting.
  • 1 enhanced webcam/TV/holodeck that creates a 3D version of the person you’re talking to?
  • 2 AI functionalities for Maps that feel less impressive than the ability to finally delete recent searches directly from Maps.
  • 2 new Maps functionalities for developers, including a new coming-soon Maps API that lets developers integrate birds-eye videos of an area and Photorealistic 3D Tiles to create immersive maps on your site.
  • Magic Editor will make your photos perfect. Or terrible. Either way you’ll be able to change elements like making a perfectly blue sky or repositioning the subject of the photo.
  • 31 fancy Android/Pixel enhancements for that one friend who makes the text thread green.
  • An AI-powered notebook, but I don’t know if they mean like Google Cloud Platform’s Vetex or like my Rocketbook I already never use.
  • 2 new safety and security APIs including a safe browsing API and a timely application of Jigsaw’s LLM toxicity mitigating Perspective API.
  • 8 ways to restrict what information is collected and about who.
  • 3 new models for Google Cloud’s Vertex and an AI-powered collaborator called Duet.
  • 1 new A3 Virtual Machine (based on NVIDIA’s H100 GPU) to power Google Cloud’s AI-optimized infrastructure.
  • 1 new hands-free gaming mouse that uses facial movements. My overly expressive eyebrows got too tired reading the announcement to check out the open-source version on Github.
  • 5 enhancements for smart TVs, watches, and Chrome products.

10. BigQuery efficiency tips for Search Console bulk data exports – Google Search Central

If you got excited when Google shared that SEOs could finally binge their hearts out on unsampled data via BigQuery, your heart probably dropped when you heard the estimated costs of all your big, big dreams.

It happened to enough folks (or he got tired of being asked about it enough) that Daniel Waisberg created new documentation on being an economical evil genius with BigQuery.

To be fair, most of the folks sulking over dashed hopes had no idea that you can set billing alerts and limits or if they did, that those features are in Google Cloud Console under “Billing”.

This article could well be the key to unlocking your best unsampled life. It’s richly linked with sample code and clear alerts – like how the benign phrase “Expiration time” is less forgiving than you might realize.

The hottest topics in our Slack community

Get a glimpse at what the hottest topics from our exclusive Slack community were during Q2 2023:

1. When TTT became a GA4 support group…

We’re sure you’ll be shocked to hear that GA4 chaos dominated the chatter in our Slack Mastermind this past quarter. 

One of our members couldn’t get GA4 and Segment to play nice together. Luckily for them, TTT OG Aarne Salminen helped walk them through that particular headache.

Another member had a dataLayer added to a button and wanted to create an Event in GA4 based on when the dataLayer fires using GTM. Shoutout to Oleg Korneitchouk and Ryan Levander (who literally wrote the book on getting to grips with GA4), who were on hand to help them get that set up (down to giving a second opinion on code snippets) in a thread that racked up 25 replies.

2. TTT’s approach to client reporting

“What’s your approach for client reporting?”, asked one TTT member.

The consensus? GA4 exported to BigQuery and then visualized in Looker Studio.

Ryan Levander was back again to drop a masterclass on creating client dashboards that actually get used. He recommendations included:

  • Focusing on creating dashboards to use on client calls to talk them through your work, not for them to use on their own (since they never do).
  • Make your dashboards actionable in less than a minute.
  • Set up notifications when a metric is too low/high so you and the client are alerted when you need to do something.

Outside of Looker Studio, Brandyn Morelli recommended DataBox for reporting, while Eric Covino gave Semrush’s white label reporting a shoutout, saying:

Semrush has good looking reports you can whitelabel. You can add in custom text and headings as well (which display in the TOC) and send on whatever cadence works for you and the client.

You can integrate all the Semrush tools as well as GA, FB ads, Google ads, GSC as well as custom images (sometimes I take screenshots from local map tools and incorporate for local).

Check out Nick’s Semrush review for more information about the all-in-one tool.

3. The best Ads bidding strategy for a new campaign

When a member asked for recommendations for the best bidding strategy for a new campaign they were running for a HVAC company, they got plenty of food for thought:

  • Jonathan Gorham (who knows a thing or two about ads…) recommended ‘Maximum Conversions’ for generating local leads.
  • Eric Rohrback recommended making sure all conversion goals are set up first, then setting the bidding method on the campaign to one of the Click-based bidding to start collecting data. Once you get enough conversion data, he recommended switching to Maximize conversions.
  • David Carle said he prefers to start a new campaign on ECPC, in order to have a better control on his bids and adjust as necessary

4. Should you send Ads to a landing page or a website?

When a member asked the community’s thoughts on sending Google Ads to a website or a landing page built in the likes of Unbounce, the community was split.

Jordan Choo said: “Landing pages absolutely crush it for us. I am talking about double digit conversion rate from local service providers.”

Andrew Peluso echoed that: “Sending traffic to a website is likely to get you a 2% – 5% conversion rate. Sending traffic to a landing page makes a 10% – 30% CR possible.” He even recorded a Loom video walking through why dedicated landing pages are so effective.

However, Jonathan Gorham: “A Landing page = A landing page. There is no difference between building on your own site vs a 3rd party provider like Unbounce. The reason people see better results on Unbounce is because they don’t have the same level of CRO baked into the lander on there website. It doesn’t matter what tool you use to build the lander, what matters is your offer and CRO for that page.”

5. Do years of experience matter in SEO?

“Why does everyone boast about “x years of experience” in such an ever changing career field? Wouldn’t speed of learning and adaptability trump outdated experience?” asked a TTT member. 

Kal and Evgeni from Ehancv said: “Experience gives you some kind of assuredness in your actions. The more battered you are, the easier it will be for you to make choices. There’s also the issue with noise – and there’s a lot in this niche. With experience, you learn how to sift through it.”

And Ian added: “Like anything in SEO, it depends 😂

A long career of doing very shallow work for a revolving door of clients doesn’t really mean anything.

But there’s certainly something to be said for being around for a while. 

It’s… interesting watching folks on twitter talk about stuff like it’s new, when it’s actually been around for a good ten years. When you’re new, you just… don’t know what you don’t know, so a lot of folks have wildly misplaced confidence.

Another good example is some of the niche site folks on twitter – talking up an approach… and then ten months later noting that their site tanked in a recent update. Some relatively new folks are speaking with high confidence about stuff that other people have already tested and dismissed… but the confidence makes it seem legit.”

6. How to use ChatGPT to 10x your content output (while cutting your effort drastically)

After creating a course on conversion optimization and joining Ian for a Fireside Chat, the wonderful Wes McDowell popped into Slack to pull back the curtain on how he’s using ChatGPT to create content that’s actually worth reading/watching/listening to in a fraction of the time it used to take him.

He shared his seven step process for allowing ChatGPT to do what it does best, while allowing space for all the human elements that will make content read as ultimately human.

7. TTT’s agency owners’ net margin goal

TTT MVPs Morgan and Greg from Jolly SEO ran an impromptu survey of TTT’s agency owners’ net margin goal.

Here’s a peak at the results 👀

8. GA4 parody trailer

Plenty of TTTers appreciated this surprisingly accurate GA4 parody trailer that Seppo Puusa shared.

9. TTT’s favorite link building software

“Is Pitchbox still the best software for link building as an agency?”, asked a TTT member.

Long story short: yes, with an honorable mention for Respona.

10. How Brian Dean Grew Exploding Topics from 0 to 400,000 Monthly Visitors

After joining Matt for a Fireside chat in May, the brilliant Brian Dean shared the exact process he used to grow his SaaS startup, Exploding Topics, to 400,000 monthly visitors – as well as reveal his plan to reach 1M+ visitors per month 👀

That’s all folks!

Want to keep up to speed with all things SEO without having to wait until our Q4 update? Sign up to our weekly Rich Snippets newsletter and join Traffic Think Tank to get in on these discussions.

And if you missed them the first time round, don’t miss the insights from our quarterly highlights from Q1 2023.