The year of generative AI


The year 2023 was unlike any other I’ve seen in SEO and Search. I say this as someone who has been in this industry since 2007. Many others I spoke to expressed similar sentiments at various points throughout 2023. 

By March, Search Engine Land had published multiple stories that, in any other year, all could have been the story of the year. 

And the news and updates kept coming, day after day, week after week and month after month. Until at last, here we are, with no more year left to go! Although, with how this year has gone, I may be tempting fate – it’s entirely possible that a major Search story will break on Dec. 31.

Search Engine Land has covered all the biggest stories for 17 years, and we did it again in 2023. 

Here’s our look back at the biggest SEO news and updates of 2023. 

Google

Search Generative Experience

The all-new, AI-powered Google Search — officially: Google Search Generative Experience — was announced May 10 after months of speculation and rumors. It was powered by multiple large-language models (LLMs), including PaLM2 and MUM. 

It came with a waitlist (a recurring theme in 2023). Google opened access May 25. Here’s our SGE hands-on and early reactions.

Dig deeper. Test driving Google’s Search Generative Experience

The AI-generated answers are presented in a variety of forms via a snapshot, with links, images, videos and the ability to ask follow-up questions. SGE has received countless feature updates and expanded to 120 more countries, all while its content formats continue to evolve

Initially, SGE failed to cite sources in its answers (like Bard). Google began testing SGE links before officially adding links in August.

Leading up to this:

Dig deeper. Google patent describes how the Search Generative Experience works

Bard

Google tried to make it clear — Bard is not Search. This didn’t stop people from confusing Bard with some AI features it teased at the same time (which we later learned was SGE). 

But the arrival of Bard — Google’s answer to ChatGPT — was huge news in the Search world. Bard is Google’s experimental conversational AI service, powered by LaMDA. 

Bard was upgraded to Google’s Gemini model in December. Assistant with Bard, a “personal assistant powered by generative AI,” will soon be added to Google Assistant on iOS and Android.

  • When Bard was introduced, there were no links or citations to the sources used to generate its AI answers, with some seeing this a declaration of war on publishers. Google’s initial explanation was that Bard was “intended to generate original content and not replicate existing content at length.” Citations were added later.
  • Google seemed to rush the Bard announcement (widely considered to be so underwhelming that Google lost $100 million in market value the following day) to Feb. 6 so Google could upstage a Feb. 7 event at which Microsoft announced its new AI-powered version of Bing Search, powered with GPT-4. 
  • A month after announcing Bard, Google opened a waitlist. SEOs who got early access shared early Bard issues, which included hallucinations and getting bad SEO advice that went against Google’s guidelines (e.g., Bard thought buying links is a good idea). Overall, SEOs weren’t impressed with Bard — Google Bard was called disappointing compared to ChatGPT and Bing Chat. 

Dig deeper: ChatGPT vs. Google Bard vs. Bing Chat: Which generative AI solution is best?

AI content

The arrival of generative AI led to brands, including BankRate, CNET and others, experiment with publishing AI-generated content, as we saw in January. This emergence of generative AI-written content reminded us of the old content farms wiped out by Google’s Panda updates.

Google seemed to change its stance on AI content this year, less than a year after warning against AI written content

Content that is helpful and created for people first (vs. solely for earning search rankings) was now OK, according to Google’s Danny Sullivan. Google reiterated its stance a month later, with Sullivan saying Google’s focus is “on the quality of content, rather than how content is produced.”

Meanwhile, content creators quickly became concerned about AI answers stealing traffic and revenue:

We also saw bad examples of AI content this year:

Ranking revelations

SEO wasn’t front and center at the U.S. vs. Google antitrust trial, but we learned a lot about how Google actually ranks pages. 

  • How Google Search and ranking works, according to Google’s Pandu Nayak: This is an absolute must-read. Learn how indexing, algorithms, deep learning systems, human raters, click and query data and more shape Google’s Search results, based on Nayak’s testimony.
  • 7 must-see Google Search ranking documents in antitrust trial exhibits: We learn a lot from internal presentations and documents, including Google’s pillars of ranking, what user interaction signals Google looks at (clicks, reads, scrolls, hovers); how Google learns from users and uses that data to improve Search; 18 aspects of search quality and much more.
  • Former Googler: Google ‘using clicks in rankings’: Eric Lehman, a former 17-year employee of Google, said during his testimony: “Pretty much everyone knows we’re using clicks in rankings.” When this was published, we didn’t yet have the full context (provided in the two stories above) around just how much click data was using due to less-than-stellar reporting from people who just don’t understand much about how search works.

Also this year (separate from the antitrust trial), Google’s Gary Illyes told us that links are no longer a “top 3” Google search ranking factor, which is in line with what Google said a year ago and told us would happen nearly a decade ago.

Links clearly still play a role in SEO. However, for Google, links are less important for ranking webpages than in years past.

Hidden gems, personal search and Notes

Google announced a trio of updates in November:

Algorithm updates

Although it felt like an incredibly volatile year, and we were warned to “buckle up” for more, Google only released nine algorithm updates this year – less than the 10 it has released the previous two years. You can read the annual recap of 2023 Google algorithm updates by Barry Schwartz.

Reminder: Our history of Google algorithm updates page features all the latest news and guidance around the latest algorithm updates.

Google shared new link best practices in their SEO and search developer documentation. 

This help document evolved from covering the basics of crawlable links to covering anchor text placements, how to write good anchor text, internal links and external links.

Content pruning

CNET got “exposed” for deleting thousands of pages (a.k.a., content pruning), which is a fairly common advanced SEO practice. CNET wrongly believed that content deprecation “sends a signal to Google that says CNET is fresh, relevant and worthy of being placed higher than our competitors in search results.”

However, Google’s Sullivan wanted to make Google’s stance on this tactic clear:

  • “Are you deleting content from your site because you somehow believe Google doesn’t like ‘old’ content? That’s not a thing! Our guidance doesn’t encourage this. Older content can still be helpful, too.”

Read all about it in Google warns against content pruning as CNET deletes thousands of pages as well as my follow-up guide, Improving or removing content for SEO: How to do it the right way.

In memoriam: Google Analytics UA

We knew the end of Universal Analytics (UA) was inevitable. Google published blog posts, sent us emails, posted reminders on social platforms and showed us an intrusive interstitial every time we logged in. Google even threatened to set up Google Analytics 4 for us if we didn’t

While it seemed everybody was talking about AI, GA4’s switch-or-else date — July 1 — eventually came. Despite all the advanced notice, marketers still felt unprepared. Our coverage:

We thought UA would stop processing data. It didn’t. UA properties kept processing data. Ten days later. A month later. Two months later

Search Engine Land’s UA property finally stopped processing data on Sept. 8 — 68 days past the date on which we were expecting, and told repeatedly, it would stop. I wonder if there are still any UA properties collecting data as we close out 2023?

10 more Google Search updates and changes

Dig deeper. Inside Google’s massive 2023 E-E-A-T Knowledge Graph update

Microsoft 

New Bing / Bing Chat / Bing Copilot

In January, we learned Microsoft was planning on adding ChatGPT features to Bing. By February, we learned it would be powered by GPT-4, (OpenAI released this model in March) and the new interface was spotted in the wild.

Microsoft revealed the new Bing at an event in February. Here’s our hands-on review from February. 

New Bing earned praise from SEOs (e.g., New Bing is mind-blowingly fast and better than I expected) despite seeming to have multiple personalities (or “confused”) early on and received multiple quality improvements since.

It also came with a waitlist and would only be open to Edge users on desktop initially. Over a million people signed up for the AI-powered Bing over the next 48 hours. 

Dig deeper. Microsoft explains how Bing AI Chat uses ChatGPT and Search with Prometheus

Microsoft’s AI-powered search earned much media attention and created the perception that the company might finally gain ground on its longtime rival, Google. But hype, as it often does, turns out not to be reality.

It looked like Bing made some small gains in search market share by March. We later learned that the new Bing attracted many new Edge users, who then chose Google for Search instead of Bing.

It became clear by May that Microsoft Bing had failed to gain market share. Even clearer after six months of the new Bing — Microsoft disputed the numbers but failed to provide any figures.

By the time Microsoft CEO Satya Nadella spoke at the U.S. vs. Google antitrust trial, he sounded like a defeated man, saying at one point:

  • “Yeah, I mean, look, that’s called exuberance of someone who has like 3% share, that maybe I’ll have 3.5% share.”

In November, Microsoft announced a rebranding of Bing Chat to Copilot. When that change will actually become visible remains to be seen, as Bing still refers to its chat experience from Search as “Bing Chat” or “Chat.”

Dig deeper:

Yandex 

A former Yandex employee allegedly leaked source code, part of which contained 1,922 search ranking factors. This was huge news when it broke, but has almost been forgotten now. 

It turned out that 1,922 figure was low — there were actually 17,854 Yandex ranking factors. See Michael King’s excellent analysis: Yandex scrapes Google and other SEO learnings from the source code leak.

Also, Russia’s largest search engine is now reportedly for sale.

Yahoo

Yahoo started dropping hints in January about its return to competing in the search space. In addition to hiring, a tweet promised Yahoo was going to make search “cool again.”

Now we know Yahoo’s new Search experience will start rolling out in the first few weeks of 2024, Brian Provost, SVP & GM, Yahoo, told me at SMX Next in November.

Neeva

The ad-free search engine, founded in 2019, shut down.

Search Engine Land

SearchBot

We turned Search Engine Land into a chatbot this year. Yes, we trained ChatGPT on our content so you can explore, experiment and learn more about search marketing.

Later in the year, SearchBot got a huge upgrade, including new personas and image generation.

Sign up here for free access.

SMX Advanced and Next

We ran two digital events this year – SMX Advanced in June and SMX Next in November. Both shows were packed full of actionable SEO tips and insights. 

Below are links to our coverage of some SEO session from Advanced:

You can expect to read lots of coverage of SEO sessions from SMX Next over the coming weeks on Search Engine Land.

Plus, congratulations to all the 2023 Search Engine Land Award winners.

Salary & Career Survey

Here’s what you told us:

20 years of Barry Schwartz

Search Engine Land’s own Barry Schwartz has now been covering all things search for 20 years — 17 of those here at Search Engine Land. JR Oaks did a fascinating breakdown of 20 years of search based on Search Engine Roundtable analytics data.

SEO in 2033

What’s next for SEO? More AI. We are only at the dawn of our generative AI journey and AI-driven Search as we enter 2024.

Pichai said this year Google Search will evolve substantively in next 10 years. And 2024 will be one year closer to the type of search that is more “personalized” (hello, Bard Assistant) and “ambiently available to users in radically different ways.” And he promises Google will get SGE right.

DeepMind co-founder Mustafa Suleyman echoed this, saying Google will look much different in 2033 – where conversation is the interface, not a Search box.

If you’ve never seen the 2013 movie “Her,” watch it. Or if you have seen it, watch it again. That could be the general direction Google is heading. The future of AI Search could be:

  • Virtually assisted.
  • Conversational.
  • Predictive.
  • Adaptive — to evolve and understand you.

We aren’t there yet. Learn all you can about how modern search engines work, generative AI, LLMs, retrieval augmented generation and generative engine optimization (GEO) in 2024.



Source link

Leave a Comment

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