When focusing on AI news and developments related to SEO, May 2024 was arguably one of the most eventful months in the last few years, particularly for Google – perhaps since the launch of ChatGPT 3.5 in November 2022.

In this month’s roundup, we cover everything that has happened, what we think and how to plan ahead to ensure the effectiveness of your marketing.

What is happening? 

AI Overviews rollout in the US

Google’s Search Generative Experience (SGE) went through a series of name changes in May. First, “AI Overview” (previously “AI overviews are experimental”) switched to “AI Answer,” before finally reverting back to “AI Overview”. The removal of the term “experimental” led some users to suspect Google’s growing confidence in the quality of the feature, but others saw the  change back to from “answer” as indication they are not fully comfortable with defining the response as concrete (perhaps partly due to reports of extreme misinformation in responses). 

Despite this, Google announced a full US-rollout of AI Overviews mid-month – with more countries following soon – claiming availability to over a billion users by the end of the year.

As with the announcement of Featured Snippets (that incidentally will not be phased out with the release of AI Overviews) online publishers fear that this new Search Engine Result Page (SERP) feature will decrease website traffic. Google countered this claiming that AI Overview links get more clicks than regular web listings, promising that “..as [they] expand this experience, [they’ll] continue to focus on sending valuable traffic to publishers and creators.” It’s worth noting though that AI Overview clicks and impressions won’t be distinguished from standard search in Google Search Console reports.

More AI-powered features

As well as the main rollout of the feature, Google has announced a number of enhancements in the pipeline:

Adjust your query: Soon, users will be able to adjust the AI Overview by selecting from various options such as providing a more simple response – or breaking the response down in more detail.

Multi-step reasoning and planning: One of the key aims of AI Overviews is to answer complex questions in one go. Google uses the example of using it to find the best yoga or pilates studios in Boston, with details on intro offers, and walking time from Beacon Hill –  a task that would usually require multiple searches. Similarly, Google has stated that AI Overviews will be useful for planning, using the example of creating a three-day meal plan with recipes, which can then be exported into Google Docs or Gmail.

AI-organised search results: Another upcoming feature is the promise of an SERP aimed at inspiring users, this will include diverse content categorised under unique AI-generated headlines. This will first be rolled out for dining/recipe queries, followed by movies, music, books, hotels, and shopping.

Beyond text queries: Finally, another exciting development will allow users to search using video combined with voice queries. For example, imagine filming a broken item and asking Google how to repair it. AI Overviews will then provide a step-by-step guide and the resources you need to complete the fix. 

While these announcements may sound exciting, the reality of drastic changes to search and Google’s products in general typically results in pushback from a large number of users and this was certainly the case with the rollout of AI Overviews. 

The response to AI Overviews so far

Immediately following the rollout, searchers were asking how they could turn off the new feature, not only due to further reports of severe errors in the provided information – and even dedicated X accounts created to compile the worst examples – but also because users do not necessarily want results derived from generative AI to be forced upon them without choice. It’s worth noting that Bing lets its users turn off Copilot responses simply by toggling an off switch within its search settings.

Of course it’s not just searchers who might want to control what information is presented to them – certain online publishers may also want to prevent their content from appearing in this new format, and while there are options available, they currently require a fairly blunt approach that initial tests reveal result in collateral damage to a publisher’s visible content in the SERPs. 

Interestingly, Google has now finally announced a feature that has been requested for many years – a pure ‘Web’ filter for its search results. Once applied, the filter only returns text-based web page links – so no AI Overviews, but also no Featured Snippets, Images, Videos or other rich media. 

As the month came to a close, Google clearly addressed the widespread backlash against AI Overviews and released a statement that seemed to simultaneously explain, acknowledge, and refute many of the accusations – along with a list of technical improvements made following on from vast amounts of user feedback. 

Google I/O and the announcement of  ‘Gems’

May’s significance was further increased by Google’s I/O 2024 event which showcased a wave of innovative developments.

Google compiled a list of 100 announcements made and, outside of the launch of AI Overviews, many of these were related to improvements to Gemini. One of the most interesting Gemini announcements was that subscribers to Gemini Advanced will soon be able to create ‘Gems’ which are customisable versions of Gemini, similar to OpenAI’s ‘GPTs’. Users just need to describe what they want their Gem to do and how it should respond, and Gemini will create the custom Gem that then can act as a personal expert on any topic – such as being a coding partner, for example.

The launch of GPT-4o 

Pressuring Google to do the same, OpenAI continues to heavily innovate at a rapid pace. Their latest flagship model, GPT-4o (the ‘o’ stands for ‘omni’), was announced with the ability to reason across audio, vision, and text in real time. This means it can accept any inputted combination of text, audio, image, and video and can generate any similar combination as an output.

Designed for more natural human-computer interation, GPT-4o shows significant improvements in vision and audio understanding compared to previous models. It also matches GPT-4 Turbo’s text and code performance, while running faster than its predecessor and costing 50% less in the API.

It was also revealed in May that users of the free version of ChatGPT would be able to access custom GPTs – an option that was previously only available to subscribers.

Not for the first time, Google attempted to pull the rug from under OpenAI’s feet by publishing a demo of the new Gemini video capabilities just 45 minutes before the announcement of GPT-4o. This backfired with most viewers deeming it fairly lacklustre compared to the impressive capabilities of GPT-4o.

Screaming Frog integrates with ChatGPT

Outside of news from the usual big three was a very promising announcement from the company behind one of the most popular website crawlers, Screaming Frog. The latest version offers custom JavaScript integration during crawls. This opens up a range of new possibilities including the option to communicate with Large Language Model APIs such as ChatGPTs and Geminis with potential use cases from the preset library of snippets ranging from using AI to generate alt text for images, analysing content sentiment or intent and even creating custom snippets. 

What we think

The constant innovation from OpenAI appears to be pressuring Google to accelerate its own AI dominance in search. Over the last 18 months, Google’s launches have felt rushed compared to their usual thorough planning and testing.

The rollout of AI Overviews arguably seems a little premature with errors and a lack of user control leading to backlash. The “Web Filter” option feels like a retrograde step –  perhaps a middle ground option should have been provided, at least until AI Overviews were more stable and error free. 

Searchers aside, online publishers’ concerns about decreased engagement likely hold merit (despite Google’s claims to the contrary). Again this aspect, along with user concern, seems to have been overlooked at launch.

Despite announcements of additional AI developments in quick succession, given the choice, searchers and publishers would likely prefer a pre-polished AI Overviews launch over these auxiliary updates. Google’s end-of-month statement arguably acknowledges this.

OpenAI plays a valuable role in ensuring Google does not have a monopoly on AI chatbots, but the pressure it applies may be pushing Google to release features prematurely, impacting user experience. That said, the AI race will not slow down, and so having innovative releases such as GPT-4o at least ensures some incredible capabilities are made available to the public that offer genuine improvements in how we search, work, and live. 

For SEOs, the integration of chatbots with tools like Screaming Frog is a welcome addition. At RocketMill, we’re already testing use cases that will benefit us, and most importantly, our clients. 

Planning ahead

With AI Overviews scheduled to reach the UK SERPs later this year, it’s important to analyse their live rollout in the US – something that we’re able to do already due to our work with US-based brands. 

That said, it’s paramount that we do not lose sight of what has, does, and will continue to work very effectively as part of more traditional SEO. AI Overviews absolutely cannot be ignored, but equally there are many aspects of improving online visibility that need to remain firmly in focus. 

A technically well-maintained website, coupled with content that provides genuine value to users will continue to win, regardless of how the information it provides is extracted and displayed to online searchers. 

At some point in 2024, it’s clear that SEO for UK sites will change radically, and we are already prepared, eager, and open to embrace this new direction.