Google has seemingly scaled back immediate penetration of AI Overviews since it recommended adding glue to pizza and other faux pas last month, causing a launch debacle. Now CEO Sundar Pichai is mooting a more cautious approach as it plugs its Gemini AI large language model into its search cash cow, re-engineering results in the process.

But make no mistake, “they are opening the floodgates little by little”, per Authoritas CEO, Laurence O’Toole. “It’s the future of search. It will be pervasive eventually and this is just the start point of generative content in the search results.”

As such, brands need to rapidly rethink SEO strategies – or risk getting pushed off the first page and plummeting click through rates.

Authoritas has been running the numbers on both publisher and brand search terms and how they compare with Google’s new AI-compiled amalgamated answers for the last year. Commissioned by publishers Daily Mail Group, Trinity Mirror and Hearst, Press Gazette in the UK as well as Mi3 locally – and spearheaded by Future Media's Ricky Sutton – Authoritas has recorded major variances over that time as Google tested ratios between organic search rankings versus serving AI-created overviews in response to queries. These vary from circa 90 per cent of results tested last year, to around 10 per cent in the last weeks, depending on the keywords and categories tracked, per O’Toole.

But he thinks it will trend towards the upper end of that range sooner rather than later – certainly over the next year. Any suggestion Alphabet is downsizing its ambition is “wishful thinking,” he added, “because I don't think Google can cede the technical leadership position in search to Microsoft and its relationship with ChatGPT”.

Sinking costs

Google’s shift from gatekeeper-organiser of internet content to aggregator-publisher in its own right means “the impact for publishers could become more and more profound,” warns O’Toole. (Those search impacts are detailed in Authoritas’ study by Future Media's Ricky Sutton here, with sites such as BBC News and the London Evening Standard respectively notching 37 per cent and 32 per cent falls in search visibility during the test period.)

Early results have publishers scrambling to work out how to reconfigure SEO strategies in order to regain top rankings and minimise revenue loss. For brands, the risk is likewise that a big chunk of the money sunk into SEO over the last couple of decades becomes similarly redundant (as it did when Facebook downgraded organic social ten years ago and watched revenues rocket).

That’s because Google is now serving additional context from different third party sources – “the Quoras, the Reddits, the third party review sites” – within its overview results for brand terms, per O’Toole. (The suggestion of adding glue to pizza to stop the cheese sliding off came from an 11-year-old joke Reddit post, the downside of letting LLMs decide what constitutes additional context and a broader range of sources.)

“So that is going to have an impact … But I think there is also an opportunity for brands.”

It just requires a tonne of work. So how should marketers prepare?

Research, then re-write

“I wouldn't be asking for more resource for writing content right now … I would be looking for some resource to properly research, understand and track this on a regular cadence, to be really on top of how this is going to impact me as Google rolls out – because it's still in its infancy,” said O’Toole.

“Google has said it is rolling out [AI Overview] worldwide, but it’s really [only affecting] logged-in users in the US right now, and a little bit in the UK. But over the next couple of months, this is going to be rolling out and different categories are going to be affected in different ways,” he said.

O’Toole thinks its unlikely that generative search results will not eventually be served across the board, i.e. to all search users, logged-in or otherwise.

Which means brands and their SEO partners need to start modelling now.

“Google Search Console tells you the most popular keywords that are generating traffic to your site today, and where you're in striking distance of getting more traffic. I'd be analysing those keywords and segmenting them in terms of the type of [AI compiled] content that Google's showing in the search results, and I would be looking at my category to understand where I'm impacted most based on where I'm currently performing in organic,” he added.

“If I'm number one in organic for a term that's got a decent search volume, and suddenly the AI Overviews come in and I'm not ranking and I'm pushed 1,000 pixels down the page [or off the page entirely], I know I'm going to lose track. So I would break that out into different opportunities,” said O’Toole.

“You really need to understand the relationship between the top organic rankings and the generative URLs, and whether there's a strong overlap. If there is, great. That means with your organic rankings today, you can manipulate generative results.”

Which means marketers and their SEO teams can breathe easy. For now.

“But if they don’t, and Google's showing a diverse set of URLs, either you need different content or different domains, [i.e.] to partner with the sites that are ranking in those … otherwise the chances are you will find it very difficult to rank.”

Paid search impacts?

Authoritas is primarily SEO-focused, but has observed some differences in paid search.

“In our analysis we've looked at the placement of paid results and you can see for both shopping and paid search – appearing across a broad range of keywords across categories – that in many examples, the AI Overviews are appearing above the ads,” said O’Toole.

Per Authoritas' May study, 77.1 per cent of adverts appeared below the AI Overviews. Even in the shopping category, traditional paid ads appeared below AI results 54.9 per cent of the time.

Where that ultimately ends up "remains to be seen," said O'Toole. "I think one thing is for sure, Google's not going to do anything that's going to be detrimental to shareholders. So it will be net [revenue] positive … and they have started to test ads in the AI Overviews … but I haven’t seen enough examples in the wild to gauge what they are planning.”

Testing times

Whatever those ad formats look like, O’Toole thinks brands and publishers are in for a bumpy ride as Google bids to reel-in AI rivals.

“I don't think people are properly prepared for what's coming … Google could roll this out to nine out of 10 queries, and it would have a huge impact … [Brands] have spent millions of dollars over the years getting these top organic search positions and then suddenly those organic positions have fallen well down the page,” he suggested.

“It may take Google months, maybe even a year or more, to reach that level of penetration, given the rocky start. But Google is going to integrate generative AI in search in many different ways. And this is just the start before multi-step reasoning, planning tools, multiple AI agents [creating a world of search within search].

“So brands should have a dedicated person looking at this – it’s such a big channel – and there really are opportunities to rank in areas where perhaps you found it difficult to rank before. If you understand the additional context Google’s looking for, then you can potentially focus on that kind of content,” added O’Toole.

“The only way you can do that right now is to collect and analyse that data and start to formulate a roadmap – but you need to be agile, because Google is chopping and changing constantly.”

He thinks it may end up that brands deploy AI and large language models to second-guess Google’s own LLM “and then use them to deliver that type of content.”