This newsletter is summarised from our latest piece of research, AI in SEO Part One: Google AI Overviews.
Google’s AI Overviews (AIOs) are the hottest topic in search. They were rolled out in May 2024, after the experimental Search Generative Experience was released a year earlier.
Why are they such a big deal?
AIOs appear at the top of search results for certain search terms, taking up valuable real estate where the ‘traditional’ search results would appear. They look like this:
The searcher is now often presented with an AI generated summary, with a citation link, and more sources when they click Show more.
The appearance of AIOs has caused concern amongst people working in search because they can give answers to searches directly on the results, which could possibly lead to a decline in clicks. There has been much discussion on social media that they are driving what’s known as ‘zero click search.’
How often do they appear?
This is debatable depending on which study you look at. Some studies have reached percentages as high a 47%. However, Semrush Sensor, which has a very large dataset, shows visibility at about 15% for the last 30 days. Notably, it does appear their visibility is growing.
Their appearance is more common in certain search categories than others, with differing visibility for device.
AIO visibility across 14 different categories on Semrush Sensor.
So at least according to the Semrush Sensor data, AIOs are not shown anywhere near the majority of searches. In transactional categories like Shopping and Real Estate, they’re shown for fewer than 3% of searches. However, for more informational categories like Science and Health, they are shown over 20% of the time on both desktop and mobile.
15% of all searches doesn’t sound very much, but when you weigh up that 14 billion searches occur on Google every day, AIOs are showing something like 2.1 billion times a day. This is about 40x the total number of searches that occur on ChatGPT every day.
What do they appear for?
Mostly ‘informational’ searches. Research from Advanced Web Rankings (which also stated overall visibility at about 31% - double the Semrush Sensor data), shows higher visibility for informational and commercial searches.
Informational searches are often pretty lengthy, including words like ‘who, what, where, when’ and asking related questions, and don’t have any specific buying intent (like ‘what was the biggest dinosaur’). Commercial searches have top of funnel buying intent, e.g. ‘best laptops for under $500’.
Given the presented data, websites creating lots of informational and commercial content - ie publishers - are most likely to be affected by AIOs. However, the effects are not necessarily negative.
What’s the effect on clicks?
It’s hard to know exactly. Upon their release, Google claimed that AIOs, ‘...get more clicks than if the page had appeared as a traditional web listing for that query.’ This is plausible due to AIO’s prominence - particularly for the top citation.
However, it gets complicated by AIO’s impact on overall click through rate. February Seer Interactive provided an analysis of 10,000 keywords, which at face value indicate that CTR is diminished by AIOs.
But AIOs probably also drive more searches overall. A SparkToro/Datos study stated that Google searches grew by 21.64% in 2024. Typically, more searches could lower CTR, just as higher volumes of traffic generally mean lower engagement metrics. Paradoxically, overall clicks could have grown at the same time as click-through-rate dropping when AIO is shown.
How are they created?
AIOs rely on Google’s existing infrastructure for crawling and indexing web pages. Google’s algorithm assesses whether a query would benefit from a generative AI response, then selects a subset of the top search results and feeds them into a custom version of the Gemini Large Language Model (LLM).
The Gemini model synthesizes information from the retrieved sources, generating an overview response that aims to concisely answer the user’s question. This AI generated content is displayed at the top of the search results as an AIO, accompanied by citation link cards.
In essence, AIOs are an example of retrieval-augmented generation (RAG) in action, relying primarily on the search results (which are as current as Google’s standard index) but also able to call on a pre-trained language model if needed.
An October 2024 study of over 400,000 searches by SurferSEO found the following:
52% of sources cited by AIOs also rank in the ‘traditional’ top 10 search results.
AIOs mention an average of 5 sources per query.
AIOs are on average 157 words long, with the top range being about 330 words. The average character length is 984.
Roughly 70% of top 10 ranking domains have higher traffic and rank for more keywords than those cited in AIOs. Although we should caveat that with the first bullet.
What do website owners need to do to appear?
We don’t currently see any reason to deviate away from established SEO practices to appear in AIO. Turnover of results is also very high, research by Authoritas from February states that 70% of the pages cited change within two to three months, independent of organic ranking shifts.
There is nothing special for creators to do to be considered other than to follow our regular guidance for appearing in search, as covered in Google Search Essentials.
If you stared at LinkedIn or X for long enough (as we do), you could be in grave danger of being persuaded into a new paradigm of ‘zero click search’ or (worse still) believing the ‘AIpocalypse’ has occurred because of AI Overviews.
But the data suggests otherwise for now. They appear for a minority of searches, it is unclear to exactly what degree they effect overall clicks, and there is no real need to change tack to appear in them.
They are really an intermediary stage of how Google is evolving towards an AI future, rather than being the radical step change they are often made out to be.
However, we do feel change is in the air, and if you’re involved in digital marketing, then it’s worth monitoring the situation closely.
It's less 'this is a big deal now' to 'the early signs are that this is probably going to be a very big deal, so you need to be thinking about approaches that can mitigate it as much as possible when the time comes.
You can check out our latest research, AI in SEO Part One: Google AI Overviews for a more detailed view.
