Now that we’re nearly a month onwards from Google’s official launch of AI Overviews (at least, in the US – the rest of the world is still waiting), we’re getting more data about when these AI-generated text snippets appear and how they change the traffic that publishers get from Google.
When an AI Overview appears on a query, it takes up most of the space at the top of a Google search result. This will push the regular organic results down quite a bit, far enough that most users will have to scroll to see them.
We know from previous research done on click-through rates from search results that the further down a link sits on a Google results page, the fewer clicks it gets. The prominence of AI Overviews means that regular organic results below it are likely to suffer painful decreases in traffic, as much as a quarter or more of their existing organic Google clicks.
Related
Now, AI Overviews do show source links as well, which users could click on. This may alleviate some of the pain, were it not for the fact that many of the cited sources aren’t webpages that show in the top ten organic results. The source links in AI Overviews have much greater diversity, something Google has admitted themselves.
So websites that have worked hard to get their pages into Google’s top ten may see most of that effort wasted, as Google will now give lower-ranking pages the opportunity to get clicks from the AI Overview at the top of the result. That is, if users will even click on the source links in the AI Overview in the first place. With the depth and detail of content in the AI Overview, this seems unlikely.
The good news is that AI Overviews, at the moment, don’t appear all that often. The most recent data shows that, after an initial high following Google’s announcement, AI Overviews started showing much less frequently. On average, AI Overviews now appear for around 4% of the tested searches.
(Research commissioned for Press Gazette has found AI Overviews are currently offered for around 10% of leading news search terms in the US).
This seems to be a response to the huge backlash against these AI-generated texts, which regularly contained drastic inaccuracies and sometimes even downright dangerous advice. The glue on pizza example is probably the most notorious.
The criticism was so overwhelming that Google felt the need to publish a damage limitation post in which their VP of Search, Liz Reid, stressed that Google “worked quickly to address these issues, either through improvements to our algorithms or through established processes to remove responses that don’t comply with our policies.”
In that same piece, Reid said that Google aimed to avoid showing AI Overviews for ‘hard news’ topics, where accurate up-to-date information is crucial.
However, the low percentage of AI Overviews and their exclusion from news topics doesn’t mean we’re in the clear. In fact, I’d argue that the hullabaloo about AI Overviews detracts from the real problem publishers should be addressing with Google.
Algorithm updates are biggest threat to publishers
The biggest threat to a publishing site’s traffic is not a new search feature like AI Overviews. Nor is it the site’s own journalistic practices, or its approach to SEO.
Many publishers are, at the very least, doing the basics right: Publishing newsworthy content produced by real journalists according to proper editorial practices and applying a baseline of optimisation to these articles.
However, publishing sites can suddenly find their traffic drastically reduced through no fault of their own. The reason is that Google has rolled out another algorithm update, framed as an attempt to improve search results and reduce spam but in fact often penalising legitimate publishers and causing massive traffic losses.
For some sites, ending up on the wrong end of a Google algorithm update can mean seeing their search traffic halved or losing all of their Discover traffic. These are immensely painful moments for publishers. People lose their jobs when traffic – and, as a result, revenue – suddenly evaporates. Some publishing sites even have to close down entirely.
I regularly get asked to help such sites, to identify a root cause that may have tripped Google’s algorithmic failsafes and caused the site to be interpreted as a source of spam or ‘unhelpful content’.
In most cases, no such root cause exists. The site wasn’t doing anything ‘wrong’, as defined by Google’s woolly-worded guidelines. Yet for some reason, it fell on the wrong side of an algorithm change, as part of Google’s ongoing attempts to salvage its reputation as the world’s best search engine.
By their own admission, Google’s systems aren’t perfect and such collateral damage from an algorithm change is not unexpected. Google asked us to provide examples of sites we feel are unfairly affected via an online feedback form which was briefly advertised on that very authoritative and highly official channel known today as ‘X’. That form has since closed, you’ll not be surprised to hear.
AI Overviews could cause havoc
AI Overviews may yet cause havoc for publisher’s search traffic. Google has the ability to ramp up the appearance of AI Overviews and insert them into nearly every search result.
Users may even prefer these AI generated snippets rather than reading multiple websites to find the information they’re looking for, with all the usability challenges that those websites offer: Cookie consent banners, email signups, pervasive advertising, affiliate links, and endless streams of third party sponsored content.
Google’s reticence to push AI Overviews further shows that, for now, users are not yet sold on the feature. If and when that changes, we’ll quickly know.
But in the meantime, the real threat to websites’ traffic isn’t whatever new shiny toy Google chooses to incorporate into their search results. It’s the less visible, but much more impactful, tinkering that Google’s engineers do behind the scenes with their core algorithm changes.
This is less of a public lightning rod for criticism. After all, Google can easily hide behind their motivations to ‘improve search’ and ‘reduce spam’. If, in the pursuit of such lofty goals, a few websites have to suffer some unfortunate side-effects, well, that’s the cost of getting a better web.
But it shouldn’t be up to Google to decide which websites succeed and which fail. Especially if such decisions are made without proper human oversight. These algorithms are black boxes that punish (and rarely reward) websites on factors that not even Google’s own engineers fully understand.
Imagine living in an enclosure, trapped with countless of your peers and one wild unpredictable animal, which may just ignore you or may at any moment and without any warning decide to eat one of you. That’s basically the state of the web today.
At some point in time we’re going to have to decide to work together and get rid of the animal among us.
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog
