There is a spirit of optimism in web search. Since Generative AI made its breakthrough with ChatGPT, Google’s competitors have sensed their chance to break the search monopoly. Can this succeed and how is Google positioning itself?
- What do we already know about Google’s AI Overviews?
- Which sources are included in the AI Overviews?
- Do all organic results have a chance of being AI source?
- Are all sources in the AI overviews equally important for Google?
- How many sources does Google actually link to in an AI overview?
- How common are AI overviews in search results now?
- Do AI reviews lead to fewer clicks on websites than before?
- And now? How to prepare for AI overviews in EU
For a long time, Google’s monopoly in Internet search seemed invincible – well over 90 percent market share in the western world. Control over relevant access technologies such as the Chrome browser or the Android mobile operating system, and lucrative deals with gatekeepers such as Apple and Firefox have nipped all attempts at competition in the bud over the last ten years.
Since OpenAI achieved its breakthrough in the general usability of Generative Artificial Intelligence (GenAI) with ChatGPT at the end of 2022, Google’s challengers have seen the cards reshuffled. Microsoft entered into a close cooperation with OpenAI early on and Microsoft CEO Satya Nadella is challenging Google: “I want people to know that we made them dance,” he said in an interview in reference to the AI capabilities of Bing and Google.
While Bing Search, enhanced with AI features, has not been able to meet the high expectations in recent months (Bing’s market share has remained largely stable), the next generation of AI search engines is just entering the market. ChatGPT Search was launched a few days ago and Perplexity has seen a significant increase in user interest in recent months.
What both search engines have in common is that they take a different approach to answering searches than Google’s well-known “10 blue links”. They use the strength of Large Language Models in processing language and formulate information from different sources to form an overall answer .
But is this really the future of web search? Forecasts from market research companies such as Gartner have often been wrong in the past. A good example – predictions about voice search.
On the other hand, there is currently a high level of dissatisfaction with the quality of Google’s search results, a general willingness to switch and a new technology that enables a leap in the user experience. Possible scenarios:
- Nothing happens. Similar to voice search, there are no significant changes in the demand for AI features in search and user behavior.
- ChatGPT or Perplexity become market leaders. Competitors can offer a significantly better search experience that users switch to en masse.
- Google slowly but steadily expands AI integration. User habits and technological advances are good enough to allow Google to defend its market position as it gradually integrates AI features.
From my perspective, scenario 3 is currently the most likely.
Even if Google does not currently seem particularly confident in the implementation of AI features, the search engine giant still has a big lead and will very likely know how to use this. So it’s a good time to take a closer look at Google’s AI Overviews.
What do we already know about Google’s AI Overviews?
Google is quite transparent about how the AI Overviews generally work. In contrast to the “normal” ChatGPT, the answers in these boxes are not generated completely from the entire AI corpus. The first step is a classic web search and the selection of trustworthy sources.
Only then does the AI language model come into play and formulate the contents of the AI overview on this basis. If you like, AI is currently mainly used as a front-end technology that supports the presentation of the results. Even if we can assume that Google will also rely on AI for the actual web search (i.e. ranking), this will already be the case and will have little impact on the user.
Which sources are included in the AI Overviews?
To verify this statement from Google, we first looked at how many of the sources from the AI Overviews actually appear in the traditional, organic web search results. Since Google has not yet rolled out the AI Overviews in Europe, all of the following evaluations refer to the feature integration in the United Kingdom. The results are clear:
- In 37% of cases, all URLs from the AI Overview appear exactly as they do in the Top 100 SERPs
- In 62% of cases, all domains from the AI Overview also appear in the Top 100 SERPs
- In 98% of cases, at least one URL from the AI Overview also appears in the Top 100 SERPs
Do all organic results have a chance of being AI source?
We wanted to find out whether Google considers all content from the organic index as a potential source for the AIO box or whether the search engine makes a pre-selection. To do this, we analysed how many different domains actually appear in the organic index, in the featured snippets and in the AI overviews:
- In the organic results of the UK SERPs we see 18,404,705 different domains.
- There are 491,606 different domains seen in the featured snippets .
- In the AI overviews, only 274,455 domains are currently cited and linked.
Here too, the conclusion is clear: Google is extremely selective when selecting sources for the AI reviews. Only top domains have a real chance.
Are all sources in the AI overviews equally important for Google?
The possibility of being named as a source in the AI overviews is one thing, but appearing there frequently is a completely different matter. Therefore, in the next step, we examined how often individual domains are actually named as a source by Google. We sorted all the domains found there in descending order according to the number of keywords for which they appear in the AI overviews. Here is a selection:
Pos. | Domain | Number of Keywords |
---|---|---|
1 | youtube.com | 329.924 |
2 | wikipedia.org | 329.276 |
3 | nih.gov | 191.096 |
4 | healthline.com | 159.688 |
5 | clevelandclinic.org | 147.102 |
6 | webmd.com | 119.280 |
7 | medicalnewstoday.com | 115.342 |
8 | study.com | 101.975 |
9 | mayoclinic.org | 101.970 |
10 | britannica.com | 86.084 |
... | ... | ... |
196 | texas.gov | 3.248 |
197 | spotify.com | 2.570 |
198 | netflix.com | 2.503 |
199 | ssa.gov | 1.945 |
200 | paypal.com | 1.791 |
The number of keywords drops massively between positions 1, 10 and 200. The air at the top is even thinner here than in the organic index – fewer domains receive additional visibility in the AI overviews. The competition will therefore be tougher than before.
How many sources does Google actually link to in an AI overview?
A big difference between featured snippets and AI overviews is that with featured snippets only one source is cited and linked, while in the AI overviews Google brings together several sources and links them accordingly. We wanted to know how many sources Google cites on average in an AI overview. Here is the result of our analysis:
It is easy to see that usually between five and six different sources are linked in the AI overviews. It is extremely rare for Google to link to just one source there, as is the case with the featured snippets. Ten or more sources are also the exception.
How common are AI overviews in search results now?
Since Google rolled out the AI reviews in the search results first in the USA and then in other countries, we have been measuring their distribution in SISTRIX. Unfortunately, there are no figures from the EU yet, as Google is still holding back on rolling out the AI reviews here. However, we do have current figures from UK search:
While growth was quite rapid in the first few weeks, it has recently slowed down somewhat and is currently stagnating at between three and four percent of keywords . We are keeping an eye on this and assume that the value will continue to rise. Why? Here is a comparison of the distribution of featured snippets and AI overviews:
It is clear that there is a connection: When Google displays more AI reviews, it comes at the expense of featured snippets. This is not surprising, as both SERP features serve the same search intent and AI Overviews are likely to deliver better results for both Google and searchers in most cases.
Do AI reviews lead to fewer clicks on websites than before?
The big question that concerns the entire SEO industry is, of course, what effect the AI overviews have on the number of visitors to your own website. In order to be able to evaluate this, we compared current click-through rate (CTR) data from UK search for featured snippets and the AIO feature. The CTR indicates what percentage of searchers click on a result in the search results. Here are the values for featured snippets:
And here is a direct comparison of the values for AI Overviews in UK search.
The data is surprising. Currently, the AI Overviews in position #1 are leading to more clicks than featured snippets. This is in line with what Google has communicated recently – but honestly, who would have trusted Google on this topic?
The fact that the data confirms Google here is, on the one hand, encouraging, but on the other hand it also shows that we in EU do not have to be afraid of the introduction of AI reviews – at least not in their current form.
And now? How to prepare for AI overviews in EU
Although Google has now rolled out the AI overviews in over 100 countries and for more than a billion people, the EU has so far been left out, presumably for competition law reasons. However, I assume that Google will not ignore the important European market forever and that we too will soon be able to enjoy these new SERP features. So how should you prepare for them ?
In order to be represented in the various AI searches, it is first important that your domain can be found at all – and for that, it must be indexable. While this is not a big problem with Google, as Google cleverly hides the crawling behind a user agent, the situation is different with competitors such as ChatGPT, Perplexity and also Bing: These providers are blocked in many robots.txt files. Even if the reasoning behind this is understandable, now is a good time to think about whether this is still the right path.
In my opinion, the first “E” in EEAT is becoming even more important: expertise . Since the breakthrough of AI, search engines can produce as much mediocrity as they want. What will continue to be scarce, and therefore in demand, is unique expert knowledge. Nobody needs a summary of five test results from the Internet. What is in demand, however, are your own authentic photos, videos, opinions, assessments and experiences.
Even if it almost sounds like a cliché by now, this recommendation also belongs here: Become a brand . In the age of AI search, links are increasingly being replaced by mentions in the semantic environment. However, it is no longer URLs that are mentioned, but brands, people and generally known entities. Building a well-known and popular brand is difficult, expensive and time-consuming, but in my view there is increasingly no alternative.