AI Volatility
- Admin
- 3 days ago
- 3 min read
There’s increasing talk about how brands need to be prepared to ensure their content appears in the growing ecosystem of AI agents. At the same time, as always happens when a disruptive player emerges, there’s a wave of hype predicting the end of traditional search engines.
Let’s break it down:
1- AI will grow — that is guaranteed, but today AI agents represent just over 5% of searches worldwide.
2- If we already consider search engines a black box, imagine what AI agents are.
3- Just as we have robots.txt and sitemaps, in AI there is still no universal, official and massively adopted standard:
The proposals gaining the most traction are ai.txt and llms.txt (but there are more)
They are driven by companies and communities (each pulling in its own direction depending on its interests)
There is no consensus among the major AI players
4- What is working is a variant of robots.txt adapted to AI, which basically “blocks”, not provides “positive signals”.
5- It is important to understand that these variants only tell the different AIs what they can take and what they cannot, but not to come and prioritize our content.
And more importantly, we must change the “mindset” regarding how AI works — different LLMs look for content that answers prompts. This is no longer about proactive user searches, but about being able to provide answers. This is a radical change: brands must boost the creation of useful, relevant and differentiated/unique content that fits the questions/doubts users have.
At the same time, we must not forget that the company that owns the main search engine- Google also has its own interests in AI, and its current approach is to integrate everything. So whether we like it or not, we now have to do more to try to achieve the same.
And I say the same because the evidence we are seeing today is that organic results are suffering (increasingly less visibility) and consequently less organic traffic to websites.
But if that were not enough — more complexity, “black box” behavior, lack of standards… there is a variable that is increasingly being discussed: “AI volatility”, unpredictability of its reference sources.
This volatility also differs depending on the model. Semrush, through its AI VISIBILITY INDEX, has been monitoring this effect for some time:
If you want, we can complicate it even more, because this volatility varies by category- meaning that depending on the type of content, behavior differs quite a lot.
This volatility is highly unpredictable. In ChatGPT, changes over 3 months are much more drastic than in Google AI Mode.
If we keep digging, what we also see is that the sources most used to generate answers are highly concentrated and top players represent 75% of the sources used (very few actually make up the answers).
The TOP 15 brands/sources most used to generate answers are:
ChatGPT: Wikipedia, Forbes, Amazon, Reddit, Techradar, Nerdwallet, Investopedia, USNews, Tom’s Guide, Medium.
Google AI Mode: Amazon, Reddit, Youtube, Nerdwallet, Google, Bankrate, Investopedia, Bestbuy, Forbes, Yahoo
Among all these sources, it is worth highlighting how Reddit has been “penalized/corrected” in ChatGPT but has started to grow significantly in Google AI Mode.
Another relevant point is that Google, to slow down ChatGPT (and its crawling), removed the option to view 100 results per page and that caused the number of brands/sources used by ChatGPT to increase (they had to adapt).
Volatility exists, but it is also true that the volume of the most used sources is very stable — it is difficult to be considered a relevant source for AI models.
So yes, AI is going to change many things, especially behaviors, but for brands having presence is not trivial- I would say much more complex than in search engines and, above all, much more “opaque” today.
Comments