Search engines becoming summary engines
We should stop calling the Internet search engines, ‘search engines’ and rename them ‘approximation’ or ‘close enough’ or ‘not sure’ engines.
Thanks to the endless promotion of Large Language Models (LLMs) as Knowledge Models by their makers and AI prophets, the industry is slowly moving away from presenting facts and moving towards presenting (its own) opinion.
It started with OpenAI and its ChatGPT, which was able to answer questions using authoritative language, leaving no doubt about the correctness of the answer. Yours truly wrote in this post 'ChatGPT, another step away from the truth' about the major shortcomings of this approach. That was in 2023. Two years later, and everything I wrote still stands.
As a side note, about the same time I wrote 'Pause AI? Naïveté or stupidity?' describing an effort of people from all walks of life to suggest that we should stop AI development for a half a year and '... focus on making current AI systems safer, transparent, and trustworthy.' You might recognize one of the signatories of this document - Mr. Elon Musk. The one who at that time was already working on his own AI and is now suing OpenAI to prevent it from changing into a for-profit company. It's ironic and cute. It should serve as a reminder before you sign any well meaning open letters.
Not sure if the AI systems since then became more 'safe, transparent, and trustworthy.'
To keep the hype going, in 2024, OpenAI introduced ChatGPT Search which was basically a front end for Microsoft's Bing search engine. The search engine which most people have only heard of but have never even tried. That gave the illusion that any answer was properly augmented with up to date information. If only.
However, this illusion, seen as a market differentiator, made the other companies blink.
One of them, Perplexity, when asked about Hiswai, provides information from the company website. It also offers an answer to the question 'How does Hiswai personalize my dashboard'. The answer contains links to HUAWEI documentation about HUAWEI CLOUD Stack 8.1.1. For them it is POTAYTO POTAHTO.
Naturally, Microsoft is testing this too since it has preferential access to ChatGPT.
That forced Google to expose Gemini (its version of ChatGPT) in the search results. The outcome? Questionable. It clutters the result page with the snippets of potential answers. Each snippet also has a link to the actual content.That has two side effects. Many people don't bother clicking on any of the links because the answer sounds good enough.
How do I know? That answer was provided by Google with the following snippet 'Yes, there is evidence that Google's introduction of Gemini and its AI-powered features, like AI Overviews, has led to a decrease in traffic to websites from Google Search. This is because Gemini provides more comprehensive answers directly in the search results, reducing the need for users to click on external links.' Google then provides 4 articles in the sidebar. But who would click on them ...
In the meantime, a few more things happened.
Google was declared a monopoly for its dominance in the search engine market and the US Department of Justice is discussing potential remedies. One of them will be a ban on companies getting paid by Google for an exclusive placement of Google Search, namely Apple, which gets about $20 billion a year for that.
Apple has been playing with AI for some time. It tried to somehow use OpenAI a few years ago, but the outcomes are not even close to the level hoped for or what its customers would tolerate. To add to that, the latest news suggests that Apple is trying to implement a similar search augmented feature in its Safari browser. It is not known if Apple will use its own technology or partner with somebody else.
You might wonder about the 'other' search engines. Yes, there are a few with single or half digit market share, but they are mostly powered by Microsoft's Bing. And that's the other funny thing. On May 12, 2025, Microsoft announced that its Bing Search API will be retired by August 2025. Without any warning, it gave all its customers a 3 month window to change. But change to what?
I am glad you asked. You could change to Grounding with Bing Search which '... allows your Azure AI Agents to incorporate real-time public web data when generating responses. You need to create a Grounding with Bing Search resource, and then connect this resource to your Azure AI Agents. When a user sends a query, Azure AI Agents decide if Grounding with Bing Search should be leveraged or not. If so, it will leverage Bing to search over public web data and return relevant chunks. Lastly, Azure AI Agents will use returned chunks to generate a response.'
Here you have it. Any future search results by Bing will be decided by AI - what/if/how should they be returned to you.
This is the new recurrent pattern. With AI, we are inserting a new layer between us and the actual information. This new layer - which nobody knows how the inside of the black box works - will further obfuscate the source and accuracy of any answer provided by the search engines. Each of us will have to decide for themselves how much we trust this new tech. Otherwise the choice will be made for us.