Around 93% of all web traffic is via a search engine - Google is king and the average CTR (clicks divided by views) for the first position on a Google search query is 19.3%.
I’m not a search specialist but it doesn’t take one to understand that ChatGPT is about to turn the whole search engine game upside down.
Will traditional search engines be relevant? Will SEO specialists have to quit their jobs and beg the AI overlords for a crumb of bread? Will GPT-5 turn into Skynet?
What is the future of search?
ChatGPT is estimated to have reached 100 million monthly active users in January, just two months after launch. It took TikTok about nine months to reach 100 million users and Instagram 2.5 years, according to data from Sensor Tower. That’s mind boggling growth.
Who’s to say GPT-5 powered ChatGPT/Bing or something even better powering Google search won’t be the future? I’m discarding Bard (Google’s chatbot) because Bard isn’t very good.
I’ve read a few articles that point out the deficiencies of ChatGPT and how search still has a place in a world where ChatGPT exists. This article talks about how ChatGPT hallucinates facts, how search is more for a guided experience and search is cheaper. My counter argument to that would be that all these flaws can and will be fixed. AI chatbots are getting cheaper, GPT-4 is much better than GPT-3, and LLMs (large language models) are being worked on to hallucinate less.
I’d even argue that in the future, most people will be ok with search not being a similar guided experience, not having to click through articles or search products in stores, not having to search for videos and instead just be presented the exact thing they were looking for.
Chatbots can be guided experiences as well - they’re all conversational and as you tweak your prompts (the same way we do multiple searches) you get the answer you’re looking for. They also might soon be able to embed content and thus replicate product experiences (such as buying stuff on Amazon), video websites (Youtube and Netflix content links) and search for music. What I mean is that chatbots could potentially take over all search (Google, Youtube, Netflix, Spotify, etc.) and do it better.
Just imagine a future where you go on ChatSkynet and ask it to recommend a review on the flaws of an iPhone 18. ChatSkynet gives you a list of flaws on the iPhone 18 summarized into bullet points but also recommends a video from a prominent tech youtuber highlighting the part with the flaws, a few websites that detail the flaws of the iPhone 18 and it also links to a deal on the Samsung S33 on Amazon. Search will be replaced by LLMs - it’s just a matter of when.
Search will be replaced by LLMs - it’s just a matter of when.
The future of search could go in a few ways…
Traditional search with LLM enhanced features
This seems very unlikely although Google would probably want that to be reality. The way traditional search could be enhanced by LLM features would be search that knows exactly what you’re looking for - search that is able to summarize and understand context and then do the same summarization and pattern matching on websites to give you the most relevant websites for your search. It could give one line summaries for those websites and their content so you know exactly what you’re getting.
Google might actually try this as a last resort to keep traditional search alive- they might even go one step further and add search summaries via LLMs.
It’s more likely that people are lazy and want LLMs to summarize multiple websites worth of content and give them the bits they need. Who likes to do research after all?
AI-chatbot that learns fast and has up-to-date data
An AI chatbot that is trained on new content regularly and has the ability to search would require a completely different revenue stream.
Imagine a scenario where I search “top destinations in France this summer” - the chatbot could look up weather conditions in summer in France, find articles of possible destinations, maybe even look through youtube video scripts to give me a curated list of top destinations in France this summer. Or it could also just use past data it was trained on to suggest some destinations. Both ways, content will be ripped off other sites and put on a singular chatbot platform. No (or negligible) traffic goes to those sites it was trained on or looked up.
Now, websites could pay the chatbot firm to be prioritised first and to feature a link to their site. Most people would read the summary answer and a smaller proportion would want to explore more on the original site. Under the status quo - the top ranked site gets 19.3% of the people that view that link to click on it. Under the new chatbot regime, they’d get a much lower figure.
It’s pretty much a massive loss for any website that isn’t the dominating AI chatbot (my bet is on Google but maybe that’s for another article). The chatbot however gets so much more traffic and which chatbot firm wouldn’t like that?
Maybe websites will make it harder to scrape their data (why try to rank higher if there’s nothing to rank), maybe they’ll charge these companies to scrape their data for training or maybe GPT-10 will tell exactly what bit of information from training data was most relevant to the user query and give some revenue back to that site that helped it generate the answer. The latter makes for a much better ecosystem.
Maybe GPT-10 will tell exactly what bit of information from training data was most relevant to the user query and give some revenue back to the site that helped it generate the answer.