ChatGPT vs Google Search-the content publishers worry
When OpenAI released its chabot ChatGPT on November 30th 2022, the world went crazy as they used, interacted with this software that mimicked human like conversations and seemed to know everything. People shared statuses on their twitter threads, WhatsApp statuses, blogs about the results they got after prompting ChatGPT on various questions.
While people were enjoying this, OpenAI, a Microsoft backed up start up in Artificial Intelligence was riding on its newfound success, perhaps having launched the most successful tech product ever. Some users began speculating Google Search, famous for the Google Search results page was dead, Google has been killed and so forth. However, this article tries to explain why Google Search is not going away even in the dominance of AI chatbots like ChatGPT and will remain the preferred way to search for information online, for quite some time into the future.
How the internet is organized and the various types of websites and Google’s role on the internet
The internet is made up of tens of millions, if not hundreds of millions of websites. These websites provide diverse services ranging from static content e.g., targeted how to (DIY) articles, dynamic content, entertainment content, e-commerce, utility services etc. There needs to be a way for people to reach this websites on the internet –and that’s where Google comes in. You go to Google search box/bar in your browser-mostly all major browsers (Safari, Opera, Firefox, Google Chrome) have a quick link to Google, and type in a word or a phrase of what you want to access. Google then gives you a list of websites ranked from the highest relevance all the way to millions of results in decreasing relevance to your search term in a fraction of a second in a results page. This way, Google, the biggest and most robust search engine is the directory or nervous system of the internet for you to get served. In doubt? Let’s look at scenario one:
Scenario one-searching for utility sites: Let’s assume you doubt this, and I will ask you a question. Suppose you are looking for a way to make your ‘About’ section on your LinkedIn profile bold, how do you go about it? You go to Google and search phrases like ‘making about section bold linked in’, ‘how to make about me section bold linked in’etc. The result is that Google will give you a link of websites that can sort this issue. In this scenario you are looking for a site that offers this service and not looking for information/queries.
Conclusion: Searching for utility sites on the internet, Google wins, ChatGPT does not have internet connection to try analyzing these websites or cannot offer this service, but Google will direct you where to get the resource.
Scenario two-searching for website name/domain names: You know a specific site called boldmyaboutme.com but haven’t memorized the domain name/URL to access the website. What do you do? You search ‘boldmyaboutme’ and hit Google Search. Google in its first result will have the domain name that exactly matched that keyword, so Google is a gateway to reach unmemorized domain names .It will be ridiculous to ask ChatGPT ‘which websites can I use to bold my about me section on LinkedIn’ only to be slapped with a response that ‘I am a large language model with no access to the internet…’.
Conclusion: Google Search wins for this use case.
Scenario three-access to websites with a comprehensive curriculum vs response to only what you asked: Let’s say you want to learn something like rain forests, you will go to Google and search ‘rain forests’. Google will respond with a list of websites e.g., https://education.nationalgeographic.org/resource/rain-forest,that talk about the term rain forests. This are educative blogs/websites that will go from basic description of these term to advanced topics such as types of rainforests, benefits of rainforests etc. This type of information is static and ChatGPT can respond with text on this particular kind of queries. But these websites win over ChatGPT in that they have info beyond that the user was looking for. To put this into perspective, the user wanted to know about ‘rain forests’ but ended up learning more about various types of rain forests. The author has tackled that keyword extensively and the user gets more than they asked for by learning more about a topic they never knew beforehand existed. To get more feedback from ChatGPT you have to prompt more on a topic you might not even be aware the subtopics that exist under it as a beginner. This case is similar to a user looking for a Python programming tutorial. Google search results on top blogs offering this course are more effective that ChatGPT search. Reason is that if you are at beginner level, you know nothing about Python programming from syntax, variables, data types. The blog with this tutorial will feed you the knowledge from zero to hero as you progress reading their content. To achieve a similar goal with ChatGPT, you will have to prompt it every subtopic to get results back, a not so effective way of learning as a starter. Conclusion Google Search wins.
Scenario four-e commerce sites: You are looking for seat covers for your car from dealers/shops within your locality. You go to a search engine, Google or Bing and type ‘mercedes seat covers near me’ or ‘car seat covers in nairobi’ if you live in Nairobi city like me.The result is a list of websites whereby you go to each link and compare to get the best bargain for the best price. This kind of search is not possible to work with ChatGPT because it’s not connected to the internet to know the latest listings but Google will know the latest products added across e commerce sites in your area as pertains to what you are looking for. In addition, the decision of where to buy ,even after calling the various seat cover outlets near you is a decision you would like to make after seeing some images of the seat covers available on those companies from their respective websites.
Conclusion: In this use case, Google search wins.
Scenario five: Construction company, consultation services online (e.g. lawyers, online doctors)
You are searching for a good contractor within your city or country for your retirement home .Would you use ChatGPT or Google Search? Definitely, you will google terms like ‘home construction companies in kenya’ for me who lives in Kenya,’ affordable house construction in kenya’ etc. From the results you click on sample construction company websites that shows up, view their designs, ask for a quote, organize to meet one or two contractors for further discussions and hopefully close on a deal. In this scenario, you need diversity of choice and that’s why Google Search results is the best fit. For this scenario, ChatGPT will not give you authoritatively who the best contractors are as it does not access internet to fetch these listings, and this is a decision you would like to make after much consideration.
Conclusion: Google Search wins.
There are endless scenarios whereby Google search wins over results from ChatGPT.
Scenarios where ChatGPT wins over Google Search
Expert knowledge search and troubleshooting. -Let’s say you an civil engineer, biotechnology researcher, software engineer and working on an advanced problem about data structures using Java. You are stuck but have a hint but not sure how to implement the solution, you can prompt ChatGPT and get a solution faster compared to googling many phrases in search of an optimal way to reach your solution.
Simple scientific, mathematical formulae and facts check=There are formulae that don’t change or scientific laws that don’t change. ChatGPT output is impressively better than Google Search results as it tries to explain such in a clearer manner. It can also give examples on the fly, you can also twist the examples so that you better understand the concept clearer compared to a site giving similar output which mostly will not be interactive.
Simple DIY
Easy to accomplish tasks such as recipes, how to troubleshoot devices at home, questionnaire questions generation are tasks that ChatGPT can help you better than Google Search results. You can prompt further to modify the steps and you end up discovering a new meal.
Sources of content and what if ChatGPT and other generative AIs take over the internet?
End of humans generating new content
As mentioned in the introduction, those saying Google has met its match or killer are wrong. We need to understand that Google does not create the content on the internet. It just emerged as the strongest rule maker/referee to how access to information is done via its algorithms of which website owners and webmasters adhere to. These websites are written by humans and are the source of information on the web-easily accessible information for free. Most of these websites survive by displaying ads to pay for domain names, domain hosting and also the servers where they run. If generative AI takes over and people no longer click on search page results, these websites are going to lose traffic and chatbots are going to have a huge success at expense of these sites. This means the internet is going to experience a meltdown as content creators will no longer get traffic to their websites, which means no revenue. They then stop producing new knowledge, publishing stories and experiences and human interaction online, by far the biggest mode of people interacting dies. They switch off their websites, passive income is lost, they become jobless. Large content producing websites fire their writers in tens of thousands. Millions of domains a deregistered or not renewed, domain registrars lose clients and close business. Web development CMS like WordPress become obsolete as no one will need them. These websites also run on cloud servers like AWS, Microsoft Azure, Digital Ocean and the like, clients are lost and business impacted.
Impact on online advertising
With websites closing, there is nowhere to run online promotion ads for businesses like ecommerce websites, consultancy websites and the like. Imagine ChatGPT giving out responses with ads to websites. Yes, there will be clicks but remember most of these websites that are paying for advertisements also get tons of organic traffic that converts to sales, traffic acquired without purchase. If then chatbots are the only place to get traffic from, then online business becomes insanely expensive as running a paid campaign requires a budget, and for big success, a lot of budget. And since there is no mix of purchases from free organic traffic, that means financial bleeding for the few surviving websites that are willing to pay.
Google’s response to generative AI like ChatGPT
Google is a technology behemoth and the most advanced company in AI research and application globally. It has similar technology to ChatGPT namely PaLM and LaMDA which it can deploy to its billions of daily users to deal a blow to ChatGPT’s success or rather take away considerable market share from ChatGPT. But there’s is a catch. How to implement this without disrupting the egg that lays its golden egg-Google Search result page. The most probable way to approach this is have the first result being interactive in a conversational style similar to ChatGPT .This means there is still a higher chance a user will click the results below the AI response and websites that rank on page 1 for what the user searched still continue to get traffic. However, if this AI response is good enough, long enough and a user decides not to even look at the multiple results below the AI response, then websites will begin to experience reduced traffic.
The other approach is to have a separate tab for generative AI response type or rather ChatGPT type of interaction. This way, when a user intends to use the Google chatbot, they will activate this tab similar to the way a user can view results in Images, Videos, Books etc. tabs. This way the result page remains and content publishes get a shot at getting decent traffic.
Google for dynamic sites like ecommerce sites- this is a use case that remains to be seen how Google will approach it as we don’t expect to see chatbot response when googling to access an ecommerce website ,government services portal, movie streaming site, we will just want to get links to these resources, so chatbot response will not apply to all scenarios.
Google's chatbot capacity at scale
The compute costs of running queries against these chatbots is quite expensive. This was made clear by Sam Altman, the CEO of OpenAI who admitted to large server bills for their service. However, being a Microsoft backed venture, OpenAI enjoys large discounts on Azure cloud. However ,that does not mean misuse of the facility and access to ChatGPT is often throttled, meaning many are the times you will need to use and you get a message to wait or leave your email to get a notification when the service has capacity .However a paid version of ChatGPT is being rolled out which will allow users access without throttling issues. The subscribers will also help OpenAI provide ChatGPT to the rest of the users who have not subscribed by offsetting server costs from subscriptions.
With entry of Google into the game, this is probably going to create pricing wars for the paid services. Google have their own cloud platform ,GCP the infrastructure that Google runs its services such as Gmail,Youtube,Google Drive etc. Therefore,they have the infrastructure to support their chatbot at scale and will strive to have a higher uptime that ChatGPT which has throttling issues for the free version.They will also likely price their paid version of the chatbot lower than ChatGPT’s price to deter users from using the competitor’s version.
Also to note is that Google can slash a huge chunk of ChatGPT’s traffic ,simply because before visiting the https://chat.openai.com/ website where ChatGPT runs, most users pass through Google by searching ‘chatgpt’(Scenario two discussed above).Once they wake up and find a chatbot option on their browser, the journey stops there and they use the Google version.
Conclusion
Response by Google is eminent. It’s a matter of when and not if. But Google, which is seen a source of truth on many fronts due to its large index of the web, is cautious of this move. Google knows too well its seeders is the millions of websites on the internet and will try to achieve a balance of fairness when deploying AI tools that could impact the millions of its sources of information. But change is constant. And to whom much is given, much is expected, AI is a powerful tool and we as humans should tread carefully with this technology lest the consequences will be unpleasant. What we are seeing, is just the tip of the iceberg. We need ethics and control by regulation on how AI is applied in our lives.
